Below every paper are TOP 100 most-occuring words in that paper and their color is based on LDA topic model with k = 7.
(It looks like 0 = ?, 1 = ?, 2 = ?, 3 = ?, 4 = ?, 5 = ?, 6 = ? etc.)
Toggle LDA topics to sort by:
TOPIC0
TOPIC1
TOPIC2
TOPIC3
TOPIC4
TOPIC5
TOPIC6
Dimensionality Reduction of Massive Sparse Datasets Using Coresets
Dan Feldman, Mikhail Volkov, Daniela Rus
Dan Feldman, Mikhail Volkov, Daniela Rus
[weighted, sum, subset, whose, thus, complete, acm] [algorithm, theorem, every, proof, let, show, set, prove, problem, general, case, cardinality, since] [approximation, size, reduction, large, compute, computing, sampling, supplementary, latent, efficient, log, end] [matrix, sparse, can, error, wikipedia, rank, computation, first, vector, svd, analysis, small, see, low, running, vit, row, pca, symposium, solution, subspace, main, relative, paper, sparsity, synthetic, sketch, eps] [coreset, dimensionality, coresets, data, random, squared, section, independent, given, result, based, project, practical] [time, research, english] [input, using, used, weight, approach, use, original, different, please, recent]
Nearly Isometric Embedding by Relaxation
James McQueen, Marina Meila, Dominique Joncas
James McQueen, Marina Meila, Dominique Joncas
[graph, average, find] [loss, algorithm, set, function, learning, will, let, convex, choose, define, case, obtain, every] [gradient, compute, smooth, method, convergence, operator, initial, step, dual, coordinate, size, requires, machine, computed, end, subsample] [dimension, can, noise, matrix, subspace, principal, one, proposition, isometry, norm, denote, rank, spectral, symmetric, existing, via, order, low, noisy, linear] [data, embedding, metric, riemannian, point, manifold, distortion, laplacian, space, isomap, embeddings, lossk, isometric, given, pushforward, measure, mvu, journal, kernel, hlle, euclidean, sphere, sample, tangent, rincipal, hourglass, distance, based, geometric, dimensionality, null, estimator, embedded, definite] [intrinsic, optimizing, along] [using, propose, use, output, figure]
Deep Submodular Functions: Definitions and Learning
Brian W. Dolhansky, Jeff A. Bilmes
Brian W. Dolhansky, Jeff A. Bilmes
[partition, represent, cycle, strictly, many, graph, represented, include, possible, normalized, decomposable, number, called] [submodular, function, set, learning, matroid, dsfs, concave, dsf, modular, laminar, may, show, monotone, since, defined, general, might, theorem, greedy, antitone, scms, matroids, arbitrary, define, scmms, class, maximizing, even, maximization, summarization, context, lemma, submodularity] [machine, learnt, size, large, optimization] [can, rank, one, via, also, linear, analysis, require] [given, family, useful, data, based, result, section, associated] [form, within, extend, showing, value] [feature, ground, used, learn, approach, deep, figure, generalize, neural, layer, training, use, using, trained, image]
Causal meets Submodular: Subset Selection with Directed Information
Yuxun Zhou, Costas J. Spanos
Yuxun Zhou, Costas J. Spanos
[causal, directed, subset, structure, definition, monotonic, larger, covariate, normalized, theory, degree, denoted] [greedy, function, submodular, smi, set, submodularity, algorithm, bound, guarantee, cardinality, problem, index, learning, general, maximizing, show, causally, defined, lemma, causality, optimal, dependence, since, theorem, maximization, lower, placement, close, provide, class, possibly, every, near] [objective, constrained, derivative, although] [can, also, analysis, first, one, much, solution, note, second, proposition, reconstructed, ieee, local] [random, selection, two, theoretical, data, covariates, given, selected, universal, journal, provides, measure] [information, sensor, location, research, process, value, series] [performance, figure, network, conditioned, work, used, proposed, better]
A Consistent Regularization Approach for Structured Prediction
Carlo Ciliberto, Lorenzo Rosasco, Alessandro Rudi
Carlo Ciliberto, Lorenzo Rosasco, Alessandro Rudi
[consistency, number, induced, probability] [loss, surrogate, learning, problem, least, considered, set, algorithm, case, prove, risk, let, satisfy, function, inequality, since, conference, consider, general, minimizer, binary, lemma, ask] [argmin, large, processing, machine, step] [can, regression, comparison, following, rank, ranking, note, linear, paper, decoding, sampled, solution, analysis, regularization, also, vector] [kernel, estimator, generalization, given, statistical, universal, reproducing, space, robust, finite, derivation, gaussian, corresponding, infinite, particular, kde, empirical, result, hellinger, hilbert] [whether, information, framework] [structured, prediction, approach, proposed, training, work, using, natural, question, neural, classifier, image, output, input, classification, reconstruction, compact]
Bayesian latent structure discovery from multi-neuron recordings
Scott Linderman, Ryan P. Adams, Jonathan W. Pillow
Scott Linderman, Ryan P. Adams, Jonathan W. Pillow
[structure, adjacency, block, number, probability] [function, binary, bernoulli, consider, since, algorithm, provide] [latent, bayesian, inference, glm, auxiliary, stochastic, sampling, standard, likelihood, posterior, fit, efficient, logistic, scalability] [can, matrix, linear, synthetic, correlated, one, via, also] [true, distribution, gaussian, distance, data, given, discrete, underlying, random, conditional, gibbs, population, mean, collapsed, functional, journal, joint, conditionally, independent] [spike, model, cell, time, neuron, retinal, ganglion, inferred, binomial, spiking, activity, response, neuronal, modeling, interpretable] [neural, network, connection, activation, table, approach, shown, figure, augmentation, using, receptive, weight, like, capture]
Lifelong Learning with Weighted Majority Votes
Anastasia Pentina, Ruth Urner
Anastasia Pentina, Ruth Urner
[majority, base, weighted, number, probability, theory, vote, total] [learning, set, algorithm, lifelong, complexity, hypothesis, every, class, theorem, conference, will, lsi, encountered, learner, setting, upper, show, assume, let, exists, bounded, consider, implies, max, formulate, case] [machine, solved, marginal, draw, end] [can, error, linear, one, note, assumption, first, dimension, arg, observed, related, solving, small, also, analysis, need, following, min] [sample, section, data, well, distribution] [information, new, previously, current, model, therefore, autonomous, required] [task, learned, used, training, sequence, ground, previous, neural, transfer, representation, classifier, work, international, performance, relatedness, using, feature, similar, prediction]
Global Optimality of Local Search for Low Rank Matrix Recovery
Srinadh Bhojanapalli, Behnam Neyshabur, Nati Srebro
Srinadh Bhojanapalli, Behnam Neyshabur, Nati Srebro
[number, present, probability] [problem, theorem, consider, case, lemma, proof, close, satisfies, learning, show, conference, convex, let, even, now, general, algorithm, property, best] [optimization, gradient, convergence, descent, saddle, operator, machine, showed, efficient, stochastic] [local, matrix, global, rank, first, order, can, also, recovery, noiseless, spurious, condition, linear, initialization, low, noisy, orthonormal, optimality, measurement, second, minimum, isometry, optimum, rip, via, following, svd, sensing, error, success, semidefinite, require, hai, restricted] [random, point, result, gaussian, given, section, stationary, space, well] [search, information, significant] [work, using, preprint, used, similar, arxiv, recent, neural, international]
More Supervision, Less Computation: Statistical-Computational Tradeoffs in Weakly Supervised Learning
Xinyang Yi, Zhaoran Wang, Zhuoran Yang, Constantine Caramanis, Han Liu
Xinyang Yi, Zhaoran Wang, Zhuoran Yang, Constantine Caramanis, Han Liu
[detection, definition, number, thus, probability] [query, problem, oracle, minimax, bound, function, lower, sup, consider, algorithm, theorem, learning, defined, label, rate, asymptotically, weakly, hypothesis, setting, constant, define, binary, uncorrupted, assume, risk, complexity, focus, let, set, exists, known, optimal, absolute, upper] [log, parameter, tractable, efficient] [computational, sparse, denote, high, gap, first, matrix, phase, analysis, satisfying, via, detecting, principal, condition, observed, also] [statistical, test, testing, random, two, sample, computationally, estimation, given, dimensional, data, covariance, polynomial, efficiency, distribution] [model, information, characterize] [supervised, sequence, classification, arxiv, preprint, compared, use, unsupervised, accuracy, work, supervision]
Disease Trajectory Maps
Peter Schulam, Raman Arora
Peter Schulam, Raman Arora
[number, probabilistic, include, many, obtained, clustering] [lower, learning, let, may, known, study, set, make, depends, focus] [variational, inducing, log, posterior, inference, approximate, bayesian, compute, stochastic, respect] [can, sparse, denote, vector, also, matrix, principal, see, first, sampled, analysis, column] [clinical, data, gaussian, basis, two, functional, distribution, longitudinal, given, covariance, test, mean, important, kernel, space, multivariate] [disease, dtm, trajectory, time, model, lmm, process, across, scleroderma, pfvc, tss, fpca, pulmonary, series, lung, skin, ini, complex, observation, prior, new] [using, use, representation, similar, figure, marker, learned, used, learn, capture]
Edge-exchangeable graphs and sparsity
Diana Cai, Trevor Campbell, Tamara Broderick
Diana Cai, Trevor Campbell, Tamara Broderick
[graph, edge, number, exchangeability, exchangeable, vertex, probability, many, kallenberg, fox, caron, collection, multiplicity, crane, greater, projective, growing, infinity, grows, added, multigraph, according] [binary, set, consider, active, let, every, since, setting, may, notion, will, theorem, function, drawn, slope, rate] [step, latent, bayesian, beta, iid] [frequency, sparse, can, sparsity, via, order, invariant, power, one, see] [random, measure, distribution, section, finite, given, data, point, infinite, nonparametric, permutation, based, asymptotic] [model, process, new, poisson, form, behavior, demonstrate, framework] [sequence, network, arxiv, dense, figure, single, use, generated, generate, generative, work, different]
Density Estimation via Discrepancy Based Adaptive Sequential Partition
Dangna Li, Kun Yang, Wing Hung Wong
Dangna Li, Kun Yang, Wing Hung Wong
[partition, piecewise, number, find, cluster, tree, total, split, probability] [function, constant, algorithm, defined, set, binary, let, opt, theorem, bound, bounded, problem, achieves, learning, class] [method, carlo, convergence, monte, bayesian, size, objective, supplementary] [can, one, good, error, first, following, see, dimension, denote, second, order, initialization, also, analysis] [density, estimation, based, dsp, discrepancy, data, distance, hellinger, star, true, given, section, estimate, bsp, nonparametric, distribution, estimator, kde, two, random, kernel, sample, estimated, computationally, underlying, ingredient] [time, demonstrate, along, location] [used, table, use, domain, figure, propose, three, different]
Supervised learning through the lens of compression
Ofir David, Shay Moran, Amir Yehudayoff
Ofir David, Shay Moran, Amir Yehudayoff
[equivalence, theory, whose] [compression, scheme, agnostic, pac, learning, loss, learnability, theorem, hypothesis, let, implies, learnable, every, proof, function, general, class, rate, equivalent, show, learner, multiclass, appears, version, exists, compactness, subsection, property, case, follows, dichotomy, algorithm, realizable, study, consider, constant, combinatorial, manfred, defined, inf, context, setting, may, even, israel, set, label, argument, known] [size, log, convergence, approximate, full, machine, shai] [dimension, following, linear, error, can, first] [sample, selection, uniform, statistical, distribution, given, based, two, finite, section, empirical, journal] [showing, statement] [part, use, input, work, categorization, similar, using, compressing]
Clustering with Same-Cluster Queries
Hassan Ashtiani, Shrinu Kushagra, Shai Ben-David
Hassan Ashtiani, Shrinu Kushagra, Shai Ben-David
[clustering, number, cluster, probability, subset, conforms, clusterability, definition, ssac] [algorithm, set, query, oracle, problem, satisfies, show, let, complexity, property, will, notion, optimal, hardness, theorem, instance, access, provide, prove, lemma, proof, lower, bound, case, least, setting, even, appendix, assume, cost, function, exists, niceness, balcan, consider] [log, efficient] [can, solution, one, computational, also, note, phase, first, following, condition, hard] [euclidean, data, two, result, given, center, polynomial, provided, mean] [framework, target, another, whether, form, time, expert] [domain, supervision, without, using, work, natural, different, approach, help]
Deep ADMM-Net for Compressive Sensing MRI
yan yang, Jian Sun, Huibin Li, Zongben Xu
yan yang, Jian Sun, Huibin Li, Zongben Xu
[graph, number, average] [function, fast, defined, algorithm, loss, learning, general, ratio, achieves, set] [stage, sampling, method, update, computed, optimization, gradient, compute] [admm, mri, can, reconstructed, regularization, sensing, nmse, computational, sparse, magnetic, multiplier, resonance, compressive, imaging, following, ieee, also, significantly, first, min, iterative, chest, dictionary, arg, pano] [data, nonlinear, test, corresponding, shrinkage, given] [brain, operation, time, direction] [reconstruction, layer, image, deep, network, flow, transform, using, output, convolution, accuracy, figure, different, learned, initialized, training, shown, filter, psnr, four, three, learn, architecture, novel, compared, train, dct, net]
Temporal Regularized Matrix Factorization for High-dimensional Time Series Prediction
Hsiang-Fu Yu, Nikhil Rao, Inderjit S. Dhillon
Hsiang-Fu Yu, Nikhil Rao, Inderjit S. Dhillon
[graph, tar, forecasting, among, dependency, many, structure, incorporate, negative] [consider, set, theorem, conference, appendix, learning, let, general, kxt, problem, might, even] [latent, standard, large, autoregressive, method, usually] [can, matrix, missing, regularizer, trmf, factorization, regularization, existing, handle, also, see, dlm, following, min, regularizers, gar, highly, yit, lag, note, electricity, regularized, formulation, linear, one] [data, section, two, gaussian, corresponding, given, embeddings, covariance, dimensional] [time, temporal, series, model, framework, unlike, simple, traffic] [use, approach, figure, international, used, weight, table, proposed, novel, scale, learn, prediction]
Rényi Divergence Variational Inference
Yingzhen Li, Richard E. Turner
Yingzhen Li, Richard E. Turner
[negative, definition, propagation] [bound, learning, conference, algorithm, special, consider, case, theorem, obtain, function, now] [variational, approximation, log, inference, bayesian, posterior, stochastic, approximate, divergence, likelihood, method, optimisation, datasets, machine, energy, importance, expectation, vae, however, frey, monte, carlo, lvi, iwae, marginal, processing, sep, wmax] [can, also, exact, one, following, first, local] [test, section, gaussian, distribution, fixed, family, mle, point, finite, bias, theoretical, sample, useful, alpha] [new, model, framework, information, considers, upon] [figure, neural, using, different, international, proposed, used, face, applied, network, recent, work]
Fast Algorithms for Robust PCA via Gradient Descent
Xinyang Yi, Dohyung Park, Yudong Chen, Constantine Caramanis
Xinyang Yi, Dohyung Park, Yudong Chen, Constantine Caramanis
[partial, probability, number, thus] [algorithm, complexity, consider, setting, problem, let, show, guarantee, theorem, convex, assume, case, fast, constant, set, annual, even, satisfies] [gradient, method, log, processing, step, deterministic, descent, faster] [matrix, sparse, running, can, observed, rank, phase, svd, via, pca, sinit, corruption, completion, linear, error, norm, following, ieee, first, row, suppose, alternating, projected, altproj, exact, sujay, singular, min, partially, note, initialization, nonzero, low, symposium, ialm, emmanuel, factorized] [robust, given, robustness, two, based, estimator, statistical, result] [time, information, model, observation] [using, figure, preprint, neural, arxiv, propose, compared, fully, produce]
Memory-Efficient Backpropagation Through Time
Audrunas Gruslys, Remi Munos, Ivo Danihelka, Marc Lanctot, Alex Graves
Audrunas Gruslys, Remi Munos, Ivo Danihelka, Marc Lanctot, Alex Graves
[definition, total, number, established, reverse] [cost, algorithm, optimal, define, will, strategy, may, general, budget, case, learning, every] [core, step, size, consumption, standard, capacity, argmin, respect] [computational, can, one, also, computation, suppose, following, comparison] [length, fixed, given, curve, particular, two] [state, internal, time, forward, backpropagation, operation, policy, backward, remember, usage, measured, dynamic, next, memorization, execution, value, simple, within, typically] [memory, hidden, rnn, single, sequence, figure, neural, network, recurrent, approach, using, input, used, previous, per, different, position, proposed, output, alex, storing, shown, store, deepmind, able, google, deep]
Learning and Forecasting Opinion Dynamics in Social Networks
Abir De, Isabel Valera, Niloy Ganguly, Sourangshu Bhattacharya, Manuel Gomez Rodriguez
Abir De, Isabel Valera, Niloy Ganguly, Sourangshu Bhattacharya, Manuel Gomez Rodriguez
[opinion, sentiment, average, social, forecasting, message, steady, number, eht, find, april, influence, simulate, bvu] [set, may, show, appendix, consider, theorem, algorithm, will, posted, depends, property] [stochastic, efficient, parameter, compute, latent, forecast, consensus, key, sampling] [can, user, accurate, first, linear, following, real, one, matrix, leverage] [given, estimation, data, conditional, two, based, differential, journal, denotes, multivariate, theoretical, twitter, converge] [model, time, hawkes, simulation, poisson, modeling, history, temporal, state, intensity, framework, predictive, evolution, information, process, markov, value, recorded] [using, figure, network, performance, proposed]
Professor Forcing: A New Algorithm for Training Recurrent Networks
Alex M. Lamb, Anirudh Goyal ALIAS PARTH GOYAL, Ying Zhang, Saizheng Zhang, Aaron C. Courville, Yoshua Bengio
Alex M. Lamb, Anirudh Goyal ALIAS PARTH GOYAL, Ying Zhang, Saizheng Zhang, Aaron C. Courville, Yoshua Bengio
[loop, number, either, quality, negative] [learning, set, algorithm] [sampling, likelihood, method, evaluation, sequential, processing, gradient, step, objective, stochastic] [can, also, one, running, observed, note, much] [distribution, data, sample, two, length, conditioning, given, selected] [forcing, professor, teacher, model, behavior, time, human, mode, information, van, raw] [training, sequence, neural, generative, network, rnn, recurrent, hidden, discriminator, arxiv, using, used, scheduled, generator, generated, input, output, preprint, validation, language, handwriting, use, generation, adversarial, better, prediction, generating, task, figure, open, classifier, trained, randomly, synthesis, train, architecture, layer, image, mnist]
Full-Capacity Unitary Recurrent Neural Networks
Scott Wisdom, Thomas Powers, John Hershey, Jonathan Le Roux, Les Atlas
Scott Wisdom, Thomas Powers, John Hershey, Jonathan Le Roux, Les Atlas
[number, quality, average, represent] [set, learning, best, loss, rate, achieves, show, problem, consider, since, function, theorem, argument, differentiable] [unitary, urnn, gradient, urnns, parameterization, recurrence, evaluation, capacity, stiefel, stft, permuted, descent, processing, optimization, suggests, intelligibility] [matrix, dimension, can, lie, restricted, synthetic, solution, vector] [manifold, test, space, data, section, given, theoretical] [state, system, optimize, determine, model, baseline, time] [using, recurrent, neural, speech, use, lstm, hidden, training, performance, memory, network, figure, table, validation, task, lstms, consists, sequence, proposed, natural, prediction, perceptual, used, shown, mnist, output]
Learned Region Sparsity and Diversity Also Predicts Visual Attention
Zijun Wei, Hossein Adeli, Minh Hoai, Greg Zelinsky, Dimitris Samaras
Zijun Wei, Hossein Adeli, Minh Hoai, Greg Zelinsky, Dimitris Samaras
[detection, number] [set, consider, return, svm, diversity, show, function, might] [method, divergence, evaluation] [auc, also, local, sparse, can, sparsity, significantly, localization, ranking, one, first, computational] [center, bias, selected, test, mean, people, distribution, gaussian, based, searching] [model, human, search, target, information, eye, new, failure, evidence] [visual, sdr, image, object, rrsvm, attention, region, priority, classification, multiple, dataset, map, poet, used, using, predict, different, figure, inhibition, score, training, fixation, mechanism, feature, performance, pascal, predicting, saliency, generated, prediction, spatial, learned, three, use, trained, pet, resized, vision, computer, work, voc, table, category, neural]
Interaction Networks for Learning about Objects, Relations and Physics
Peter Battaglia, Razvan Pascanu, Matthew Lai, Danilo Jimenez Rezende, koray kavukcuoglu
Peter Battaglia, Razvan Pascanu, Matthew Lai, Danilo Jimenez Rezende, koray kavukcuoglu
[interaction, string, relation, potential, whose, external, many, contained, represents, represent, represented, relational] [learning, show, learnable, constant] [mse, applies] [can, also, one, matrix, first] [two, generalization, test, well, data] [model, complex, time, across, abstract, gravitational, future, system, velocity, receiver, spring, simulation, baseline, reason, control, effect, state, another, exploit, ability] [physical, input, reasoning, object, neural, predict, network, prediction, used, learn, training, deep, trained, different, output, mlp, using, rigid, applied, three, engine, cnns, static, use, novel, ground, truth, hidden, scene]
Discriminative Gaifman Models
Mathias Niepert
Mathias Niepert
[gaifman, knowledge, relational, base, neighborhood, tuple, number, relation, structure, formula, graph, tuples, probability, connected, logical, probabilistic, negative, whose, path, include] [learning, set, conference, every, query, problem, now, class, let, complexity, theorem, confidence, algorithm] [inference, machine, large, compute, free, size] [can, local, one, also, locality, completion] [positive, data, embeddings, embedding, given, locally, random, corresponding] [model, within, form, artificial, target, complex] [figure, learn, training, neural, perform, feature, object, generated, domain, representation, work, per, used, generate, language, table, learned, input]
Generative Adversarial Imitation Learning
Jonathan Ho, Stefano Ermon
Jonathan Ho, Stefano Ermon
[causal, interaction] [learning, cost, function, will, algorithm, problem, optimal, max, lemma, conference, convex, minimize, show, constant, defined, cloning, appendix, set] [large, dual, step, gradient, respect, primal, machine, optimization, efficient, expectation, due, objective] [can, proposition, linear, arg, regularizer, following, exactly, certain, min, optimum, running] [measure, entropy, data, given, true, section, sample, maximum, random] [policy, expert, occupancy, reinforcement, irl, imitation, apprenticeship, gail, inverse, behavioral, control, environment, directly, trpo, behavior, guided] [performance, using, neural, generative, adversarial, international, dataset, discriminator, approach, matching, work, used, learned, training]
Probabilistic Inference with Generating Functions for Poisson Latent Variable Models
Kevin Winner, Daniel R. Sheldon
Kevin Winner, Daniel R. Sheldon
[probability, variable, runtime, number, detection, degree, resulting, enough, graphical] [algorithm, eliminate, class, set, let, function, theorem, will, fast, show, known] [pgf, posterior, inference, parameter, line, pgfs, factor, compute, latent, nmax, likelihood, hmms, unnormalized, marginals, abundance, end, approximate, faster, countably, implementation, develop, standard, marginal, apply] [can, exact, tail, first, truncated, also, observed, see, proposition, related, one] [joint, population, infinite, polynomial, discrete, based, given, data, two, estimation, mean, finite, true, hmm, journal] [poisson, forward, model, time, elimination, form, value, series, simple, count] [generating, figure, using, perform, representation, approach, use, instead, previous]
Active Nearest-Neighbor Learning in Metric Spaces
Aryeh Kontorovich, Sivan Sabato, Ruth Urner
Aryeh Kontorovich, Sivan Sabato, Ruth Urner
[number, probability, possible] [active, label, set, sin, let, theorem, learning, compression, passive, gottlieb, learner, algorithm, complexity, marmann, general, selectscale, guarantee, distmon, setting, kontorovich, defined, binary, lemma, uin, hnn, depends, since, competitive, show, define, generatennset, margin, provide, smallest, proof, requested] [size, end] [error, can, denote, following, analysis, first, also] [metric, sample, procedure, nearest, neighbor, given, selection, empirical, section, estimate, generalization, based, random, selected, bayes, consistent, two] [search, model, therefore, left, value, information, new] [scale, classification, output, using, labeled, different, prediction, similar, propose, classifier, approach, input, unlabeled]
Regularization With Stochastic Transformations and Perturbations for Deep Semi-Supervised Learning
Mehdi Sajjadi, Mehran Javanmardi, Tolga Tasdizen
Mehdi Sajjadi, Mehran Javanmardi, Tolga Tasdizen
[combination, contains, number] [loss, set, function, learning, randomized, achieve, rate, conference, example] [large, five, method, processing, size, minimizes] [can, error, sparse, also, one, first, see, vector] [data, sample, two, based, random, test, difference] [model, another, information] [labeled, training, unlabeled, use, using, unsupervised, convolutional, accuracy, neural, different, proposed, dropout, network, trained, dataset, prediction, computer, used, similar, train, perform, multiple, svhn, mnist, augmentation, norb, improve, randomly, international, available, layer, table, feature, convnets, deep, vision, ladder, single, supervised, classification, convnet, last, task, pattern, arxiv]
Generalization of ERM in Stochastic Convex Optimization: The Dimension Strikes Back
Vitaly Feldman
Vitaly Feldman
[probability, number, many, describe, computable] [convex, bound, lower, let, complexity, erm, every, function, set, will, setting, define, general, sco, exists, algorithm, known, defined, binary, dependence, even, theorem, now, problem, constant, show, lipschitz, bpd, upper, imply, asymptotically, learning, case, consider, class, max, risk, prove, mink, vitaly] [stochastic, convergence, optimization, smooth, term, standard, additional, efficiently, gradient, log] [can, stability, regularization, linear, also, one, necessary, note, analysis, following, denote, support, convexity] [sample, uniform, distribution, generalization, construction, given, maximum, based, empirical, radius] [therefore, ball, code, whether, time] [use, approach, work, unit, question, natural]
On Explore-Then-Commit strategies
Aurelien Garivier, Tor Lattimore, Emilie Kaufmann
Aurelien Garivier, Tor Lattimore, Emilie Kaufmann
[rule, possible, identification, number] [regret, strategy, etc, algorithm, optimal, lower, arm, bandit, theorem, known, bound, upper, stopping, unknown, minimax, show, asymptotically, best, let, ucb, lai, proof, kaufmann, uniformly, inf, set, choose, problem, learning, bounded, appendix, garivier, case, even, now, confidence, conference, prove, chooses, suboptimal, perchet, chosen, horizon, exploitation, setting, robbins, active] [sequential, log, end, factor, efficient, sampling] [can, also, arg, gap, relative, one, phase, analysis, following] [based, given, gaussian, two, lim, asymptotic, section, empirical, difference, uniform, mean] [exploration, action, time, simple, design] [using, fully, improve, presented]
Improved Deep Metric Learning with Multi-class N-pair Loss Objective
Kihyuk Sohn
Kihyuk Sohn
[negative, number, mining, clustering, identification, many] [loss, learning, example, verification, class, since, online, function, set, observe, improvement] [batch, large, log, standard, efficient] [can, one, product, hard, also, formulation, following, comparison] [metric, embedding, data, positive, distance, construction, two, equation, well, selected, section, test] [model] [triplet, deep, recognition, face, performance, training, different, using, proposed, softmax, evaluate, accuracy, per, multiple, dataset, object, trained, network, train, output, image, table, use, contrastive, figure, similarity, feature, tri, visual, propose, neural, classification, input, composed, pair, better, vrf, randomly, nmi, instead, representation]
Greedy Feature Construction
Dino Oglic, Thomas Gärtner
Dino Oglic, Thomas Gärtner
[number, definition, subset, probability] [function, set, algorithm, learning, hypothesis, greedy, let, banach, considered, problem, defined, theorem, appendix, rate, convex, expected, bounded] [descent, gradient, step, method, convergence, incremental, constructive, fitting, machine, capacity, respect, optimization, large, smooth, walltime] [linear, can, error, also, regression, good, regularization, following, one] [space, data, functional, ridge, kernel, constructed, empirical, hilbert, squared, measure, sample, generalization, construct, construction, section, provided, smoothness, random, specified, carte, basis] [choice, target, model, information, hyperparameter, goal, time] [feature, approach, sequence, training, using, representation, proposed, residual, neural, single, compact, performance, presented]
Learning HMMs with Nonparametric Emissions via Spectral Decompositions of Continuous Matrices
Kirthevasan Kandasamy, Maruan Al-Shedivat, Eric P. Xing
Kirthevasan Kandasamy, Maruan Al-Shedivat, Eric P. Xing
[number, chebyshev, variable, theory, probability] [algorithm, will, bound, lemma, let, learning, theorem, case, setting, concentration] [method, hmms, latent, efficient, compute] [can, error, spectral, singular, first, via, matrix, one, denote, linear, analysis, recover, following, main, perturbation, rank] [nonparametric, density, discrete, joint, parametric, true, observable, conditional, estimation, hmm, estimated, estimating, kernel, emission, two, given, kde, qmatrix, distribution, sample, data, space, estimate, procedure] [continuous, state, time, markov, predictive, model] [hidden, used, using, training, use, representation, prediction, recent, sequence, compare]
Multivariate tests of association based on univariate tests
Ruth Heller, Yair Heller
Ruth Heller, Yair Heller
[independence, resulting, third, aggregation, number, partition, level, fact] [let, will, hypothesis, may, show, set, since, consider, every, best, function] [bivariate, method, processing, log] [can, power, one, see, following, vector, order, computational, first, also, good] [test, univariate, center, multivariate, sample, distribution, point, consistent, based, null, statistic, random, two, measure, testing, fxy, minp, significance, result, data, permutation, journal, retton, distance, section, specific, useful, procedure, versus, annals, comparing, population, normal, powerful, joint] [value, choice, information, carry] [using, single, approach, figure, use, pooling, table, multiple, different, referred, novel, category, neural]
Generating Long-term Trajectories Using Deep Hierarchical Networks
Stephan Zheng, Yisong Yue, Jennifer Hobbs
Stephan Zheng, Yisong Yue, Jennifer Hobbs
[quality] [learning, study, class, may, set, case, general, every, function] [latent, large] [can, one, also, preference, via] [two, data, space] [hpn, policy, hierarchical, model, basketball, raw, rollouts, rollout, player, behavior, agent, state, action, towards, gti, modeling, planning, trajectory, court, reinforcement, move, issn, expert, depicts, ait, sit, macro, straight, velocity, generates, spatiotemporal] [attention, using, network, figure, gru, cnn, approach, ground, neural, weak, output, realistic, truth, training, use, work, transfer, deep, accuracy, instead, conv, used, trained, generated, predicted, recurrent, prediction, sequence, natural]
R-FCN: Object Detection via Region-based Fully Convolutional Networks
jifeng dai, Yi Li, Kaiming He, Jian Sun
jifeng dai, Yi Li, Kaiming He, Jian Sun
[detection, mining, vote, bin, average] [set, loss, example, learnable, bank, microsoft, fast] [faster, method, proposal, standard, size] [following, can, one, computation, also, paper, hard] [test, data, result] [roi, box, information, time, evaluated, design] [convolutional, object, layer, score, fully, pooling, training, image, map, table, using, train, rpn, subnetwork, voc, region, use, figure, classification, pascal, deep, shared, spatial, feature, depth, per, accuracy, entire, trainval, bounding, last, semantic, neural, single, candidate, architecture, val, translation, evaluate, trous, scale, network, adopt, position, conv, used]
Large-Scale Price Optimization via Network Flow
Shinji Ito, Ryohei Fujimaki
Shinji Ito, Ryohei Fujimaki
[cut, programming, number, find, graph] [price, algorithm, problem, optimal, function, submodular, revenue, property, binary, supermodular, strategy, upper, set, define, cost, relaxation, might, best, lower, bound, even, satisfies] [optimization, method, approximate, objective, quadratic, compute, efficient, large, gradient, solved] [can, demand, computational, gross, profit, qpbo, regression, minimum, qpboi, solution, sdprelax, elasticity, following, product, exact, paper, substitute, ieee, satisfying, assumption, prescriptive, also, regarding, good, analysis, fij] [basis, data, construct] [value, time, subject, model] [proposed, using, network, flow, cross, pattern, table, complementary, use]
A Theoretically Grounded Application of Dropout in Recurrent Neural Networks
Yarin Gal, Zoubin Ghahramani
Yarin Gal, Zoubin Ghahramani
[many, sentiment, probabilistic, probability] [learning, will, appendix, rate] [variational, bayesian, variant, standard, inference, approximate, approximating, large, technique, naive, experiment, posterior, log, step] [can, matrix, error, one, following, analysis, existing, see, small, note] [distribution, test, embedding, word, given, embeddings, random] [model, time, decay, new, early, medium, next, identical, prior] [dropout, recurrent, neural, lstm, weight, rnn, different, using, use, perplexity, layer, language, used, gru, input, deep, single, applied, compared, network, validation, proposed, output, performance, moon, approach, sequence, modelling, rnns, overfitting, regularisation, penn, mask, randomly, figure, lstms, untied, seen]
Unsupervised Learning for Physical Interaction through Video Prediction
Chelsea Finn, Ian Goodfellow, Sergey Levine
Chelsea Finn, Ian Goodfellow, Sergey Levine
[interaction] [learning, show, conference] [method, processing] [can, also, one, interactive] [test, two, transformation, well] [model, future, human, information, prior, robot, action, state, new, internal, next, robotic, time, previously, raw, directly, predictive] [prediction, video, motion, image, previous, object, pixel, figure, neural, predicted, lstm, cdna, predict, using, physical, different, multiple, dataset, convolutional, proposed, trained, vision, performance, predicting, conv, applied, predicts, used, without, use, frame, international, approach, computer, stp, learn, mask, including, unsupervised, training, spatial, produce, explicitly, million, appearance, feedforward, pushing, work, reconstruct, evaluate]
Convolutional Neural Fabrics
Shreyas Saxena, Jakob Verbeek
Shreyas Saxena, Jakob Verbeek
[number, connected, obtained, node, grows, many, sum, thus, contains] [learning, show, best, exponentially, since, obtain, consider] [large, size, factor] [can, sparse, one, also, error, first, connectivity, see, related, signal] [two, data, embedded] [across, response, state, model, process, along] [convolutional, input, network, fabric, using, channel, use, scale, neural, figure, dense, used, layer, image, per, pooling, part, semantic, table, training, last, cnn, architecture, output, deep, filter, trellis, coarser, single, classification, three, work, maxout, convolution, memory, segmentation, dataset, resolution, augmentation]
Multiple-Play Bandits in the Position-Based Model
Paul Lagrée, Claire Vernade, Olivier Cappe
Paul Lagrée, Claire Vernade, Olivier Cappe
[cascade, list, expression, number] [lower, bound, regret, algorithm, optimal, bandit, learning, arm, theorem, learner, pbm, will, asymptotically, inf, expected, round, upper, appendix, obtain, click, considered, problem, provide, examination, uniformly, assume, confidence, best, may, decreasing, chosen, close, suboptimal, general, every] [efficient, log, machine, step, parameter, stochastic, end] [can, one, user, item, first, also, order, analysis, following, proposition, observed] [section, given, corresponding, two, data, based, estimator, lim, independent, associated] [model, feedback, reward, time, information, search, action, simple] [position, using, multiple, figure, use, content, performance]
Sublinear Time Orthogonal Tensor Decomposition
Zhao Song, David Woodruff, Huan Zhang
Zhao Song, David Woodruff, Huan Zhang
[number, total, probability, variable] [algorithm, theorem, let, achieve, show, constant, fast, smaller, set, since, wang, even, lemma] [sampling, importance, method, inner, latent, approximate, faster, datasets, large, factor] [tensor, can, one, sublinear, power, symmetric, also, running, preprocessing, decomposition, norm, pprox, dimension, sketching, small, main, need, synthetic, orthogonal, spectral, analysis, much, order, denote, first, noise, still, suppose, contraction, asymmetric, product] [slice, random, robust, based, squared, two, dirichlet, estimate, sample, theoretical] [time, take, reading, median] [residual, work, using, without, use, previous, different, generating, input, table, instead]
Generating Videos with Scene Dynamics
Carl Vondrick, Hamed Pirsiavash, Antonio Torralba
Carl Vondrick, Hamed Pirsiavash, Antonio Torralba
[often] [learning, show, since, may, will] [large, usually, latent] [can, one, also, order, first, relative, believe] [two, random, data, interested, people, suggest] [model, temporal, future, action, except, human] [video, generative, network, use, frame, generation, unlabeled, adversarial, generated, using, unsupervised, generator, generate, train, labeled, figure, static, discriminator, background, representation, architecture, convolutional, scene, work, learn, stream, foreground, input, deep, recognition, prefer, vgan, image, motion, visual, performance, approach, realistic, training, instead, antonio, golf, layer, hidden, strided, learned, better, initialized, experimented, accuracy, object, outperforms, promising]
Differential Privacy without Sensitivity
Kentaro Minami, HItomi Arai, Issei Sato, Hiroshi Nakagawa
Kentaro Minami, HItomi Arai, Issei Sato, Hiroshi Nakagawa
[probability, adjacent, strong, proportional, definition, obtained, becomes, whose] [privacy, satisfies, loss, function, let, theorem, private, inequality, boundedness, lipschitz, convex, lsi, bound, randomized, upper, learning, constant, corollary, set, algorithm, general, proof, assume, satisfy, bounded, differentially, problem, ratio, defined, risk, example, arbitrary, said, consider, depends, define, concentration, property, tokyo, case] [log, posterior, parameter, size, dkl, twice, university, gradient, sampling, datasets, respect, monte, requires, langevin] [can, following, analysis, assumption, condition, suppose, certain, proposition, sufficient, related] [gibbs, exponential, differential, space, density, estimator, measure, two, distribution, difference, sensitivity, given, gaussian, statistical, exp, empirical, provides, shrinkage, classical] [prior] [mechanism, dataset]
SPALS: Fast Alternating Least Squares via Implicit Leverage Scores Sampling
Dehua Cheng, Richard Peng, Yan Liu, Ioakeim Perros
Dehua Cheng, Richard Peng, Yan Liu, Ioakeim Perros
[amazon, review, many, number, probability] [algorithm, least, randomized, cost, optimal, will, theorem] [sampling, factor, approximation, efficient, importance, size, method, gradient, stochastic, kronecker, step, due, efficiently] [tensor, matrix, can, leverage, row, decomposition, krp, linear, alternating, rank, spals, error, product, via, computational, sparse, first, solving, also, related, low, main, regression, sublinear, one, spectral, sketching, singular, running, much, square] [statistical, data, estimation, two, sample, section, numerical, random, provides] [time, design, allows, form, directly, value, target] [score, input, work, recent, using, previous, similar]
Estimating Nonlinear Neural Response Functions using GP Priors and Kronecker Methods
Cristina Savin, Gasper Tkacik
Cristina Savin, Gasper Tkacik
[] [rate, learning, function, defined, example] [log, inference, kronecker, additional, posterior, approximate, sampling, due] [can, spectral, also, linear, stability, matrix, see, overall, sparse] [place, data, estimate, covariance, gaussian, mixture, mean, functional, estimator, space, kernel, multidimensional, way, estimation, familiar, distribution] [tuning, firing, complex, information, selectivity, animal, time, speed, model, within, direction, cell, process, modulation, exploration, environment, found, simple, nature, poisson, across, hippocampal, behavior, temporal, activity, prior, artificial, grid] [field, spatial, neural, traditional, input, network, using, use, position, used, motion, several, open, scale, figure, recent, representation]
Learning What and Where to Draw
Scott E. Reed, Zeynep Akata, Santosh Mohan, Samuel Tenka, Bernt Schiele, Honglak Lee
Scott E. Reed, Zeynep Akata, Santosh Mohan, Samuel Tenka, Bernt Schiele, Honglak Lee
[] [set, learning, show, case, gray, class, provide] [black, draw, variational] [can, global, also, local, vector, noise, small, tensor, addition] [conditioning, embedding, conditional] [box, model, location, white, human, control, long, blue, form] [bird, image, text, keypoints, part, spatial, keypoint, bounding, generative, generate, figure, beak, deep, using, red, network, conditioned, recurrent, adversarial, gawwn, pointy, use, depth, convolutional, yellow, discriminator, neural, pose, object, caption, generator, generating, face, shown, trained, pathway, realistic, training, concat, man, visual, gan, fed, synthesis, position, several, final, feature, generation, generated, cub, orange, encoder, approach, golf]
A Multi-step Inertial Forward-Backward Splitting Method for Non-convex Optimization
Jingwei Liang, Jalal Fadili, Gabriel Peyré
Jingwei Liang, Jalal Fadili, Gabriel Peyré
[identification, partial, definition] [mifb, function, example, algorithm, theorem, let, scheme, bnd, optimal, problem, rate, near, property, kxk, proper, set, since, satisfies, loss, bound, case, choose, consider, class, bounded, show, will] [convergence, smooth, optimization, siam, gradient, method, faster, proximal, splitting, objective, iterates, supplementary] [inertial, def, can, local, linear, also, rank, following, partly, relative, thm, see, condition, global, critical, sparse, submanifold, ifb, analysis, lsc, noted, min, proposition, comparison] [journal, finite, locally, point, given, empirical, smoothness, section, numerical, mathematical, riemannian, space] [choice, continuous] [figure, sequence, generated]
Depth from a Single Image by Harmonizing Overcomplete Local Network Predictions
Ayan Chakrabarti, Jingyu Shao, Greg Shakhnarovich
Ayan Chakrabarti, Jingyu Shao, Greg Shakhnarovich
[find, number, coefficient] [set, confidence, since, learning] [derivative, standard, log, method, inference, efficient, large, zeroth, factor, respect] [local, can, also, order, global, relative, second, qij, error, first, vector, note, high, related] [estimation, corresponding, estimate, mixture, mean, geometry, useful, geometric] [form, location, value, characterize, individual, along, information] [depth, network, map, scene, different, image, neural, single, training, using, use, approach, output, globalization, monocular, accuracy, surface, train, used, overcomplete, trained, figure, able, various, performance, color, convolutional, multiple, produce, predicted, instead, validation, predicting, task, nyu, input, better]
Efficient High-Order Interaction-Aware Feature Selection Based on Conditional Mutual Information
Alexander Shishkin, Anastasia Bezzubtseva, Alexey Drutsa, Ilia Shishkov, Ekaterina Gladkikh, Gleb Gusev, Pavel Serdyukov
Alexander Shishkin, Anastasia Bezzubtseva, Alexey Drutsa, Ilia Shishkov, Ekaterina Gladkikh, Gleb Gusev, Pavel Serdyukov
[cmicot, team, number, sfs, scoring, interaction, cmi, subset, among, variable, whose, iwfs, account, rcdfs, cmim, identify] [binary, set, greedy, since, complexity, optimal, problem, function, strategy, appendix, algorithm, learning, best, will, let, class] [method, datasets, technique, select, approximation, opposing, large, optimization, size, step, end] [one, following, can, computational, dimension, also, popular, note, existing, order] [selection, based, mutual, selected, two, joint, given, conditional, sample, equal, distribution] [information, search, target, built] [feature, score, approach, complementary, novel, performance, candidate, used, able, already, classification, using, different, several, higher, neural]
Training and Evaluating Multimodal Word Embeddings with Large-scale Web Annotated Images
Junhua Mao, Jiajing Xu, Kevin Jing, Alan L. Yuille
Junhua Mao, Jiajing Xu, Kevin Jing, Alan L. Yuille
[negative, collected, number, base] [learning, strategy, click, loss, query, will, set] [evaluation, datasets, large, standard, pure] [related, user, one, can, sampled, also] [word, embedding, embeddings, positive, data, sample, based, given, section] [model, information, search, state] [dataset, visual, image, phrase, text, training, multimodal, million, neural, gold, rnn, trained, similar, figure, similarity, using, weight, softmax, semantically, proposed, layer, final, language, learned, hair, recurrent, use, previous, sentence, sharing, adopt, effective, relatedness, score, pinterest, shown, billion, work, gru, annotated, table, several, compared, learn, scale, evaluate, semantic, coco]
SDP Relaxation with Randomized Rounding for Energy Disaggregation
Kiarash Shaloudegi, András György, Csaba Szepesvari, Wilsun Xu
Kiarash Shaloudegi, András György, Csaba Szepesvari, Wilsun Xu
[number, integer, programming, total, find, normalized] [problem, algorithm, set, randomized, relaxation, convex, function, best, binary, minimize, learning] [method, energy, inference, approximate, quadratic, large, consumption, optimization, fhmms, objective, variational, end, efficient, factorial, solve, fhmm] [power, kolter, can, appliance, disaggregation, solution, admm, sdp, error, rounding, jaakkola, load, also, zgs, semidefinite, matrix, sparse, need, ieee, signal, smart, vector, min, home, monitoring, via, running, arg, synthetic, park, following] [data, denotes, based, given, section, length, hmm, point, additive] [time, model, subject, state, usage, change, information, new, form, individual] [use, using, used, performance, neural, work, better]
Orthogonal Random Features
Felix X. Yu, Ananda Theertha Suresh, Krzysztof M. Choromanski, Daniel N. Holtmann-Rice, Sanjiv Kumar
Felix X. Yu, Ananda Theertha Suresh, Krzysztof M. Choromanski, Daniel N. Holtmann-Rice, Sanjiv Kumar
[structure, number] [ratio, show, fast, let, lower, provide, function, since, almost, lemma, set, theorem, cost, achieve, close] [variance, approximation, mse, method, large, approximate, log, var, unbiased, distributed] [orthogonal, can, matrix, linear, also, first, error, following, note, computation, diagonal, sampled, comparison, small, orthogonality, order, hence] [random, kernel, orf, sorf, rff, gaussian, transformation, fourier, circulant, based, fastfood, bias, provides, theoretical, nonlinear, mean, section, distance, space, empirical, distribution, estimation, empirically, korf, given, krff, estimator, fixed] [time] [structured, figure, feature, use, proposed, memory, classification, mnist, using, used, different, similar, compared, cifar, table]
Unsupervised Learning of Spoken Language with Visual Context
David Harwath, Antonio Torralba, James Glass
David Harwath, Antonio Torralba, James Glass
[branch, many, association, discovery] [learning, set, max, function] [machine, timit, processing] [computational, also, analysis, highly, can, one, approximately, error, dot, product, dimension, first, vector] [embedding, word, data, mean, embeddings, well, given] [model, search, across, information, time, take, along, form, within, modeling, framework] [image, audio, caption, network, speech, spoken, similarity, used, training, using, neural, language, recognition, visual, text, annotation, score, use, acoustic, spectrogram, learned, work, different, google, figure, object, ground, truth, layer, learn, perform, multimodal, deep, semantic, able, dataset, train, unsupervised, table, shown, vgg, similar, convolutional, recognize, recent]
Pruning Random Forests for Prediction on a Budget
Feng Nan, Joseph Wang, Venkatesh Saligrama
Feng Nan, Joseph Wang, Venkatesh Saligrama
[pruning, node, tree, rune, udget, number, ensemble, reedy, average, integer, pruned, leaf, constraint, path, ccp, program, iser, account, tradeoff, acquired, subset, prune, total, forest, majority, rule, indicates, impurity] [cost, example, set, learning, algorithm, budget, problem, optimal, expected, obtain, conference, consider, focus, label] [large, datasets, solve, dual, optimization, competing, due, constrained, method] [can, error, one, first, solving, also, high, matrix, low] [test, based, random, associated, corresponding, entropy, given] [acquisition, decision, usage, resource, time] [feature, used, accuracy, prediction, use, using, approach, classification, training, figure, propose, network, shown, international, outperforms, work]
Stochastic Online AUC Maximization
Yiming Ying, Longyin Wen, Siwei Lyu
Yiming Ying, Longyin Wen, Siwei Lyu
[probability, pairwise] [online, learning, solam, algorithm, sup, problem, maximization, function, loss, spp, let, theorem, convex, lemma, complexity, max, optimal, rate, roc, equivalent, surrogate, proof, defined, hinge, minimization, set, inst, assume, drawn, achieves] [stochastic, convergence, objective, gradient, optimization, datasets, method, saddle, update, step, descent, iteration, feat, approximation] [auc, can, min, formulation, following, main, linear, running, one, denote, solution, existing, need, analysis] [space, data, given, point, estimator, positive, true, sample, two, covariance, curve] [time, area, information, ability] [training, used, use, table, performance, proposed, classification, using, work, store, previous, several]
Adaptive Averaging in Accelerated Descent Dynamics
Walid Krichene, Alexandre Bayen, Peter L. Bartlett
Walid Krichene, Alexandre Bayen, Peter L. Bartlett
[variable, theory, thus, ezi, many, average, weighted, unique] [function, convex, adaptive, rate, theorem, algorithm, set, give, feasible, defined, whenever, consider, will, case, show, studied, since, example, lipschitz, problem, adaptively, learning, corollary, study, guarantee, existence] [accelerated, averaging, lyapunov, ode, convergence, energy, mirror, primal, descent, restarting, method, gradient, dual, replicator, strongly, quadratic, discretization, optimization, supplementary, derivative, restart, uniqueness] [can, solution, condition, second, following, note, suppose, one, also] [given, discrete, section, positive, equation, mathematical, two] [heuristic, trajectory, simple, speed, taking, time, corresponds, significant, form] [using, original, propose, weight, proposed, figure, used, different]
Preference Completion from Partial Rankings
Suriya Gunasekar, Oluwasanmi O. Koyejo, Joydeep Ghosh
Suriya Gunasekar, Oluwasanmi O. Koyejo, Joydeep Ghosh
[total, partial, monotonic, ordering, subset, entity, definition, dag, directed, often, number] [algorithm, let, set, learning, arbitrary, convex, case, complexity, monotone, problem, margin, index, may] [standard, log, proximal, operator, efficient, parameter, gradient] [matrix, preference, completion, ranking, rank, can, low, collaborative, observed, isotonic, norm, following, listwise, vector, dbl, nuclear, order, onto, error, projection, recommender, letor, regression, retargeted, exact, singular, retargeting] [estimator, estimate, data, numerical, estimation, denotes, underlying, generalization] [within, brain, cognitive, value, simple, information, voxels, representing, evaluated] [using, proposed, performance, jointly, task, used, score, dataset, table, propose, affinity, approach, single]
Hierarchical Deep Reinforcement Learning: Integrating Temporal Abstraction and Intrinsic Motivation
Tejas D. Kulkarni, Karthik Narasimhan, Ardavan Saeedi, Josh Tenenbaum
Tejas D. Kulkarni, Karthik Narasimhan, Ardavan Saeedi, Josh Tenenbaum
[knowledge, number] [learning, function, set, conference, chooses, algorithm, rate, consider, expected] [key, stochastic, end, processing, efficient] [can, order, also, success, first, sparse] [space, critic, provides] [goal, reward, agent, reinforcement, controller, state, intrinsic, exploration, policy, hierarchical, value, temporal, model, extrinsic, learns, framework, time, game, delayed, action, intrinsically, information, decision, motivated, process, terminal, future, receives, atari, environment, autonomous, artificial, door, motivation, new, reaching, current, episode, internal, transition, left] [deep, figure, learn, using, neural, approach, different, training, use, international, representation, preprint, arxiv, proposed, architecture]
Multistage Campaigning in Social Networks
Mehrdad Farajtabar, Xiaojing Ye, Sahar Harati, Le Song, Hongyuan Zha
Mehrdad Farajtabar, Xiaojing Ye, Sahar Harati, Le Song, Hongyuan Zha
[social, intervention, number, average, programming, find, relation, influence, level, total] [function, optimal, problem, algorithm, maximization, online, appendix, cost, rate, let, constant, best, budget, will] [objective, stage, optimization, due, size] [can, user, one, real, minimum, first, overall, linear, synthetic, still, suppose, following] [point, given, data, random] [exposure, intensity, time, process, campaigning, control, cll, opl, rnd, exogenous, dynamic, activity, event, hawkes, policy, campaign, temporal, form, shaping, history, information, state, future, framework, desired, modeling, wei, world, next, grd, wfl, prk, target, steer, current] [network, performance, prediction, accuracy, different, used, previous, outperforms]
Double Thompson Sampling for Dueling Bandits
Huasen Wu, Xin Liu
Huasen Wu, Xin Liu
[number, normalized, thus, according, present, probability] [dueling, regret, arm, copeland, algorithm, condorcet, bound, learning, bandit, thompson, conference, general, lower, lemma, pij, achieves, pji, show, refine, rucb, may, bij, ccb, mab, will, winner, appendix, bounded, confidence, optimal, expected, suboptimal, rcs, achieve, study, let, consider] [log, sampling, posterior, machine] [can, first, preference, one, also, analysis, second, comparison, much, significantly, following, existing, proposition, user] [two, theoretical, practical, comparing, based, reduces, robust, selected] [information] [candidate, using, traditional, international, compared, propose, double, shown, compare, back, work, pair, score, similar]
Even Faster SVD Decomposition Yet Without Agonizing Pain
Zeyuan Allen-Zhu, Yuanzhi Li
Zeyuan Allen-Zhu, Yuanzhi Li
[block, find, rayleigh] [algorithm, theorem, obtain, let, best, even, corollary, dependence, depends, since, guarantee, provide, call, make] [stochastic, accelerated, method, faster, convergence, approximation, sampling, full, lanczos, compute, respect, due, approximate, apply] [running, matrix, krylov, can, singular, one, lazysvd, norm, musco, spectral, paper, also, first, svd, frobenius, satisfying, gap, yes, knd, relative, nnz, denote, column, rank, appxpca, symmetric, following, main, power, zeyuan, orthonormal, alternating, stated] [needed, result, given, type, two] [time, found, state, framework, value] [using, table, use, arxiv, used, four, output, performance, open, multiplicative, question]
Estimating the class prior and posterior from noisy positives and unlabeled data
Shantanu J. Jain, Martha White, Predrag Radivojac
Shantanu J. Jain, Martha White, Predrag Radivojac
[negative, probability, added, theory, labeling, identifiable] [class, algorithm, learning, set, let, label, setting, general, defined, function, lemma, theorem, binary, follows, will] [posterior, size] [noise, can, also, noisy, one, first, related, component] [data, positive, estimation, sample, mixture, true, proportion, nonparametric, parametric, univariate, distribution, practical, identifiability, estimating, mixing, estimate, density, transformation, random, alphamax, space, gaussian, msgmm, dimensionality, section, robust, statistical, equation, two, maximum] [prior, form, model] [unlabeled, labeled, classification, using, used, approach, input, use, classifier, learn, performance, different, explicitly, pair, generate]
Blind Regression: Nonparametric Regression for Latent Variable Models via Collaborative Filtering
Dogyoon Song, Christina E. Lee, Yihua Li, Devavrat Shah
Dogyoon Song, Christina E. Lee, Yihua Li, Devavrat Shah
[probability, according, variable, whose] [algorithm, function, provide, bounded, lipschitz, bound, set, lemma, follows, unknown, exists, theorem, assume] [latent, variance, method, variant, parameter, due, around, approximation] [collaborative, matrix, error, filtering, row, can, observed, regression, rating, analysis, local, user, movie, completion, also, column, factorization, dimension, order, noise, netflix, fraction, movielens, first, recommendation] [sample, kernel, neighbor, empirical, classical, mean, estimate, nearest, nonparametric, given, based, true, metric, difference, two, taylor, squared, exp, gaussian, commonly] [framework, information, long, model] [using, similar, similarity, use, predict, input, used, work]
Beyond Exchangeability: The Chinese Voting Process
Moontae Lee, Seok Hyun Jin, David Mimno
Moontae Lee, Seok Hyun Jin, David Mimno
[quality, voting, cvp, negative, helpfulness, vote, display, chinese, number, amazon, trendiness, average, social, crp, positional, polarity, review, probability, many, community, popularity, presentational, stackexchange, conformity, urn, stackoverflow, sentiment, opinion] [conference, function, ratio, since, online, every, learning, show, even, study] [log, parameter, due] [can, rank, user, first, phase, item, also, qij, one, existing, order, relative, note, restaurant] [two, based, positive, selection, given, length, estimated, bias] [response, time, model, process, intrinsic, new, information, whereas, behavioral, predictive, early, trajectory] [different, table, figure, better, using, single, instead, previous]
Stochastic Variance Reduction Methods for Saddle-Point Problems
Balamurugan Palaniappan, Francis Bach
Balamurugan Palaniappan, Francis Bach
[number, thus, probability, split, proportional, sum, many] [convex, algorithm, consider, may, monotone, appendix, complexity, problem, function, defined, get, always, assume, minimization, show, loss, learning, rate, fast, constant, theorem, every] [stochastic, operator, saga, proximal, svrg, convergence, accelerated, batch, bilinear, gradient, method, variance, separable, optimization, machine, sampling, factored, efficient, siam, reduction, iterate, iteration, variational, saddle, acceleration] [need, following, can, leading, matrix, vector, note, analysis, see, existing, extension, regularizer, linear, solution] [section, journal, two, associated, given] [individual, simple, value] [using, use, several, table, shown]
Towards Unifying Hamiltonian Monte Carlo and Slice Sampling
Yizhe Zhang, Xiangyu Wang, Changyou Chen, Ricardo Henao, Kai Fan, Lawrence Carin
Yizhe Zhang, Xiangyu Wang, Changyou Chen, Ricardo Henao, Kai Fan, Lawrence Carin
[monomial, variable, total, becomes] [function, drawn, show, uniformly, algorithm, may, adaptive, consider, case] [sampling, monte, momentum, standard, carlo, parameter, large, energy, method, convergence, initial, auxiliary] [can, one, solving, also, generalized, analysis, via, symmetric, first] [slice, hamiltonian, hmc, distribution, theoretical, numerical, sampler, analytic, sample, point, conditional, density, mixing, transformation, space, provided, practical, gamma, two, section, legendre, exponential, family, gaussian, random, acceptance, chain, journal, described, equation, resampling, uniform, independent, given] [target, system, kinetic, time, trajectory, corresponds, integration, value, along, autocorrelation, form, new, defines] [performance, using, original, figure, shown, approach, table]
Parameter Learning for Log-supermodular Distributions
Tatiana Shpakova, Francis Bach
Tatiana Shpakova, Francis Bach
[probabilistic, base, probability, thus, gumbel, review] [function, submodular, bound, consider, learning, convex, binary, max, may, show, lower, minimization, polytope, upper, std, set, since, known, equivalent, algorithm, subgradient, modular, problem] [logistic, log, parameter, alogistic, stochastic, likelihood, approximation, optimization, inference, approximate, expectation, efficient, gradient, variational] [can, min, noise, noisy, one, missing, also, note, following, linear] [maximum, section, two, based, random, conditional, given, data, journal, equal, distribution, sample, independent, well, discrete, estimation] [model] [image, use, using, used, learn, figure, perform, better, supervised, table, unsupervised, several, approach]
The Parallel Knowledge Gradient Method for Batch Bayesian Optimization
Jian Wu, Peter Frazier
Jian Wu, Peter Frazier
[knowledge, number, report] [function, algorithm, learning, set, regret, ucb, expected, deviation, show, improvement, maximizing, choose, will] [parallel, batch, optimization, gradient, machine, bayesian, method, size, immediate, qkg, posterior, standard, logistic, sampling, hyperparameters, especially, compute, efficient, develop, proposes, evaluation, processing, initial] [can, one, synthetic, noisy, min, regression, error, global, also, solution] [test, gaussian, section, mean, point, distribution, practical, two, sample, independent, finite, provides, testing, journal] [acquisition, tuning, process, optimize, next, prior, information] [evaluate, scale, domain, neural, using, better, figure, several, training, previous, performance, different]
Stochastic Optimization for Large-scale Optimal Transport
Aude Genevay, Marco Cuturi, Gabriel Peyré, Francis Bach
Aude Genevay, Marco Cuturi, Gabriel Peyré, Francis Bach
[probability, thus] [problem, optimal, algorithm, defined, set, function, maximization, known, arbitrary, consider, since, max, show, define, rate] [stochastic, dual, gradient, convergence, sgd, optimization, sag, processing, sinkhorn, compute, solve, iterates, incremental, method, approximation, proxy, supplementary, faster, expectation, large, iteration, primal] [can, solution, regularized, computational, note, plot, linear, solving, regularization, one, norm, denote, proposition, vector] [discrete, transport, kernel, two, section, finite, metric, sample, distance, wasserstein, word, density, empirical, converge, random, space] [continuous, averaged, information, corresponds, another] [using, figure, different, compare, propose, used, neural, use, shown, approach, three]
Learning Kernels with Random Features
Aman Sinha, John C. Duchi
Aman Sinha, John C. Duchi
[consistency, base, probability] [problem, learning, randomized, provide, consider, set, let, risk, convex, conference, misclassification, concentration, define, function, lemma, show, best] [optimization, method, machine, respect, iid, sampling, efficient, solve, employ, standard, requires, approximate, large, processing, speedup] [error, can, linear, regression, computational, solution, matrix, one, following] [kernel, random, generalization, data, space, procedure, empirical, joint, gaussian, distribution, estimator, section, rahimi, reuters, test, selection, well, denotes, given, sample, ridge] [optimized, time, model, benchmark, simple, information] [feature, performance, approach, training, use, alignment, figure, using, supervised, learn, original, dataset, neural, international, input]
CliqueCNN: Deep Unsupervised Exemplar Learning
Miguel A. Bautista, Artsiom Sanakoyeu, Ekaterina Tikhoncheva, Bjorn Ommer
Miguel A. Bautista, Artsiom Sanakoyeu, Ekaterina Tikhoncheva, Bjorn Ommer
[resulting, number, clique, negative, transitivity, obtained, average, many, thus] [learning, problem, set, now, since, show, query, label, wang, cost] [batch, large, method, optimization, sgd, compute, due, initial] [can, mutually, one, analysis, also, matrix, unreliable] [nearest, positive, data, estimation, sample, based] [model, information] [exemplar, training, similarity, approach, cnn, using, pose, different, similar, unsupervised, visual, object, compact, single, performance, proposed, image, posture, olympic, deep, representation, supervised, neural, classification, learned, used, pascal, feature, cnns, use, evaluate, hog, figure, train, compare, part, alexnet, compared, trained, dataset, voc, convolutional]
Adaptive optimal training of animal behavior
Ji Hyun Bak, Jung Choi, Ilana Witten, Athena Akrami, Jonathan W. Pillow
Ji Hyun Bak, Jung Choi, Ilana Witten, Athena Akrami, Jonathan W. Pillow
[rule, theory] [learning, optimal, set, expected, may, algorithm, will, adaptive, function, dependence, rate, learner, defined, toward, let] [log, method, full, gradient, likelihood, posterior, parameter, supplementary, hyperparameters, suggests] [can, vector, order, first, success, matrix] [estimate, true, space, fixed, two, given, based, random, bias, estimating, discrimination] [model, stimulus, behavior, history, trial, animal, rat, evidence, reward, prior, choice, desired, simulated, internal, effect, alignmax, policy, change, time, current, variability, reinforcement, psychometric, difficult, hyperparameter, drive, experimental, value, series, governing, correct, behavioral, infer, new, early] [training, weight, using, figure, shown, used, use, task, neural, accurately, map]
Improved Dropout for Shallow and Deep Learning
Zhe Li, Boqing Gong, Tianbao Yang
Zhe Li, Boqing Gong, Tianbao Yang
[number, definition, connected, evolving, covariate, according, present] [learning, risk, bound, smaller, let, theorem, achieves, dependent, set, upper, bernoulli, minimize, minimization, make, obtain, defined] [sampling, standard, batch, stochastic, optimization, convergence, normalization, faster, gradient, processing, logistic, usually] [can, also, error, noise, following, analysis, proposition, denote, second, order, note, one, first, vector] [data, multinomial, distribution, testing, given, theoretical, based, uniform, empirical] [internal, information, issue, experimental] [dropout, deep, evolutional, neural, training, using, different, shallow, proposed, feature, use, layer, performance, similar, network, iters, figure, propose, three, used, convolutional, compare, geoffrey, four, alex]
Using Fast Weights to Attend to the Recent Past
Jimmy Ba, Geoffrey E. Hinton, Volodymyr Mnih, Joel Z. Leibo, Catalin Ionescu
Jimmy Ba, Geoffrey E. Hinton, Volodymyr Mnih, Joel Z. Leibo, Catalin Ionescu
[rule, expression, many] [fast, learning, will, set, rate, conference] [processing, capacity, machine, normalization, term] [can, vector, one, much, product, first, need, also, computational, retrieval] [two] [model, information, state, slow, current, activity, transition, synaptic, past, time, decay, long, agent, scalar, allows, game, brain, new, process] [hidden, memory, neural, associative, attention, different, visual, recurrent, using, input, lstm, used, use, store, recent, layer, network, sequence, glimpse, task, rnn, weight, stored, learn, figure, table, storage, single, classification, temporary, convnet, performance, facial, previous, without, image, mnist, object, integrate, training, presented, shown, paddle, recognition]
Learning Parametric Sparse Models for Image Super-Resolution
Yongbo Li, Weisheng Dong, Xuemei Xie, GUANGMING Shi, Xin Li, Donglai Xu
Yongbo Li, Weisheng Dong, Xuemei Xie, GUANGMING Shi, Xin Li, Donglai Xu
[cluster, average, obtained, denoted] [learning, function, set, algorithm, conference, problem] [method, argmin, factor, solved, processing, conventional, large] [sparse, can, via, ieee, desirable, dictionary, solving, pca, recovered, also, first, matrix, sparsity, denote, regression, reconstructed, recover, anchored] [parametric, gaussian, estimated, denotes, estimate, basis, based, test, mean, selection] [prior, scaling, coding, followed, experimental] [image, learned, similar, proposed, mapping, patch, learn, used, bicubic, using, feature, ncsr, generated, computer, training, srcnn, input, performance, propose, natural, shown, psnr, international, original, extracted, downsampling, neural, novel, scsr, degradation, table, blur]
CNNpack: Packing Convolutional Neural Networks in the Frequency Domain
Yunhe Wang, Chang Xu, Shan You, Dacheng Tao, Chao Xu
Yunhe Wang, Chang Xu, Shan You, Dacheng Tao, Chao Xu
[cluster, number, larger, thus, coefficient, obtained, pruning] [compression, ratio, since, scheme, set, algorithm, will, learning, smaller, complexity] [large, method, size, although] [frequency, can, compressed, computational, matrix, also, small, sparse, note] [data, two] [operation, model] [convolutional, dct, proposed, domain, neural, filter, accuracy, deep, layer, cnns, approach, network, compressing, feature, alexnet, used, original, similar, image, using, huffman, cnn, mobile, memory, residual, cnnpack, convolution, relatively, performance, figure, impact, table, storage, arxiv, spatial, preprint, cosine, use, rdi, effective, net, weight, storing, object, applying]
Poisson-Gamma dynamical systems
Aaron Schein, Hanna Wallach, Mingyuan Zhou
Aaron Schein, Hanna Wallach, Mingyuan Zhou
[structure, negative, definition, report, subset, many] [set, conference, obtain, define, algorithm, provide] [latent, inference, sampling, bayesian, factor, auxiliary, parameter, five, draw, pass, expressive] [can, component, matrix, also, via, observed, error, one, linear, following, note, analysis] [data, distribution, gamma, two, conditional, well, nonparametric, gaussian, introduce, given] [pgds, count, model, transition, poisson, time, lds, information, therefore, dynamical, binomial, burstiness, sequentially, inferred, marginalize, process, gdelt, new, backward, alternative, prior, lkk, system, sotu, linked, marginalizing] [used, figure, top, using, feature, three, international, performance, neural]
A Unified Approach for Learning the Parameters of Sum-Product Networks
Han Zhao, Pascal Poupart, Geoffrey J. Gordon
Han Zhao, Pascal Poupart, Geoffrey J. Gordon
[spn, cccp, spns, induced, wij, sum, structure, monomial, complete, node, learnspn, decomposable, sma, number, tree, pgd, signomial, unique, fvj, program, fact, edge, often] [learning, function, convex, show, set, problem, algorithm, concave, will, let, since, equivalent, obtain, surrogate] [objective, log, optimization, convergence, parameter, update, bayesian, gradient, sequential, computed, although] [can, also, product, unified, vector, note, first, despite] [two, mixture, polynomial, data, based, univariate, random, difference, mle, procedure, distribution, maximum, expressed] [model, form, corresponds] [network, use, multiplicative, four, using, input, different, better, used]
Kernel Bayesian Inference with Posterior Regularization
Yang Song, Jun Zhu, Yong Ren
Yang Song, Jun Zhu, Yong Ren
[probability, rule, consistency, whose, variable, relation] [theorem, let, learning, problem, set, assume, show, will, since, function] [posterior, bayesian, inference, likelihood, expectation, machine] [regularization, can, following, thresholding, regression, linear, note, arg, regularized, filtering, first, formulation, via, product, also] [kernel, embeddings, rkhs, distribution, embedding, data, space, estimator, kregbayes, sample, cxx, arthur, cxy, hilbert, random, conditional, covariance, theoretical, reproducing, estimate, joint, based, given, optimizational, nonlinear, journal, consistent, two, pkbr, squared, finite] [framework, new, kalman, observation, markov] [training, used, propose, use, hidden, feature, filter, different, camera, compared, alex, international, position, china]
Toward Deeper Understanding of Neural Networks: The Power of Initialization and a Dual View on Expressivity
Amit Daniely, Roy Frostig, Yoram Singer
Amit Daniely, Roy Frostig, Yoram Singer
[node, connected, normalized, obtained, denoted, number, whose, probability, theory, incoming, induced, often] [learning, let, function, loss, show, example, every, bounded, set, theorem, assume, conference, convex, complexity] [processing, dual, machine, supplementary, log, size] [can, computation, initialization, norm, denote, note, also, analysis, ieee, following] [kernel, skeleton, random, space, corresponding, empirical, hilbert, well, polynomial, distribution, family, gaussian, theoretical, refer, reproducing, sample] [information, simple, realization, next, extend] [neural, activation, network, output, input, convolutional, layer, deep, representation, fully, used, single, training, relu, use, figure, work, supervised, sequence]
Regret Bounds for Non-decomposable Metrics with Missing Labels
Nagarajan Natarajan, Prateek Jain
Nagarajan Natarajan, Prateek Jain
[] [] [] [] [] [] []
Accelerating Stochastic Composition Optimization
Mengdi Wang, Ji Liu, Ethan Fang
Mengdi Wang, Ji Liu, Ethan Fang
[lambda] [problem, algorithm, composition, rate, function, convex, learning, optimal, case, minimization, wang, special, scgd, consider, achieves, best, complexity, general, known, theorem, show, example, choose, let, expected, necessarily, fast, online] [stochastic, convergence, gradient, objective, strongly, proximal, optimization, expectation, experiment, accelerated, inner, method, siam, machine] [linear, can, first, solution, denote, one, analysis, note, see, min, assumption, solving, error, main, regularization] [two, random, result, compositional, application, empirical, journal, denotes, given, theoretical, equation, important, section, sample] [state, reinforcement, new, bellman, policy, averaged, value, transition, reward] [proposed, generate, preprint, arxiv, using, figure, use]
Proximal Deep Structured Models
Shenlong Wang, Sanja Fidler, Raquel Urtasun
Shenlong Wang, Sanja Fidler, Raquel Urtasun
[unary, configuration, subset] [learning, function, loss, set, algorithm, problem, show, convex, considered, special, rate, exist, minimize] [proximal, inference, operator, energy, method, gradient, compute, dual, primal, pass, iteration, refinement, flownet, step, reader, size, mrfs] [can, first, note, min, arg, second, sparse] [random, family, type, refer, data, shrinkage, gaussian, given, discrete] [model, continuous, complex, forward] [deep, structured, image, depth, output, used, optical, neural, use, flow, performance, recurrent, using, convolution, approach, proposed, shown, dataset, denoising, figure, convolutional, training, learn, different, train, activation, natural, network, whole, recent, table]
Supervised Learning with Tensor Networks
Edwin Stoudenmire, David J. Schwab
Edwin Stoudenmire, David J. Schwab
[site, number, diagram, structure, thus, represent] [learning, algorithm, set, function, label, index, will, cost, optimal, consider, adaptively, implies] [large, gradient, approximation, machine, step, optimization, additional] [tensor, can, bond, matrix, one, product, vector, local, mapped, decomposition, also, svd, renormalization, error, dimension, first, small, order, singular, good, interesting, group, orthogonal] [data, test, space, basis, density, quantum, type, dimensional, two, kernel] [form, decision, model, next, optimizing, right, value, optimize, new] [feature, input, training, map, network, weight, using, use, shown, figure, classification, neural, mnist, similar, approach, different, image, used, illustrated, grayscale, pixel]
Total Variation Classes Beyond 1d: Minimax Rates, and the Limitations of Linear Smoothers
Veeranjaneyulu Sadhanala, Yu-Xiang Wang, Ryan J. Tibshirani
Veeranjaneyulu Sadhanala, Yu-Xiang Wang, Ryan J. Tibshirani
[total, theory, graph, sharp, many, often] [minimax, smoothing, rate, optimal, bound, function, lower, risk, theorem, will, class, slope, defined, bounded, learning, upper, even, known, constant, max, study, smaller, eigenmaps, now, conference, trivial, problem, lemma, setting] [log, mse, parameter, siam] [linear, also, can, see, interesting, gap, first, wavelet, via, signal, remark, though, denote] [laplacian, sobolev, estimator, mean, canonical, estimation, discrete, journal, section, nonlinear, nonparametric, given, result, annals, provides, trend, estimating, two, inscribed, ryan, space] [ball, scaling, grid] [denoising, variation, like, figure, different, image, international, using, shown, denoiser, neural]
Communication-Optimal Distributed Clustering
Jiecao Chen, He Sun, David Woodruff, Qin Zhang
Jiecao Chen, He Sun, David Woodruff, Qin Zhang
[clustering, graph, blackboard, message, passing, sparsifier, coordinator, edge, total, site, among, number, notice, centralized, david, present, partition, constructing, correctly, sends, normalized, sparsifiers, qin, undirected, probability, cluster, vertex] [algorithm, cost, theorem, let, constant, show, set, lower, every, will, since, study, define, optimal, obtain, defined, assume, bound, function, lemma, setting] [distributed, sampling, computing] [spectral, can, matrix, one, also, following, diagonal, leverage] [two, chain, data, based, constructed, laplacian, geometric, length, space, construct] [model, communication, value, broadcast, information, write] [using, different, input, used, figure, weight, similar]
Exponential Family Embeddings
Maja Rudolph, Francisco Ruiz, Stephan Mandt, David Blei
Maja Rudolph, Francisco Ruiz, Stephan Mandt, David Blei
[negative, contains, find, structure] [context, function, set, example, study, will, define, learning] [objective, stochastic, latent, sampling, gradient, factor, parameter, fit, inner, term, size] [can, one, link, also, movie, nonnegative, matrix, analysis, vector, linear, product, item, see, pca] [data, embedding, word, embeddings, conditional, exponential, point, shopping, two, family, gaussian, test, distribution, basket, mean, market, given, section, supplement, equation, surrounding, cbow, purchased, purchase, based, hpf] [poisson, model, activity, neuron, time, across, count, form, information] [neural, use, similar, different, table, language, natural, using, three, similarity, used]
Data Programming: Creating Large Training Sets, Quickly
Alexander J. Ratner, Christopher M. De Sa, Sen Wu, Daniel Selsam, Christopher Ré
Alexander J. Ratner, Christopher M. De Sa, Sen Wu, Daniel Selsam, Christopher Ré
[labeling, programming, dependency, knowledge, relation, extraction, describe, distant, crowdsourcing, base, many, number, conflict, filling] [learning, set, function, label, provide, loss, may, show, expected, will, conference, problem, risk, example, achieve, general, class, improvement] [large, machine, parameter, gradient, stochastic, processing] [can, also, user, one, first, noisy, min, error] [data, given, independent, two, distribution, section, point, empirical] [model, information, scaling, framework, new] [training, using, labeled, generative, approach, used, score, supervision, unlabeled, use, supervised, lstm, feature, neural, without, paradigm, input, dataset, accuracy, text, discriminative]
Regret of Queueing Bandits
Subhashini Krishnasamy, Rajat Sen, Ramesh Johari, Sanjay Shakkottai
Subhashini Krishnasamy, Rajat Sen, Ramesh Johari, Sanjay Shakkottai
[number, probability, cycle, thus] [regret, queueing, bound, queue, algorithm, bandit, optimal, late, service, lower, problem, regenerative, theorem, cumulative, show, arm, loaded, heavily, mab, rate, scheduling, corollary, thompson, learning, let, consider, obtain, since, upper, job, stabilize, expected, depends, instance, schedule, allocation] [log, stage, server, stochastic, standard, sampling, beginning, university] [can, analysis, one, first, phase] [given, length, two, asymptotic, regime, mean, corresponding, independent, distribution, difference] [time, early, system, exploration, policy, behavior, transition, reward, scaling, arrival, past, across] [performance, structured, different, use, like, figure, single]
Improving PAC Exploration Using the Median Of Means
Jason Pazis, Ronald E. Parr, Jonathan P. How
Jason Pazis, Ronald E. Parr, Jonathan P. How
[number, probability, possible, average, many, definition, fact] [pac, algorithm, function, will, complexity, learning, best, qmax, bounded, set, theorem, least, every, even, let, prove, cost, bound, conference, defined, inequality, known, lower, diameter, optimal, exist, max, regret, optimistic, show] [variance, operator, factor, end, expectation, approximation] [can, first, error, note, analysis, one, high, following] [sample, fixed, finite, space, estimate, discrete, mean] [exploration, policy, state, bellman, value, reward, median, action, mdps, reinforcement, transition, rather, take, decision, unbounded, mdp, concurrent, simple, pazis, markov, continuous, next, time, range, delayed] [work, using, per, available, used]
Scaling Memory-Augmented Neural Networks with Sparse Reads and Writes
Jack Rae, Jonathan J. Hunt, Ivo Danihelka, Timothy Harley, Andrew W. Senior, Gregory Wayne, Alex Graves, Tim Lillicrap
Jack Rae, Jonathan J. Hunt, Ivo Danihelka, Timothy Harley, Andrew W. Senior, Gregory Wayne, Alex Graves, Tim Lillicrap
[number, external, level] [learning, set, differentiable, since, access, scheme, defined, best, cost, constant, may] [supplementary, step, machine, large, efficient, overhead, recently, processing] [sparse, can, one, via, linear, also] [space, length, nearest, word, neighbor, test, comparable] [time, sam, write, read, model, ntm, ann, backward, wtr, operation, forward, controller, usage, curriculum, turing, copy, information, accessed, addressing] [memory, neural, dense, used, training, sequence, use, able, learn, figure, associative, using, network, priority, previous, character, perform, preprint, approach, arxiv, task, omniglot, train]
PAC-Bayesian Theory Meets Bayesian Inference
Pascal Germain, Francis Bach, Alexandre Lacoste, Simon Lacoste-Julien
Pascal Germain, Francis Bach, Alexandre Lacoste, Simon Lacoste-Julien
[negative, probability, theory, obtained, frequentist, majority, valid, number] [loss, bound, function, theorem, learning, corollary, consider, risk, study, obtain, set, least, optimal, bounded, john, equivalent, may, minimizing, show, general, upper, defined, appendix, provide, tighter] [bayesian, marginal, posterior, likelihood, factor, variance, parameter, machine, log, rely, compute] [regression, can, one, linear, note, assumption, analysis, real, first] [equation, generalization, given, distribution, section, gibbs, empirical, selection, gaussian, exp, sample, squared, data, alexandre, result, done, two, yevgeny, bayes] [model, prior, unbounded, choice, new, extend] [using, used, figure, classification, pascal, training, input]
Tree-Structured Reinforcement Learning for Sequential Object Localization
Zequn Jie, Xiaodan Liang, Jiashi Feng, Xiaojie Jin, Wen Lu, Shuicheng Yan
Zequn Jie, Xiaodan Liang, Jiashi Feng, Xiaojie Jin, Wen Lu, Shuicheng Yan
[detection, number, tree, path, level, threshold, find, average, recursively] [set, optimal, fast, best, obtain, scheme, learning, uncovered, function, active] [proposal, large, faster, sequential] [one, local, localization, can, small, also, global, group] [two, testing, based, well, searching] [search, action, agent, current, scaling, reward, state, starting, reinforcement, model, next, taking, new, taken, sequentially, value, decision] [object, window, recall, whole, proposed, multiple, image, iou, single, voc, feature, translation, different, deep, table, pascal, rpn, training, approach, using, layer, map, visual, better, figure, preprint, arxiv, generated, attended, instead]
On the Recursive Teaching Dimension of VC Classes
Xi Chen, Xi Chen, Yu Cheng, Bo Tang
Xi Chen, Xi Chen, Yu Cheng, Bo Tang
[recursive, number, sat, definition, boolean, string, present, subset, among, formula, theory, introduced] [concept, class, set, bound, tsmin, learning, every, upper, smallest, lemma, nonempty, proof, prove, lower, rtd, complexity, known, let, contradiction, vcd, cbi, shattered, kuhlmann, least, compression, learner, obtain, since, ratio, assume, conference, example, satisfiability, show, depends, follows, binary, cby, moran, always, appear, case, induction] [size, log, university, factor] [dimension, one, can, also, first, projection, denote, small, see, following, still, computational, order] [given, sample, two, family, needed] [teaching, must, teacher, desired] [figure, domain, use, computer, improve, learn, open]
Adaptive Smoothed Online Multi-Task Learning
Keerthiram Murugesan, Hanxiao Liu, Jaime Carbonell, Yiming Yang
Keerthiram Murugesan, Hanxiao Liu, Jaime Carbonell, Yiming Yang
[rule, number, knowledge, sentiment, detection, average] [learning, online, algorithm, spam, loss, hypothesis, may, set, regret, will, learner, let, argminwk, problem, show, consider, since, theorem, corollary, adaptive, setting, omtrl] [batch, optimization, updating, end, machine, university, datasets, update, large] [matrix, can, related, formulation, multitask, also, existing, paper, addition] [data, two, given, well, independent, section] [model, information, time, learns, benchmark] [task, training, relationship, pkj, proposed, learn, use, shared, similar, multiple, classification, performance, three, compared, better, different, domain, per, landmine, several, using, single, learned, consists, available, work, dataset, shamo]
Conditional Generative Moment-Matching Networks
Yong Ren, Jun Zhu, Jialian Li, Yucen Luo
Yong Ren, Jun Zhu, Jialian Li, Yucen Luo
[present, knowledge, number] [learning, function, set, appendix, class, defined, element, theorem, competitive] [bayesian, objective, stochastic, gradient, draw, size, minibatch, operator] [can, also, dimension, first, one, error, gram, interesting, via] [conditional, cgmmn, mean, kernel, distribution, embedding, data, space, sample, maximum, cmmd, discrepancy, two, estimate, given, difference, test, mmd, dark, dxy, distill, hilbert, rkhs] [model, predictive, simple, dgm, whether, range] [generative, network, dataset, training, generated, deep, hidden, using, input, learn, mnist, neural, performance, including, mlp, generating, use, image, prediction, architecture, generate, feature, various, convolutional, discriminative, layer, used, shown, unit, figure, table, cnn]
DeepMath - Deep Sequence Models for Premise Selection
Geoffrey Irving, Christian Szegedy, Alexander A. Alemi, Niklas Een, Francois Chollet, Josef Urban
Geoffrey Irving, Christian Szegedy, Alexander A. Alemi, Niklas Een, Francois Chollet, Josef Urban
[conjecture, many, number, definition, average, logic, negative] [theorem, set, proof, learning, best, defined, max, prove, since] [large, stage, evaluation] [can, rank, order, proved, first, also, relative] [selection, premise, mizar, automated, atp, mathematical, axiom, embeddings, proving, formal, itp, embedding, two, test, length, useful, based, given, corpus, formalization, formalized, pavail, word] [model, time, experimental, simple, statement, baseline] [neural, network, using, training, sequence, reasoning, figure, use, volume, fully, used, trained, convolutional, deep, recurrent, without, task, preprint, arxiv, architecture, previous, google, language, approach, different, produce]
Interpretable Nonlinear Dynamic Modeling of Neural Trajectories
Yuan Zhao, Il Memming Park
Yuan Zhao, Il Memming Park
[stable] [function, set, constant, august] [latent, around, initial, step, black, method, inference, standard, reduction, autoregressive] [computation, can, one, linear, phase, note, computational, coherence, also, global] [nonlinear, fixed, true, basis, ring, point, two, test, journal, chaotic, gaussian, locally, random, mean, finite, space, corresponding, theoretical] [model, time, dynamical, trajectory, direction, velocity, series, attractor, continuous, slow, system, simulated, within, head, stimulus, interpretable, blue, decision, driven, dynamic, current, ghost, neuroscience, november] [neural, prediction, training, input, figure, train, field, use, using, proposed, trained, network, used, recurrent, perceptual, red]
Learning Deep Embeddings with Histogram Loss
Evgeniya Ustinova, Victor Lempitsky
Evgeniya Ustinova, Victor Lempitsky
[negative, histogram, number, pairwise, probability] [loss, learning, set, conference, online, function, margin, rate, smaller, class, best, uniformly] [batch, datasets, size, sampling, large, computed, parameter] [can, ieee, also, product, mij, sampled] [positive, two, embedding, embeddings, metric, distance, sample, random, corresponding, test, estimate] [binomial, new] [deep, used, similarity, triplet, computer, use, training, figure, deviance, pair, vision, using, image, pattern, dataset, neural, person, classification, different, recognition, network, international, face, sij, last, randomly, contrastive, several, architecture, shown]
Coordinate-wise Power Method
Qi Lei, Kai Zhong, Inderjit S. Dhillon
Qi Lei, Kai Zhong, Inderjit S. Dhillon
[number, rule, social] [algorithm, show, set, greedy, function, will, learning, appendix, selecting, let, problem, strategy, even] [method, coordinate, convergence, lanczos, iterate, descent, faster, iteration, processing, optimization, select, update, machine, computing, requires, converges, large, objective] [power, matrix, cpm, sgcd, dominant, can, tan, sparse, symmetric, largest, one, also, vanilla, real, vector, see, arg, eigenvector, eigenvalue, vrpca, linear, much, global, loading, following, comparison, good, analysis] [data, selection, section, important, positive] [time, therefore, next, value] [figure, performance, part, propose, dataset, mechanism, using, better, dense, similar, neural, applied, compared, use, shown]
PerforatedCNNs: Acceleration through Elimination of Redundant Convolutions
Mikhail Figurnov, Aizhan Ibraimova, Dmitry P. Vetrov, Pushmeet Kohli
Mikhail Figurnov, Aizhan Ibraimova, Dmitry P. Vetrov, Pushmeet Kohli
[number, structure, modern, obtained] [set, cost, since, may, best, achieve, consider, function, choosing] [size, acceleration, method, speedup, increase, evaluation, reduce, decrease, reduction] [error, computational, can, matrix, tensor, see, high, first, support] [redundancy, two, data, kernel, uniform] [grid, exploit, time, value, required] [perforation, convolutional, network, layer, mask, impact, spatial, output, using, use, cpu, alexnet, neural, cnn, pooling, perforated, training, approach, table, deep, gpu, fractional, memory, several, input, perforatedcnns, proposed, used, evaluate, nin, figure, cnns, propose, combined, perform, architecture, store, calculated, lebedev, original, different]
Faster Projection-free Convex Optimization over the Spectrahedron
Dan Garber, Dan Garber
Dan Garber, Dan Garber
[probability, strong, dan] [algorithm, convex, let, problem, learning, follows, function, since, conference, set, rate, minimizing, theorem, elad, bound, optimal, setting, lemma] [method, gradient, convergence, optimization, machine, standard, faster, kvt, iteration, processing, variant, update, full, requires, strongly] [matrix, eigenvector, min, decomposition, denote, can, also, leading, linear, solving, semidefinite, computation, following, error, completion, rank, eigenvalue, analysis, one, much, norm, main, regularized, certain, nuclear, ovie, note, first, convexity, vector, largest, garber, spectrahedron] [conditional, given, distribution, important] [new, choice, information, current] [unit, international, single, neural, task, several]
An algorithm for L1 nearest neighbor search via monotonic embedding
Xinan Wang, Sanjoy Dasgupta
Xinan Wang, Sanjoy Dasgupta
[tree, average, number, possible] [query, set, hashing, algorithm, lsh, case, show, scheme, general, function, will, cost, choose, consider, rate, hash] [approximate, log, end, method, reduction] [projection, can, explicit, dimension, also, high, linear, one, rank, satisfying, practice, mapped] [data, random, nearest, embedding, neighbor, distance, metric, gaussian, two, project, space, point, erp, multivariate, corel, joint, cauchy, based, distribution, embedded, andoni, journal, embeddings, scan, given] [search, time, choice, implement, accessed, along, save] [using, use, used, table, performance, volume, generate, available, original, per, multiple]
Learning Tree Structured Potential Games
Vikas Garg, Tommi Jaakkola
Vikas Garg, Tommi Jaakkola
[tree, potential, structure, configuration, node, nij, assignment, many, edge, vote, agree, possible, graphical, supreme] [set, algorithm, learning, setting, strategy, optimal, problem, may, since, will, max, context, function, uniformly, chosen, theorem, least] [dual, method, iteration, compute, optimization, approximate, stochastic, solve] [local, global, decomposition, can, real, one, recover, synthetic, phase, following, first, note, sampled] [data, underlying, two, locally, associated, random, procedure] [game, player, across, value, strategic, form, correct, court, model, typically] [structured, prediction, using, learn, figure, training, without, dataset, different, language]
Natural-Parameter Networks: A Class of Probabilistic Neural Networks
Hao Wang, Xingjian SHI, Dit-Yan Yeung
Hao Wang, Xingjian SHI, Dit-Yan Yeung
[probabilistic, number, according] [learning, function, since, set, algorithm, show, achieve, assume, rate, class, defined, may] [variance, bayesian, compute, supplementary, datasets, inference, efficient, proxy, detailed] [can, error, link, also, note, linear, rank, first, computation, one, matrix, following, vector, support] [npn, gaussian, gamma, distribution, nonlinear, mean, data, transformation, test, equation, exponential, two, bdk, section, corresponding, estimated, random, denotes] [poisson, another, pgm, uncertainty, form, model] [natural, different, use, used, neural, deep, activation, training, dropout, performance, table, output, like, network, learned, using, figure, higher, representation, consists, input, learn]
Using Social Dynamics to Make Individual Predictions: Variational Inference with a Stochastic Kinetic Model
Zhen Xu, Wen Dong, Sargur N. Srihari
Zhen Xu, Wen Dong, Sargur N. Srihari
[social, number, probability, grows, infection, infected, transmission, collected, represents] [rate, algorithm, set, complexity, make, expected, problem] [inference, stochastic, variational, sampling, approximate, bayesian, large, linearly, log, likelihood, efficient, latent, step, tractable] [can, filtering, order, coupled, one, computation] [data, given, hmm, gibbs, introduce, positive, section, people, discrete, kernel, two] [individual, time, model, state, infectious, kinetic, event, transition, markov, dynamic, viskm, epidemic, evolution, particle, temporal, susceptible, form, collective, current, demonstrate, dartmouth, disease, sensor, modeling, backward] [hidden, three, performance, figure, using, predict, sequence, use, network, different, field, entire, shown]
Computing and maximizing influence in linear threshold and triggering models
Justin T. Khim, Varun Jog, Po-Ling Loh
Justin T. Khim, Varun Jog, Po-Ling Loh
[influence, threshold, edge, infected, graph, vertex, cascade, probability, infection, node, social, lbtrig, kempe, directed, subset, number, monotonic, preferential, live, adjacency, weighted, becomes, acm, runtime] [lower, bound, set, upper, may, theorem, greedy, maximization, algorithm, function, maximizing, general, submodular, conference, case, since, inequality, class, bji, bij, let, tighter, define, problem] [step, size] [linear, following, also, note, via, matrix, proposition, certain, denote, gap, paper, suppose] [independent, section, random, theoretical, data, two, discrete, length, maximum, selected, mathematical] [triggering, model, time, simulation, information, write] [network, international, weight, computer, applied, figure]
Anchor-Free Correlated Topic Modeling: Identifiability and Algorithm
Kejun Huang, Xiao Fu, Nikolaos D. Sidiropoulos
Kejun Huang, Xiao Fu, Nikolaos D. Sidiropoulos
[criterion, identification, mining, identifiable, thus, obtained, runtime, clustering] [algorithm, problem, since, may, considered, show, theorem, least] [optimization, size, separable, usually] [matrix, can, assumption, anchorfree, sufficiently, scattered, also, xray, nonnegative, anchor, linear, factorization, det, snpa, much, one, mined, solution, column, first, coherence, condition, cone, noise, fastachor, cec, deflation, fastanchor, still, analysis, projection, via, see] [topic, data, identifiability, document, corpus, based, spa, two, vocabulary, word, lewinsky, reuters, procedure] [model, modeling, correlation, subject] [proposed, using, compared, use, text, work, table, propose, used, better, approach]
Robust k-means: a Theoretical Revisit
ALEXANDROS GEORGOGIANNIS
ALEXANDROS GEORGOGIANNIS
[cluster, consistency, clustering, definition, influence, clear, become, valid, called, subset, number] [set, function, theorem, let, show, least, consider, rate, class, general, bounded, optimal, case, problem, defined, corollary, define, convex, every, lemma] [proximal, quadratic, unbiased, distributed, datasets, convergence] [one, error, following, min, analysis, also, can, outlier, proposition, vector, order, lie, closest] [robust, trimmed, breakdown, two, moreau, point, robustness, envelope, center, estimation, journal, positive, given, section, associated, equal, radius, modification, statistical, sample, distance, universal, based, biased, remain, mean] [form, value] [map, dataset, figure, top, different, like]
The Power of Adaptivity in Identifying Statistical Alternatives
Kevin G. Jamieson, Daniel Haas, Benjamin Recht
Kevin G. Jamieson, Daniel Haas, Benjamin Recht
[probability, number, knowledge, detection, total, bag, many, identifying, identification, presence, identify, repeat] [coin, problem, algorithm, lower, heavy, strategy, known, theorem, least, complexity, adaptive, bound, just, arm, reservoir, consider, constant, unknown, bandit, let, expected, show, difficulty, observe, max, upper, fix, hypothesis, corollary, general, whenever, bounded, absolute, armed, case, optimal, set, since, learning, may, will] [log, requires, draw, sampling] [one, can, detecting, note, first, anomaly, following] [distribution, biased, sample, mixture, two, procedure, mean, given, section, infinite, versus, random, fixed, test, exponential] [correct, prior, whether, information, player] [work, proposed, single, propose, using]
Combinatorial Multi-Armed Bandit with General Reward Functions
Wei Chen, Wei Hu, Fu Li, Jian Li, Yu Liu, Pinyan Lu
Wei Chen, Wei Hu, Fu Li, Jian Li, Yu Liu, Pinyan Lu
[base] [algorithm, arm, expected, regret, super, problem, general, utility, function, bound, combinatorial, cmab, online, known, set, learning, upper, outcome, bandit, confidence, case, offline, may, cumulative, round, supported, maximizing, provide, let, depends, assume, optimal, consider, obtain, constant, sdcb, mab, show, oracle, cdf, just, theorem, maximization, ptas, chooses] [stochastic, parameter, optimization, supplementary, stochastically, approximation, end] [can, assumption, also, first, one, linear, note, support, high, dominant, vector] [distribution, random, finite, section, nonlinear, given] [reward, framework, player, time, wei, value, typically] [used, use, work, input, previous]
An equivalence between high dimensional Bayes optimal inference and M-estimation
Madhu Advani, Surya Ganguli
Madhu Advani, Surya Ganguli
[thus, find, message, equivalence, passing] [optimal, loss, smoothing, smoothed, opt, equivalent, special, case, problem, function, convex, setting, unknown, dependent] [inference, step, log, descent, bayesian, gradient, proximal, variance, iid, approximate, likelihood, optimization, posterior] [signal, measurement, mamp, mmse, bamp, regularizer, noise, gamp, via, can, sparse, high, lopt, linear, related, also, amp, compressed, component, qht, rigorous, error, note, one, lasso, see] [gaussian, moreau, distribution, given, fixed, correction, mean, section, dimensional, additive, statistical, derivation, random, data, estimation, envelope] [time, choice, information, found, determine, evolution] [map, performance, using, dense, input, output, used, neural]
Breaking the Bandwidth Barrier: Geometrical Adaptive Entropy Estimation
Weihao Gao, Sewoong Oh, Pramod Viswanath
Weihao Gao, Sewoong Oh, Pramod Viswanath
[degree, science, consistency, basic] [show, theorem, defined, dependent, case, special, proof, maximization, adaptive, depends, general, known, unknown] [log, likelihood, standard, step, large, key] [local, can, order, one, main, small, following, also] [estimator, bandwidth, entropy, density, mutual, estimation, bias, independent, sample, asymptotic, nearest, neighbor, kernel, fixed, estimate, based, data, underlying, distribution, random, estimating, section, gaussian, numerical, theoretical, resubstitution, nonparametric, empirical, distance, boundary, polynomial, family, uniform, mathematical, result, joint, kozachenko] [information, choice, form, closed, correlation, state, simple] [proposed, use, using, work, recent, figure, approach, outperforms, used, art, unit]
Normalized Spectral Map Synchronization
Yanyao Shen, Qixing Huang, Nati Srebro, Sujay Sanghavi
Yanyao Shen, Qixing Huang, Nati Srebro, Sujay Sanghavi
[graph, probability, normalized, connected] [algorithm, problem, show, complexity, consider, lemma, proof, bound, conference, yet, pij, theorem, convex] [optimization, step, method, initial, processing, irregular] [normspecsync, matrix, xij, noise, sdp, can, specsync, synchronization, eigenvectors, cmu, spectral, leading, recovery, gsparse, diffsync, generalized, first, analysis, much, see, synthetic, second, noisy, sparse, min, ujt, stability, also, ransac, following, ieee, solving, real, exact] [permutation, described, density, distance, two, data, underlying, section, theoretical, random] [observation, model, varying, time] [input, map, object, figure, computer, similar, using, consists, proposed, quantitative, performance, evaluate, pattern, generate, vision]
Fast and Flexible Monotonic Functions with Ensembles of Lattices
Mahdi Milani Fard, Kevin Canini, Andrew Cotter, Jan Pfeifer, Maya Gupta
Mahdi Milani Fard, Kevin Canini, Andrew Cotter, Jan Pfeifer, Maya Gupta
[monotonic, lattice, ensemble, base, calibrators, monotonicity, calibration, rtl, number, subset, torsion, larger] [set, learning, function, provide, problem, algorithm, inequality, defined, selecting, consider] [machine, evaluation, size, experiment, respect, step, constrained, datasets, optimization, large, key, method, extreme] [can, linear, one, regression, see, first, error, regularizer, small] [random, test, two, joint, testing, journal, useful, mean, selection] [model, time, optimized] [feature, training, table, accuracy, used, dataset, separate, shared, different, neural, validation, train, trained, using, better, randomly, figure, single, learn, jointly, classification, compared, use, similar, per, propose]
CRF-CNN: Modeling Structured Information in Human Pose Estimation
Xiao Chu, Wanli Ouyang, hongsheng Li, Xiaogang Wang
Xiao Chu, Wanli Ouyang, hongsheng Li, Xiaogang Wang
[passing, message, tree, structure, among, flooding, loopy, graphical, probabilistic, node, pairwise, kong, chen, hong, obtained, graph] [learning, algorithm, scheme, receive, obtain, set] [latent, efficient, implementation, term, end, factor] [can, also, order, group, denote] [joint, two, estimation, denotes, estimated] [model, human, information, modeling, location, connecting] [body, pose, feature, structured, cnn, use, neural, used, deep, using, network, crf, training, output, convolutional, part, dataset, relationship, shown, lsp, table, hidden, spatial, proposed, compared, performance, image, vgg, map, preprint, different, implemented, arxiv, score, person, layer, without]
Online Pricing with Strategic and Patient Buyers
Michal Feldman, Tomer Koren, Roi Livni, Yishay Mansour, Aviv Zohar
Michal Feldman, Tomer Koren, Roi Livni, Yishay Mansour, Aviv Zohar
[obtained, probability, number, half, acm] [price, regret, algorithm, seller, buyer, revenue, bound, patience, lower, pricing, switching, consider, cost, set, posted, every, regrett, theorem, proof, assume, online, valuation, case, observe, may, best, bandit, loss, optimal, let, obtain, upper, show, expected, buy, setting, problem, constant, implies, mab, even, define, since, day, exists, might, choose] [reduction, size, step, log, full, stochastic, epoch, iteration, respect, university] [can, note, one, main, item, order, following] [given, fixed, purchase, section, two, result] [time, feedback, strategic, model, value, information, within, receives, continuous] [sequence, window, single, work, different, mechanism, use, neural]
GAP Safe Screening Rules for Sparse-Group Lasso
Eugene Ndiaye, Olivier Fercoq, Alexandre Gramfort, Joseph Salmon
Eugene Ndiaye, Olivier Fercoq, Alexandre Gramfort, Joseph Salmon
[obtained, level, introduced, rule, whose] [feasible, algorithm, set, defined, max, active, optimal, appendix, duality, convex, case, learning, since, strategy, else] [dual, screening, efficient, primal, term, compute, coordinate, convergence, sequential, descent, reduction, optimization] [lasso, norm, gap, group, can, qwg, one, sparsity, solution, computation, sparse, following, proposition, vector, regression, pxq, thanks, xgj, note, high, see, need, regularization, arg, climate, critical, first, computational, matrix] [point, data, given, section, sphere, two, corresponding, radius, equation] [safe, time, ball, value, new, dynamic] [figure, feature, using, natural, sequence, region, propose, proposed, unit]
Catching heuristics are optimal control policies
Boris Belousov, Gerhard Neumann, Constantin A. Rothkopf, Jan R. Peters
Boris Belousov, Gerhard Neumann, Constantin A. Rothkopf, Jan R. Peters
[theory, resulting, probability] [optimal, will, function, constant, cost, problem, show, ratio, fast, maximal] [fit, acceleration, away, optimization, run] [can, running, noise, tan, linear, computational, sufficiently, observed] [point, based, journal, described, tangent, two, distance, gaussian] [ball, agent, time, angle, control, catching, model, uncertainty, human, belief, interception, state, trajectory, reactive, system, policy, behavior, gaze, internal, tracking, predictive, future, bearing, reaction, backwards, towards, baseball, turn, intercept, catch, explain, direction, heuristic, catcher, planning, elevation, cancellation, motor, continuous, optic, outfielder, modeling, simulation, long, action, observation, force, speed, simple] [figure, using, optical, task, final, different, visual, prediction]
Hierarchical Question-Image Co-Attention for Visual Question Answering
Jiasen Lu, Jianwei Yang, Dhruv Batra, Devi Parikh
Jiasen Lu, Jianwei Yang, Dhruv Batra, Devi Parikh
[level, many, hierarchy, ablation, number, recursively] [set, problem] [parallel, method, full, machine] [alternating, can, first, vector, also, see, following] [word, two, based, embedding, given, test, space] [model, answer, hierarchical, light, stop, encoding, information, research, focused] [question, image, attention, visual, vqa, phrase, lstm, neural, propose, color, use, three, proposed, dataset, answering, feature, different, mechanism, language, oursa, used, table, snow, representation, top, using, layer, novel, pooling, convolution, preprint, arxiv, better, affinity, spatial, attend, jointly, dhruv, similar, final, devi, deep, four, attended, figure, recent, holding, attends, train, bird]
Local Similarity-Aware Deep Feature Embedding
Chen Huang, Chen Change Loy, Xiaoou Tang
Chen Huang, Chen Change Loy, Xiaoou Tang
[negative, mining, pairwise, many, thus, lifted] [learning, loss, margin, class, set, function, hinge, max, absolute, online, adaptive] [large, method, standard, select] [hard, can, local, retrieval, global, also, one, small] [embedding, metric, euclidean, distance, sample, positive, two, space, distribution, data, mean, test, mahalanobis] [new, heterogeneous, towards] [feature, similarity, deep, pddm, figure, image, score, using, learned, transfer, quadruplet, cnn, use, used, imagenet, proposed, pair, triplet, training, unit, structured, visual, contrastive, vision, intraclass, classification, shown, trained, learn, position, jointly, network, table, propose, effective, performance]
Provable Efficient Online Matrix Completion via Non-convex Stochastic Gradient Descent
Chi Jin, Sham M. Kakade, Praneeth Netrapalli
Chi Jin, Sham M. Kakade, Praneeth Netrapalli
[runtime, neighborhood, number, probability] [algorithm, online, lemma, theorem, will, let, offline, case, problem, show, complexity, learning, general, function, rate, proof, always, set, provide, fast, max, convex, consider, make, since, every, observe] [log, gradient, sgd, saddle, convergence, descent, stochastic, initial, update, step, efficient, away, warm, however] [matrix, rank, completion, can, local, psd, low, denote, also, svd, linear, phase, stay, symmetric, first, one, main, row, benjamin, good, via, see, order, def, analysis, computational] [sample, section, estimate, given, result] [time, start, framework] [preprint, arxiv, region, work, use, using, several]
The Generalized Reparameterization Gradient
Francisco R. Ruiz, Michalis Titsias RC AUEB, David Blei
Francisco R. Ruiz, Michalis Titsias RC AUEB, David Blei
[probabilistic, variable, report] [function, algorithm, set, obtain, learning, conference, class, show, general] [variational, gradient, reparameterization, advi, bbvi, latent, stochastic, gcorr, depend, apply, variance, standard, elbo, respect, inference, log, monte, grep, carlo, machine, method, beta, posterior, processing, logit, compute, znk, technique, optimization, expectation, nonconjugate, auxiliary, olivetti, kingma, fit] [can, generalized, also, one, xnd] [distribution, gamma, transformation, gaussian, estimate, family, two, random, test, sample, exponential, density, standardization] [model, transformed, information, time, form, control, artificial] [use, neural, score, omniglot, shape, using, used, mnist, figure, international, approach, different, deep, single, consists, dataset]
Split LBI: An Iterative Regularization Path with Structural Sparsity
Chendi Huang, Xinwei Sun, Jiechao Xiong, Yuan Yao
Chendi Huang, Xinwei Sun, Jiechao Xiong, Yuan Yao
[split, consistency, path, partial, structural, theory, enough, strong, panel, called, stanley, graphical] [let, algorithm, theorem, equivalent, will, define, consider, set, now, constant, smallest, example, defined, minimax] [standard, log, parameter, large, supplementary, iteration] [assumption, lbi, can, lasso, regularization, generalized, sparsity, condition, sign, iterative, order, genlasso, matrix, following, sparse, fiba, error, irrepresentable, ranking, auc, linear, computational, remark, weaker, singular, nonzero, recovery, comparison, one, see, support, also, admm, note, subspace] [selection, journal, fused, based, estimator, bregman, family] [model, right, left, value, information, basketball, design, simple] [figure, image, shown, top, better, denoising, compare]
Semiparametric Differential Graph Models
Pan Xu, Quanquan Gu
Pan Xu, Quanquan Gu
[graph, graphical, gene, cancer, expression, number, present, dependency, genetic] [precision, rate, oracle, setting, theorem, show, defined, will, property, concave, since, minimax] [nonconvex, latent, log, faster, large, parameter, likelihood, convergence] [penalty, matrix, can, assumption, group, also, nonzero, analysis, following, order, magnitude, norm, note, lasso, support, condition, synthetic, sparse] [differential, estimator, estimation, two, transelliptical, dpm, gaussian, semiparametric, distribution, sepglasso, mcp, difference, data, estimating, section, true, statistical, comp, estimate, based, random, tau, elliptical, covariance, ovarian, journal, molecular, clinical, mild] [model, research, correlation, world] [network, proposed, different, figure, propose, preprint, arxiv]
Joint quantile regression in vector-valued RKHSs
Maxime Sangnier, Olivier Fercoq, Florence d'Alché-Buc
Maxime Sangnier, Olivier Fercoq, Florence d'Alché-Buc
[thus, called, probability, decomposable] [learning, loss, let, problem, function, algorithm, set, since, property, risk, theorem, appendix, conference, now, convex] [coordinate, machine, datasets, descent, standard, optimization, dual, efficient, method] [regression, can, linear, one, also, order, matrix, vector, comparison, hard, following, remark] [quantile, kernel, joint, conditional, crossing, estimation, quantiles, pinball, two, curve, random, journal, rkhs, space, empirical, theoretical, based, data, estimator, sample, independent, estimating, given, mean, numerical, true, associated, provided, estimated] [intercept, framework, along] [proposed, output, training, approach, multiple, feature, figure, several, different, use, table, using, used, novel, work, weight]
Assortment Optimization Under the Mallows model
Antoine Desir, Vineet Goyal, Srikanth Jagabathula, Danny Segev
Antoine Desir, Vineet Goyal, Srikanth Jagabathula, Danny Segev
[probability, subset, number, expression, sum, program, pairwise] [set, general, problem, algorithm, theorem, chosen, optimal, outside, let, show, learning, price, revenue, consider, class, known, choosing, offered, always, studied, highest, ranked, assume, expected] [optimization, computing, parameter, compute, machine, large, key, efficiently, solve, logit, sampling] [product, can, ranking, running, preference, demand, order, formulation, also, solution, following, first, note, existing, customer, rank] [mixture, distribution, given, two, exponential, random, commonly, equal, fixed, permutation] [choice, model, offer, assortment, mip, time, management, option, location, universe, substitution, determining, preferred, inserted] [position, using, work, scale, approach]
Dynamic Network Surgery for Efficient DNNs
Yiwen Guo, Anbang Yao, Yurong Chen
Yiwen Guo, Anbang Yao, Yurong Chen
[pruning, pruned, number, connected, thus] [compression, learning, will, function, rate, loss, make, problem, since, set, achieves, even, learnable, algorithm, binary, may] [method, parameter, efficient, importance, reduce, update, updating, gradient] [can, error, also, matrix, compressed, order, following, much, sparse, one] [reference, two, section, intel, test, corresponding] [model, dynamic, process, current, making, policy] [network, neural, convolutional, deep, figure, training, surgery, prediction, han, connection, accuracy, table, splicing, alexnet, better, dnn, compress, fully, shown, hidden, propose, classification, layer, without, proposed, shall, use, consists, trained, improve, iter, china, retraining]
Stochastic Gradient MCMC with Stale Gradients
Changyou Chen, Nan Ding, Chunyuan Li, Yizhe Zhang, Lawrence Carin
Changyou Chen, Nan Ding, Chunyuan Li, Yizhe Zhang, Lawrence Carin
[number, worker, theory, average, indicates, level] [loss, defined, optimal, theorem, learning, bound, setting, algorithm, function, constant, appendix, bounded] [stochastic, distributed, variance, gradient, mse, stale, speedup, convergence, bayesian, standard, staleness, decrease, parameter, parallel, iteration, asynchronous, mcmc, due, langevin, faster, log, monte, server, optimization, maxl, size, computed, discussed, verify, step, posterior] [linear, can, running, technical, order, one] [bias, section, independent, integrator, test, based, increasing, estimation, data, chain, two, described, sample] [time, system, simple, model, typically] [deep, using, neural, architecture, figure, used, multiple, use, instead]
High resolution neural connectivity from incomplete tracing data using nonnegative spline regression
Kameron D. Harris, Stefan Mihalas, Eric Shea-Brown
Kameron D. Harris, Stefan Mihalas, Eric Shea-Brown
[number, find, present, site] [problem, since, algorithm, version, will, smoothing, unknown, function, assume] [full, method, fit, operator, inference, term, optimization] [connectivity, rank, matrix, injection, low, regional, mouse, allen, can, ninj, nonnegative, tracing, regression, also, projection, institute, completion, relative, solution, mesoscale, spatially, error, wfull, wtrue, throughout, local, rny, linear, spline, product, mserel, signal] [data, result, space, journal, test, estimator, laplacian, well] [brain, within, voxels, model, cortical, cortex, target, cell, system, varying] [visual, voxel, using, source, neural, use, similar, without, image, training, applied, available, field, spatial, region, map, higher, output]
Boosting with Abstention
Corinna Cortes, Giulia DeSalvo, Mehryar Mohri
Corinna Cortes, Giulia DeSalvo, Mehryar Mohri
[base, ensemble, present, threshold, rule] [abstention, algorithm, loss, learning, rejection, convex, function, lsb, problem, general, appendix, cost, surrogate, defined, lmb, set, expected, dhl, let, scenario, since, max, theorem, define, optimal, assume, best, finding, tsb, consider, hold, stump, price, binary, show] [boosting, standard, optimization, descent, log, step, coordinate, size, argmin] [can, following, solution, error, via, analysis, projected, first, note] [based, predictor, given, two, section, family, data, sample, well, distribution, introduce, bayes, test] [model, direction, new, option, search, decision, value, along] [classification, figure, pair, reject, using, classifier, learned, several, consists, natural]
Gaussian Process Bandit Optimisation with Multi-fidelity Evaluations
Kirthevasan Kandasamy, Gautam Dasarathy, Junier B. Oliva, Jeff Schneider, Barnabas Poczos
Kirthevasan Kandasamy, Gautam Dasarathy, Junier B. Oliva, Jeff Schneider, Barnabas Poczos
[number, querying, many, queried] [fidelity, regret, query, appendix, set, bandit, will, problem, strategy, lower, algorithm, let, bound, upper, function, capital, learning, confidence, best, maximise, smaller, consider, ucb, oracle] [expensive, optimization, bayesian, likelihood, optimisation, method, cheap, posterior, size, due, standard, large, implementation] [can, also, first, high, analysis, second, small, one, optimum, via, real, synthetic, error, global] [gaussian, theoretical, kernel, maximum, two, bad, section] [time, goal, process, simple, tuning, information, next, payoff, demonstrate] [single, using, used, use, training, different, figure, previous, computer, region, work]
Testing for Differences in Gaussian Graphical Models: Applications to Brain Connectivity
Eugene Belilovsky, Gaël Varoquaux, Matthew B. Blaschko
Eugene Belilovsky, Gaël Varoquaux, Matthew B. Blaschko
[graphical, edge, many, often] [confidence, obtain, learning, consider, known, now, show, precision, will, setting, hypothesis, problem, set, case, general] [parameter, end, unbiased, method] [lasso, can, sparsity, sparse, matrix, regression, connectivity, one, power, analysis, projected, group, support, second] [debiased, fused, difference, functional, gaussian, two, data, estimate, estimator, statistical, test, permutation, ridge, parametric, estimation, bias, testing, distribution, selection, covariance, significance, autism, given, section, joint, sample, comparing, ggm, well] [brain, fmri, control, inverse, across, van] [use, using, used, figure, work, network, approach, different, dataset, recent]
An ensemble diversity approach to supervised binary hashing
Miguel A. Carreira-Perpinan, Ramin Raziperchikolaei
Miguel A. Carreira-Perpinan, Ramin Raziperchikolaei
[number, ensemble, subset, bit, disjoint, neighborhood] [hash, binary, hashing, precision, function, diversity, ilht, learning, set, kshcut, ksh, ynm, case, may, loss, bre, bagging, algorithm, fast, lsh, since, show, will] [objective, optimization, faster, large, approximate, quadratic] [can, linear, also, one, much, alternating, matrix, orthogonality, vector, coupled, retrieval, solution] [random, laplacian, test, hamming, data, well, independent, kernel, two, nearest] [information, optimize, far, decision, optimizing] [training, using, different, use, supervised, neural, output, single, train, classifier, image, better, svms, work, cifar, recall, approach]
Learning shape correspondence with anisotropic convolutional neural networks
Davide Boscaini, Jonathan Masci, Emanuele Rodolà, Michael Bronstein
Davide Boscaini, Jonathan Masci, Emanuele Rodolà, Michael Bronstein
[partial, diffusion, triangle, many] [learning, defined, function, pointwise, follows, minimizing] [method, challenging, operator, processing, supplementary] [spectral, yes, local, can, error, matrix, one, denote] [geodesic, point, construction, geometric, geometry, reference, euclidean, radius, done, lscnn, two, well, random, tangent, functional, faust, refer, riemannian] [intrinsic, soft, unlike, model, template] [correspondence, shape, anisotropic, acnn, convolutional, neural, used, heat, different, applied, using, figure, deep, cnn, computer, training, gcnn, network, approach, filter, compared, output, surface, cnns, performance, representation, proposed, recent, deformable, outperforms, add, mesh, input, matching, work, previous, patch, several]
Brains on Beats
Umut Güçlü, Jordy Thielen, Michael Hanke, Marcel van Gerven, Marcel A. J. van Gerven
Umut Güçlü, Jordy Thielen, Michael Hanke, Marcel van Gerven, Marcel A. J. van Gerven
[number, assignment, assigned] [show, highest, sensitive, revealed, set] [music, posterior, gradient, analyzed, processing] [analysis, first, tag, musical, matrix, significantly, correlated, spectrum] [mean, corresponding, functional, test, random, data, two] [model, stg, representational, rdms, voxels, sensory, brain, auditory, control, rdm, target, human, temporal, fmri, stimulus, anterior, along, information, gyrus, subject, increasingly, followed, found, representing, time, value, ventral, whereas, significant, fdr, across] [dnn, neural, layer, performance, used, visual, candidate, similarity, deep, figure, region, dataset, shown, different, entire, convolutional, prediction, compared, deeper, training, predict, architecture, audio, dnns, dissimilarity]
Dynamic matrix recovery from incomplete observations under an exact low-rank constraint
Liangbei Xu, Mark Davenport
Liangbei Xu, Mark Davenport
[number, weighted, constraint, definition, program, nmin] [case, set, will, complexity, assume, bound, theorem, minimization, problem, optimal, consider, let, since, context, provide, convex] [operator, convergence, factor, parameter] [matrix, recovery, can, lowems, error, one, completion, measurement, sensing, noise, also, alternating, perturbation, recover, rank, note, following, min, analysis, norm, denote, recommendation, linear, arg, nuclear, rmse, ieee, via, synthetic, exact, incoherence, suppose, noisy] [sample, two, estimator, reduces, given, section, result, gaussian, fixed, theoretical] [baseline, dynamic, time, information, required, simple, model] [approach, using, use, compared, figure, proposed, accuracy, validation, similar, recent, relatively]
Graphical Time Warping for Joint Alignment of Multiple Curves
Yizhi Wang, David J. Miller, Kira Poskanzer, Yue Wang, Lin Tian, Guoqiang Yu
Yizhi Wang, David J. Miller, Kira Poskanzer, Yue Wang, Lin Tian, Guoqiang Yu
[warping, graph, dtw, path, gtw, cut, directed, edge, labeling, valid, reverse, definition, ggtw, neighboring, structure, astrocyte, dependency, find, node, propagation, gstruct, connected, induced, graphical, lmf, align, programming] [set, max, algorithm, problem, function, cost, lemma, may, conference, defined, since, example, case] [primal, dual, capacity, parameter, solved, hyperparameters] [can, one, also, imaging, noise, global, analysis, solution, calcium, comparison] [data, two, given, corresponding, curve, distance, reference, joint, equation, estimated] [time, series, dynamic, another, area, transformed, form] [flow, pair, multiple, alignment, single, different, pattern, shown, using, jointly, international, computer, approach, applied, used, similar]
Deep Learning Models of the Retinal Response to Natural Scenes
Lane McIntosh, Niru Maheswaranathan, Aran Nayebi, Surya Ganguli, Stephen Baccus
Lane McIntosh, Niru Maheswaranathan, Aran Nayebi, Surya Ganguli, Stephen Baccus
[circuit, precise, david] [function, general] [variance, glm] [noise, second, first, can, computational, accurate, linear, also, generalized, one] [mean, data, nonlinear, distribution, particular, journal] [retinal, model, response, stimulus, spike, white, sensory, spatiotemporal, ganglion, time, firing, variability, temporal, count, found, encoding, activity, markus, poisson, spiking, retina, early, cell, stephen, internal, nature, recorded] [natural, neural, convolutional, cnns, cnn, layer, figure, trained, network, capture, spatial, performance, better, used, learned, scene, using, filter, understanding, recurrent, generalize, adaptation, deep, contrast, understand, predicting, training, compared, shown, similar, different, generate, resolution, including, feedforward, architecture, work]
Review Networks for Caption Generation
Zhilin Yang, Ye Yuan, Yuexin Wu, William W. Cohen, Ruslan R. Salakhutdinov
Zhilin Yang, Ye Yuan, Yuexin Wu, William W. Cohen, Ruslan R. Salakhutdinov
[negative, probability, added, theory, labeling, identifiable] [class, algorithm, learning, set, let, label, setting, general, defined, function, lemma, theorem, binary, follows, will] [posterior, size] [noise, can, also, noisy, one, first, related, component] [data, positive, estimation, sample, mixture, true, proportion, nonparametric, parametric, univariate, distribution, practical, identifiability, estimating, mixing, estimate, density, transformation, random, alphamax, space, gaussian, msgmm, dimensionality, section, robust, statistical, equation, two, maximum] [prior, form, model] [unlabeled, labeled, classification, using, used, approach, input, use, classifier, learn, performance, different, explicitly, pair, generate]
Active Learning from Imperfect Labelers
Songbai Yan, Kamalika Chaudhuri, Tara Javidi
Songbai Yan, Kamalika Chaudhuri, Tara Javidi
[number, probability, membership, querying, threshold, find] [algorithm, learning, active, abstention, query, labeler, label, complexity, rate, theorem, lower, function, hypothesis, return, least, conference, setting, close, class, learner, optimal, case, let, instance, interval, heck, will, near, constant, checksignificant, upper, kamalika, pac, agnostic, binary, abstain, adaptive, consider, satisfies, fragment, problem, since] [end, log, smooth, variance] [condition, can, noise, noisy, one, also, small, satisfying, significantly, analysis] [boundary, space, statistically, closer, random, two, based, given, statistical, distance, universal] [decision, information, response, model, search, significant] [ground, output, work, truth, classifier, figure, use, used]
Human Decision-Making under Limited Time
Pedro A. Ortega, Alan A. Stocker
Pedro A. Ortega, Alan A. Stocker
[theory, resulting, average, made] [utility, expected, function, set, optimal, regret, bounded, version, setting] [posterior, log, experiment, fit, parameter, university, processing, normalizing, solve] [decomposition, one, can, solution, first, relative, also, percentage, following] [test, distribution, two, empirical, given, mutual, well, exp, maximum, derived, gibbs, determined, based, increasing, data] [choice, inverse, model, prior, information, stimulus, resource, temperature, time, subjective, human, subject, rational, puzzle, trial, decision, correct, change, value, economic, inferred, within, duration] [using, performance, training, used, figure, shown, neural, learned, conducted, calculated, different, untrained, task]
Unsupervised Domain Adaptation with Residual Transfer Networks
Mingsheng Long, Han Zhu, Jianmin Wang, Michael I. Jordan
Mingsheng Long, Han Zhu, Jianmin Wang, Michael I. Jordan
[dan, made, block, adding] [learning, function, adaptive, show, will, make, best] [standard, method, parameter, approximate, key] [can, tensor, penalty, perturbation, paper, assumption, one, hence] [data, mmd, entropy, kernel, bridge, discrepancy, well, based] [target, new, state, across, directly, adapting, model, module, abstract, framework, prior] [domain, deep, classifier, adaptation, residual, source, transfer, feature, rtn, labeled, learn, multiple, transferable, layer, different, unsupervised, using, neural, training, network, art, figure, convolutional, accuracy, approach, shown, enable, perform, learned, alexnet, work, previous, dataset, better, unlabeled, jointly, recent, effective, applied, input, propose, proposed, trained]
Finite-Dimensional BFRY Priors and Variational Bayesian Inference for Power Law Models
Juho Lee, Lancelot F. James, Seungjin Choi
Juho Lee, Lancelot F. James, Seungjin Choi
[average, stable, normalized, many, pyp, variable, law, crp, number, obtained] [algorithm, let, class, get, learning, proof, since, bounded] [variational, beta, bayesian, easily, datasets, inference, log, posterior, computed, develop, likelihood, efficient, machine, tractable, supplementary, latent] [can, also, one, synthetic, explicit, first] [random, bfry, measure, mixture, bayes, based, collapsed, gamma, density, dirichlet, crm, laplace, converge, finite, ggp, journal, statistical, fnspm, functional, data, buffet, fdpm, construct, indian, gibbs, ideal, test, sbp, independent, infinitely, kingman, six, divisible, corresponding, marginalized, dimensional] [process, written, simple, model] [using, generated, figure, compared, performance, table, relationship]
Optimistic Gittins Indices
Eli Gutin, Vivek Farias
Eli Gutin, Vivek Farias
[probability, number, frequentist] [gittins, index, regret, arm, problem, discount, ogi, bandit, optimistic, thompson, optimal, algorithm, will, horizon, bound, ucb, proof, bernoulli, let, consider, show, lower, expected, lemma, mab, theorem, constant, class, set, learning, agrawal, stopping, even] [bayesian, factor, sampling, approximation, parameter, experiment, large, beta, log, posterior, solves] [denote, one, computational, analysis, sufficiently, can, paper, also, following] [bayes, gaussian, mean, infinite, result, equation, two, given, random, finite, distribution] [time, policy, prior, reward, information, state, discounted, value, simplest, tuning, therefore] [use, performance, using, proposed, recent, applied]
Cooperative Inverse Reinforcement Learning
Dylan Hadfield-Menell, Stuart J. Russell, Pieter Abbeel, Anca Dragan
Dylan Hadfield-Menell, Stuart J. Russell, Pieter Abbeel, Anca Dragan
[sum, induced] [optimal, learning, best, problem, function, maximizes, regret, theorem, let, consider, algorithm, show, set, make, general, will, expected, may, complexity, case] [compute, computing, approximate, initial] [can, observed, assumption, phase, solution, reduced, one, first, solving] [distribution, section, two, joint, given, maximum, common, difference, mean] [reward, policy, robot, cirl, game, human, state, irl, inverse, reinforcement, behavior, value, response, apprenticeship, demonstration, cooperative, model, action, deployment, maximize, world, trajectory, expert, pomdp, information, dbe, belief, design, expects, prior, learns, infers, arises] [pair, use, feature, alignment, better, figure, ground, truth]
Path-Normalized Optimization of Recurrent Neural Networks with ReLU Activations
Behnam Neyshabur, Yuhuai Wu, Ruslan R. Salakhutdinov, Nati Srebro
Behnam Neyshabur, Yuhuai Wu, Ruslan R. Salakhutdinov, Nati Srebro
[path, number, adding, node, proceeding] [learning, function, problem, conference, show, consider, feasible, define, even] [optimization, term, gradient, parameter, step, size, respect, plain, standard, descent, update, sgd, calculating, large, calculate] [can, also, second, invariant, first, error, initialization, addition] [length, geometry, two, section, test, transformation] [time, modeling, therefore, long, internal, win] [neural, rnn, recurrent, shared, rnns, input, feedforward, output, using, different, network, hidden, relu, sequence, training, activation, table, wrec, international, language, use, better, performance, including, irnn, calculated, deep, train, wout, implemented]
Bootstrap Model Aggregation for Distributed Statistical Learning
JUN HAN, Qiang Liu
JUN HAN, Qiang Liu
[number, average, total, probabilistic, weighted] [learning, since, set, fix, max, show, problem, rate] [bootstrap, method, mse, distributed, log, averaging, size, variance, ppca, parameter, applicable, liu, large, efficient, reduction, ihler, due, machine, latent, likelihood] [local, linear, can, dimension, global, arg, first, see, need, also, assumption, note, exact, real, analysis] [estimator, data, mixture, practical, based, asymptotic, theoretical, true, sample, mle, gmm, empirical, estimated, section, statistical, consistent, estimating, given, positive] [model, control, information, simple, communication, world, varying] [different, using, figure, shown, use, dataset, better, learn, combined, original, approach]
Proximal Stochastic Methods for Nonsmooth Nonconvex Finite-Sum Optimization
Sashank J. Reddi, Suvrit Sra, Barnabas Poczos, Alexander J. Smola
Sashank J. Reddi, Suvrit Sra, Barnabas Poczos, Alexander J. Smola
[number, average] [algorithm, complexity, convex, constant, theorem, obtain, problem, corollary, function, optimal, let, known, set, learning, assume, general, show, cost, since, alexander, best, studied] [rox, nonconvex, convergence, stochastic, vrg, proximal, ifo, gradient, aga, nonsmooth, minibatch, size, step, method, variance, optimization, minibatches, machine, iteration, smooth, variant, suvrit, descent, end, sashank, faster, svrg, due, strongly, converges, siam, reduction, incremental, rfi, key, full, inner, saga] [following, can, solution, also, analysis, linear, one, first, note, much, global] [point, theoretical, result, journal, stationary, important] [typically] [using, use, performance, used, similar, output, figure, work, multiple]
Tagger: Deep Unsupervised Perceptual Grouping
Klaus Greff, Antti Rasmus, Mathias Berglund, Tele Hao, Harri Valpola, Juergen Schmidhuber
Klaus Greff, Antti Rasmus, Mathias Berglund, Tele Hao, Harri Valpola, Juergen Schmidhuber
[connected, many, number] [learning, even, example, class, cost, fast, make, conference] [inference, posterior, iteration, amortized, end, method, efficient] [can, group, iterative, tag, also, error, one, first] [two, based, parametric, result, given] [model, system, framework, baseline, internal, information, complex] [grouping, neural, tagger, network, using, unsupervised, fully, use, dataset, ladder, classification, trained, denoising, figure, table, training, segmentation, input, different, object, textured, perceptual, ami, preprint, zki, score, similar, attention, convolutional, help, train, arxiv, used, digit, learn, mapping, evaluate, mnist, supervised, generative, task, international, rnn, iter, texture, layer, without]
Learning Treewidth-Bounded Bayesian Networks with Thousands of Variables
Mauro Scanagatta, Giorgio Corani, Cassio P. de Campos, Marco Zaffalon
Mauro Scanagatta, Giorgio Corani, Cassio P. de Campos, Marco Zaffalon
[treewidth, number, graph, dag, parent, thus, structural, variable, moral, subgraph, node, twilp, nie, tree, reverse, structure, whose, topological, consistently, report, bic, edge, gobnilp, allow, campos, find, obtained, undirected, actual] [learning, set, bounded, consider, algorithm, problem, yet, conference, cost, every, let, since, function, will, best] [bayesian, large, method, run, initial, size] [order, can, one, first, exact, also] [data, space, given, two] [inverted, state, time, information, search, artificial, exploration] [score, approach, table, network, different, higher, randomly, figure, adopt, propose, using, novel, proposed]
Neurally-Guided Procedural Models: Amortized Inference for Procedural Graphics Programs using Neural Networks
Daniel Ritchie, Anna Thomas, Pat Hanrahan, Noah Goodman
Daniel Ritchie, Anna Thomas, Pat Hanrahan, Noah Goodman
[program, many, partial, number, probabilistic, constraint, possible] [example, function, learning, set] [likelihood, inference, variational, monte, method, importance, faster, sequential, university, efficient, amortized, size, carlo, constrained] [can, also, local, via, computation, require, one] [random, chain, well, distribution, section, mixture, given, reference] [procedural, model, target, smc, unguided, choice, time, guided, current, state, took, match, design, forward, pgm, siggraph, generates, continuous, control, modeling] [image, figure, neural, using, output, training, generate, similarity, network, train, shape, generated, performance, capture, used, use, work, generative, accumulative, matching, different, visual, mask, recent]
Graphons, mergeons, and so on!
Justin Eldridge, Mikhail Belkin, Yusu Wang
Justin Eldridge, Mikhail Belkin, Yusu Wang
[graphon, graph, cluster, clustering, level, merge, tree, probability, node, edge, height, consistency, say, connected, mergeon, equivalence, graphons, claim, theory, structure, possible, collection, identify, represents, piecewise, present, neighborhood, relabeling] [set, let, will, algorithm, every, appendix, consider, now, class, notion, general, setting, may, function, almost, said, equivalent, show] [method, stochastic] [sampled, also, can, recovery, one, suppose, matrix, see, following, zero] [random, consistent, measure, two, estimator, density, distribution, distortion, given, particular, preserving, sense, measurable, statistical, distance, result] [model, hierarchical, must, rather] [pair, single, figure, recent, shown, using, approach]
Geometric Dirichlet Means Algorithm for topic inference
Mikhail Yurochkin, XuanLong Nguyen
Mikhail Yurochkin, XuanLong Nguyen
[number, clustering, obtained, weighted, cluster, probabilistic, probability] [algorithm, polytope, convex, problem, function, will, loss, conference, let, may, learning, since, show, fast] [inference, sampling, latent, simplex, likelihood, method, objective, variational, extreme, processing] [can, extension, min, matrix, via, also, subspace, following] [topic, geometric, recoverkl, dirichlet, gdm, document, tgdm, gibbs, length, distribution, given, geometry, vem, based, increasing, lda, word, pmi, distance, two, data, section, statistical, point, nonparametric, estimation, true, vocabulary] [modeling, varying, information, model] [perplexity, performance, approach, viewed, neural, international, connection]
Robust Spectral Detection of Global Structures in the Data by Learning a Regularization
Pan Zhang
Pan Zhang
[bethe, community, clustering, localized, pairwise, adjacency, block, detection, thus, degree, graph, average, detectability, informative, structure, many, obtained, topology, panel] [problem, learning, algorithm, general, example, will, set] [stochastic, large, method, inference, however, usually] [matrix, spectral, eigenvectors, can, sparse, regularization, hessian, see, localization, rank, singular, eigenvector, completion, first, global, planted, also, eigenvalue, largest, noise, local, vector, participation, one, trimming, noisy, perturbation] [data, random, given, well, way, based, theoretical, associated] [model, information, transition, left, inverse, right] [using, work, different, figure, three, use, generated, several, approach, proposed, network, similarity]
Generative Shape Models: Joint Text Recognition and Segmentation with Very Little Training Data
Xinghua Lou, Ken Kansky, Wolfgang Lehrach, CC Laan, Bhaskara Marthi, D. Phoenix, Dileep George
Xinghua Lou, Ken Kansky, Wolfgang Lehrach, CC Laan, Bhaskara Marthi, D. Phoenix, Dileep George
[detection, graph, edge, strong, many, variable, false] [conference, learning, considered, every] [factor, inference, due, machine] [can, real, high, ieee, one, analysis, first, local, also, vector, accurate, global] [word, two, random, given, true, robust, selection, test, data, document] [model, reading, forward, process, world, system, unlike, hypothetical] [text, parsing, character, shape, training, using, scene, recognition, letter, segmentation, generative, icdar, image, candidate, font, output, discriminative, computer, used, structured, invariance, lateral, svt, approach, trained, vision, single, figure, pattern, feature, photoocr, clean, international, dataset, train, representation, perform, score, language, cnn]
Matrix Completion has No Spurious Local Minimum
Rong Ge, Jason D. Lee, Tengyu Ma
Rong Ge, Jason D. Lee, Tengyu Ma
[claim, probability, many, partial, find] [proof, function, satisfies, will, lemma, theorem, case, let, satisfy, convex, problem, known, show, close, prove, implies, learning, version, setting] [objective, gradient, descent, optimization, showed, step, sampling, nonconvex, machine] [matrix, local, optimality, order, can, first, second, completion, condition, minimum, also, global, necessary, hessian, rank, assumption, spurious, solution, norm, main, incoherent, observed, regularizer, via, low, linear, suppose, initialization, alternating] [point, section, converge, two, tij, random, equation, statistical, positive, result] [observation, must, therefore, information, time, form, starting] [use, work, arxiv, different, preprint, using, similar]
Solving Random Systems of Quadratic Equations via Truncated Generalized Gradient Flow
Gang Wang, Georgios Giannakis
Gang Wang, Georgios Giannakis
[number, many, normalized, thus] [rate, set, constant, function, algorithm, convex, since, will, even, complexity, cost, consider, problem] [gradient, quadratic, nonconvex, method, initial, large, stage] [spectral, tggf, ati, truncated, initialization, generalized, success, real, relative, truncation, phase, can, twf, solving, one, signal, retrieval, error, solution, matrix, vector, global, order, noiseless, recovery, also, computational, via, eigenvalue, sufficiently, ieee, exact, sign, formulation, linear] [gaussian, random, empirical, data, numerical, estimate, given, two, corresponding, sample, squared, stationary] [complex, model, simple, system, form, time, varying, upon, search, worth, superior] [performance, using, novel, flow, figure]
Hierarchical Object Representation for Open-Ended Object Category Learning and Recognition
Seyed Hamidreza Kasaei, Ana Maria Tomé, Luís Seabra Lopes
Seyed Hamidreza Kasaei, Ana Maria Tomé, Luís Seabra Lopes
[number, obtained, assigned] [learning, set, conference, since, known, defined] [latent, batch, incremental, evaluation] [can, view, dictionary, local, first, one, observed, ieee, also] [topic, lda, word, based, distribution, data, dirichlet, well, two, given, gibbs, selected, point, described, bow, provides] [system, new, model, hierarchical, robot, demonstration, must, information, process, simulated, increasingly] [object, category, recognition, used, using, approach, representation, visual, proposed, layer, accuracy, performance, learned, training, table, feature, learn, presented, dataset, shape, scene, candidate, mug, international, available, memory, different, per, classification, semantic, figure, recognize, three, neural, unsupervised]
Select-and-Sample for Spike-and-Slab Sparse Coding
Abdul-Saboor Sheikh, Jörg Lücke
Abdul-Saboor Sheikh, Jörg Lücke
[probabilistic, number, among, variable] [learning, study, set, algorithm, may, setting, general] [latent, posterior, sampling, large, inference, standard, variational, latents, factored, approximation, efficient, apply, scalability, approximate, expectation, scalable, preselection] [sparse, can, observed, truncated, also, first, high, subspace, following, proposition, zero, noise] [data, given, dimensional, gaussian, gibbs, distribution, dimensionality, selection, application, based, sampler, result] [coding, model, complex, previously, within, typically, continuous, encoding, sensory] [using, generative, image, used, neural, use, applied, multiple, hidden, shown, different, scale, reported, approach, work, per, visual, accuracy, including]
Adaptive Newton Method for Empirical Risk Minimization to Statistical Accuracy
Aryan Mokhtari, Hadi Daneshmand, Aurelien Lucchi, Thomas Hofmann, Alejandro Ribeiro
Aryan Mokhtari, Hadi Daneshmand, Aurelien Lucchi, Thomas Hofmann, Alejandro Ribeiro
[variable, number, neighborhood, average] [risk, set, problem, loss, algorithm, erm, consider, convex, optimal, satisfied, theorem, constant, argument, observe, since, implies, learning, upper, conference, bound, minimization] [newton, size, method, ada, stochastic, convergence, gradient, quadratic, factor, step, cvn, increase, iteration, large, line, descent, backtracking, saga, rrn, requires, initial, stepsize, solves, processing, argmin, update, strongly, compute] [condition, order, can, proposition, solution, regularized, assumption, first, one, see, second, phase, also, regularization, following, local] [statistical, sample, empirical, section, result, journal, increasing, associated] [within, information, search] [accuracy, training, use, dataset, neural, performance, single]
Constraints Based Convex Belief Propagation
Yaniv Tenzer, Alex Schwing, Kevin Gimpel, Tamir Hazan
Yaniv Tenzer, Alex Schwing, Kevin Gimpel, Tamir Hazan
[consistency, cbcbp, constraint, cbp, potential, xtp, message, passing, average, xsp, variable, program, lagrange, pairwise, propagation, enforce, include, many, slot, number] [algorithm, set, function, problem, since, define, may, consider, convex, feasible, learning, defined, lemma, show, conference] [dual, standard, inference, update, log, machine, factor, large, faster, respect] [computational, following, can, one, also, first, order, running, via] [given, word, based, two, exp, derivation, associated, random] [model, belief, value, must, time] [source, phrase, using, translation, use, region, accuracy, table, language, used, work, computer, feature, segmentation, neural]
Joint M-Best-Diverse Labelings as a Parametric Submodular Minimization
Alexander Kirillov, Alexander Shekhovtsov, Carsten Rother, Bogdan Savchynskyy
Alexander Kirillov, Alexander Shekhovtsov, Carsten Rother, Bogdan Savchynskyy
[diverse, labelings, runtime, divmbest, quality, labeling, number, pairwise, called, master, unary, solver, graphical, grows, variable, probable, definition] [diversity, submodular, problem, minimization, algorithm, will, binary, case, set, theorem, best, known, let, concave, give, minimizing, function, finding, consider, implies, since, defined] [method, energy, efficient, sequential, faster, optimization, approximate, respect, computing, inference] [can, also, solution, invariant, min, one, exact, interactive] [parametric, measure, joint, two, hamming, distance, permutation, difference, given] [time, dynamic] [use, proposed, used, work, segmentation, single, using, different, multiple, structured, pascal, approach, shown]
Tracking the Best Expert in Non-stationary Stochastic Environments
Chen-Yu Wei, Yi-Te Hong, Chi-Jen Lu
Chen-Yu Wei, Yi-Te Hong, Chi-Jen Lu
[total, number, dependency, according, theory] [regret, algorithm, loss, bound, bandit, arm, let, theorem, learning, expected, will, lower, show, switching, upper, interval, consider, constant, drifting, set, offline, achieve, rate, every, may, appendix, make, problem, online, provide, optimal, achieves, conference, prove, best, even, choose, existence, achievable, case, almost, smaller, idea, chosen, implies] [step, stochastic, variance, supplementary, large] [can, one, vector, order, small, following, denote, note, still, related, good, need] [distribution, given, two, mean, measure, well, way, estimate] [time, information, switch, dynamic, new, action, expert] [using, use, previous, different, like, whole, better]
Visual Question Answering with Question Representation Update (QRU)
Ruiyu Li, Jiaya Jia
Ruiyu Li, Jiaya Jia
[interaction, possible, weighted, resulting] [since, set, may, query, focus, max, learning] [update, updating] [one, also, global, related, following] [two, based, common, given, mean, type, space, retrieved, data] [model, next, answer, information, ability, correct, learns, modeling] [image, question, reasoning, attention, neural, vqa, object, network, layer, visual, representation, answering, preprint, arxiv, language, region, pooling, using, figure, table, use, natural, color, proposed, convolutional, qil, candidate, original, several, memory, dataset, updated, gru, spatial, feature, top, recurrent, relevant, understanding, multiple, generate, single, performance, qru, classification, input, deep, cnn, used, mechanism, content]
Learning a Metric Embedding for Face Recognition using the Multibatch Method
Oren Tadmor, Tal Rosenwein, Shai Shalev-Shwartz, Yonatan Wexler, Amnon Shashua
Oren Tadmor, Tal Rosenwein, Shai Shalev-Shwartz, Yonatan Wexler, Amnon Shashua
[number, expression, according, proportional, partition, variable, sum] [let, every, lemma, case, obtain, consider, fix, proof, observe, set, bound, appear, theorem, show, follows, analyze, since, inequality] [variance, rewrite, sampling, expectation] [row, can, first, denote, matrix, diagonal, vector] [estimate, quantity, random, two, derivation] [value, take, allowing, observing, turn] [used, without, use, entire, like]
Ladder Variational Autoencoders
Casper Kaae Sønderby, Tapani Raiko, Lars Maaløe, Søren Kaae Sønderby, Ole Winther
Casper Kaae Sønderby, Tapani Raiko, Lars Maaløe, Søren Kaae Sønderby, Ole Winther
[weighted, number, purely, recursively, obtained] [learning, show, lower, bound, best, study, tighter, set] [inference, latent, variational, lvae, stochastic, vae, importance, approximate, vaes, likelihood, deterministic, distributed, log, term, five, posterior, additional, parameterization, recently, processing, normalizing] [see, can, one, highly, regularization, also] [gaussian, distribution, test, data, provides, true, two] [model, information, new, observation, prior, process, hierarchical] [generative, using, training, figure, performance, layer, deep, arxiv, trained, preprint, used, neural, representation, different, train, learned, compared, better, omniglot, fully, seen, norb, higher, ladder, consisting, table, unit, qualitatively, improve, mnist, autoencoder]
Dynamic Filter Networks
Xu Jia, Bert De Brabandere, Tinne Tuytelaars, Luc V. Gool
Xu Jia, Bert De Brabandere, Tinne Tuytelaars, Luc V. Gool
[connected, number] [learning, set, idea, loss, show, general, function] [apply, method, applies] [filtering, can, local, also, view, one, note] [transformation, section, given, specific, two, address] [dynamic, model, module, dynamically, another, new, baseline, next, future] [network, filter, input, layer, video, convolutional, image, generated, prediction, spatial, figure, use, deep, stereo, conditioned, learned, feature, single, using, applied, similar, moving, generate, flow, like, work, proposed, residual, highway, learn, different, task, output, predict, propose, previous, convolution, transformer, neural, training, position, sequence, architecture, instead, predicting, lstm, several, driving, used, shown, optical, frame, map, unsupervised, iminds, steerable]
Refined Lower Bounds for Adversarial Bandits
Sébastien Gerchinovitz, Tor Lattimore
Sébastien Gerchinovitz, Tor Lattimore
[probability] [regret, bound, lower, loss, algorithm, bandit, proof, theorem, exists, lemma, let, eqj, round, every, prove, optimal, upper, show, arm, follows, learning, inequality, corollary, learner, defined, randomness, conference, minimax, case, randomised, randomisation, hazan, bounded, chosen, get, since, max, improvement, cumulative, sup, best, smaller, always] [log, expectation, respect, stochastic, quadratic, depend, variance, parameter, supplementary, key, divergence] [can, first, min, also, second, one, high, assumption, denote, existing, order, note] [distribution, two, section, result, difference, random, noting] [internal, range, action, value, therefore, taken, new, must] [sequence, adversarial, variation]
A Non-parametric Learning Method for Confidently Estimating Patient's Clinical State and Dynamics
William Hoiles, Mihaela Van Der Schaar
William Hoiles, Mihaela Van Der Schaar
[number, quality, unique, intervention, contains, level, possible, cancer] [algorithm, learning, hypothesis, bound, set, bernstein, function] [stochastic, bayesian, inference, parameter, increase, method, processing] [can, note, necessary, sufficient] [clinical, given, test, patient, segment, associated, estimate, construct, data, mean, covariance, multivariate, vital, therapeutic, estimated, based, clinician, distribution, estimation, hmm, statistical, estimating, provided, utilize, infinite, gaussian, maximum, bias, provides, admission, rothman, physiological, icu, sample, independent, ppv, section, constructed, dirichlet, empirical] [state, dynamic, model, new, prior, equality, medical, time, information, transition] [using, used, dataset, segmentation, several, accuracy, neural, evaluate, detected, compared]
Low-Rank Regression with Tensor Responses
Guillaume Rabusseau, Hachem Kadri
Guillaume Rabusseau, Hachem Kadri
[structure, constraint] [problem, algorithm, function, least, theorem, let, set, defined, learning, drawn, greedy, minimization, consider, show, loss, finding, convex, returned, equivalent, setting, bounded, will] [approximation, size, supplementary, efficient, method, term, material] [tensor, regression, rank, multilinear, holrr, matrix, can, linear, low, following, analysis, extension, regularization, one, product, tucker, lrr, error, decomposition, kernelized, solution, proposition, good, note, see, orthogonal] [data, kernel, multivariate, generalization, given, space, theoretical, nonlinear, based, computationally, ridge] [model, time, along] [output, approach, proposed, feature, input, using, task, training, top, use, different, image, propose]
The Multiple Quantile Graphical Model
Alnur Ali, J. Zico Kolter, Ryan J. Tibshirani
Alnur Ali, J. Zico Kolter, Ryan J. Tibshirani
[graphical, neighborhood, variable, panel, dependency, structure] [learning, function, consider, problem, set, considered, may, example, assume, studied, algorithm, minimize, smoothing] [machine, optimization, sampling, method, university, likelihood, efficient] [regression, sparse, can, also, one, lasso, matrix, following, via, technical, sparsity, solving, recovery, admm] [conditional, mqgm, quantile, distribution, given, estimation, joint, basis, selection, journal, gaussian, data, additive, section, gibbs, covariance, pseudolikelihood, corresponding, flu, random, multivariate, conditionals, space, estimate, nonparametric, quantiles, underlying, annals, conditionally, independent, univariate, ring, fitted, spacejam, statistical, positive] [model, modeling, continuous, inverse, across] [using, multiple, figure, approach, region, bottom, available]
Stochastic Gradient Geodesic MCMC Methods
Chang Liu, Jun Zhu, Yang Song
Chang Liu, Jun Zhu, Yang Song
[variable, simulate, constraint] [learning, appendix, define, known, conference, set, defined, consider, now] [stochastic, gradient, log, sampling, coordinate, bayesian, inference, monte, posterior, machine, carlo, mcmc, variational, energy, langevin, develop, draw, large, inner, processing, apply, due, scalable] [can, global, also, matrix, first, see, small, one, synthetic, local, denote] [sggmc, geodesic, distribution, embedded, riemann, space, manifold, gmc, empirical, spherical, hamiltonian, true, integrator, gsgnht, topic, estimate, sample, data, mixture, journal, two, test, embedding, expressed] [model, sam, target, information, process, system, von] [using, use, international, used, recipe, dataset, task, figure, neural, like, better]
Contextual semibandits via supervised learning oracles
Akshay Krishnamurthy, Alekh Agarwal, Miro Dudik
Akshay Krishnamurthy, Alekh Agarwal, Miro Dudik
[valid, number, pmin, probability, find, reedy] [algorithm, regret, contextual, semibandit, vcee, learning, composite, bound, set, known, bandit, unknown, oracle, best, setting, agarwal, combinatorial, observe, expected, let, problem, may, klt, achieve, semibandits, achieves, learner, theorem, play, consider, round, context, ucb, least, exploitation, optimal, dependence, make, since, show, appendix] [log, efficient, parameter, variance, unbiased] [can, linear, vector, one, first, regression, also, min, analysis, existing] [empirical, uniform, distribution, space] [reward, policy, simple, action, feedback, exploration, new, information] [use, supervised, using, weight, structured, compare, implemented, better, work, performance, generalizes, feature]
Learning to learn by gradient descent by gradient descent
Marcin Andrychowicz, Misha Denil, Sergio Gómez, Matthew W. Hoffman, David Pfau, Tom Schaul, Nando de Freitas
Marcin Andrychowicz, Misha Denil, Sergio Gómez, Matthew W. Hoffman, David Pfau, Tom Schaul, Nando de Freitas
[base, number, many] [learning, function, problem, will, conference, algorithm, show, setting, appendix] [optimizers, optimization, gradient, update, objective, machine, standard, method, step, additional] [can, one, also, see, small, avoid, plot] [test, generalization, two, well] [behavior, simple, design, allows, optimize, model, optimizing, baseline, artificial] [optimizer, neural, lstm, different, using, trained, network, performance, figure, training, work, style, hidden, learned, optimizee, used, use, content, learn, image, international, train, shown, applying, instead, approach, architecture, final, able, similar, convolutional, resolution, mlp, recurrent, coordinatewise, transfer]
Nested Mini-Batch K-Means
James Newling, François Fleuret
James Newling, François Fleuret
[assignment, thus, cluster, number, triangle, assigned, obtained, rapidly] [algorithm, bound, consider, scheme, may, conference, let, problem, inequality, learning, idea] [premature, nmbatch, energy, end, mbatch, nested, size, iteration, sculley, update, line, standard, machine, large, infmnist, initialisation, yinyang, full, discussed, extreme, accelerate, approximate] [one, first, can, exact, sparse, relative, note, second, minimum, running] [centroid, distance, data, based, redundancy, sample, random, lloyd, two, elkan, mean, difference, test] [new, time, balancing, within] [using, used, use, bounding, figure, training, approach, shown, validation, international, performed, dataset, presented, already]
Quantized Random Projections and Non-Linear Estimation of Cosine Similarity
Ping Li, Michael Mitzenmacher, Martin Slawski
Ping Li, Michael Mitzenmacher, Martin Slawski
[bit, number, expression, probability] [set, optimal, case, scheme, let, function, may, oracle, consider, lemma, theorem, even, depends, problem, learning, chosen, general] [variance, likelihood, full, inner, requires] [can, linear, one, fraction, high, relative, also, real, small, matrix] [data, mle, random, quantized, quantization, given, estimator, hamming, neighbor, nearest, retrieved, based, test, fisher, two, maximum, specific, result, estimation, statistical, empirical, farm, corresponding, estimate, quantizer, distance, denotes] [information, choice, form, search, required, inverse, cell] [approach, similarity, figure, unit, using, accuracy, performance, better, cosine]
Generating Images with Perceptual Similarity Metrics based on Deep Networks
Alexey Dosovitskiy, Thomas Brox
Alexey Dosovitskiy, Thomas Brox
[connected] [loss, learning, function, class, show, best, set, even] [method, term, latent, supplementary, variational] [can, error, interesting, good, see, local, real, first] [space, two, euclidean, based, distribution, important, random, squared, metric, given] [model] [image, feature, alexnet, training, adversarial, network, generator, similarity, generative, shown, trained, convolutional, used, natural, discriminator, perceptual, deep, different, neural, conv, generate, using, architecture, figure, approach, proposed, use, inversion, work, similar, comparator, table, unsupervised, fully, eat, layer, generation, generated, uconv, reconstruction, train, encoder, generating, extracted, videonet, learn, dosovitskiy, inverting]
Automated scalable segmentation of neurons from multispectral images
Uygar Sümbül, Douglas Roossien, Dawen Cai, Fei Chen, Nicholas Barry, John P. Cunningham, Edward Boyden, Liam Paninski
Uygar Sümbül, Douglas Roossien, Dawen Cai, Fei Chen, Nicholas Barry, John P. Cunningham, Edward Boyden, Liam Paninski
[number, graph, basic, many, connected, obtained, expression, normalized, level] [problem, set, index, binary, show, algorithm] [size, due, method, university, processing, merging, large] [can, noise, projection, real, fluorescent, ieee, cij, one, analysis, much, imaging, denoised] [maximum, two, data, distance, based, density, slice, automated] [intensity, voxels, individual, simulated, neuronal, within, neuron, nature, variability, simple, anatomy, simulation, light, adjusted, occupied, along, form] [color, image, segmentation, supervoxels, brainbow, single, stack, different, voxel, neural, microscopy, supervoxel, approach, spatial, use, accuracy, figure, patch, nervous, background, using, used, similar, rand, foreground, topographic, various, dataset]
The Multiscale Laplacian Graph Kernel
Risi Kondor, Horace Pan
Risi Kondor, Horace Pan
[graph, vertex, flg, kflg, definition, mlg, multiscale, base, level, structure, number, induced, subgraphs, karsten, subgraph, degree, topological, collection, purely] [let, defined, just, learning, set, define, conference, may, will, sensitive, now, assume, function, property] [compute, computing, machine, approximation, computed, size, requires, processing] [can, matrix, local, low, proposition, gram, also, first, small, one, eigenvalue, following, overall, orthonormal, similarly, rank, paper, spectral, eigenvectors, vector, invariant] [kernel, two, space, laplacian, corresponding, joint, given, basis, section, positive, random, risi, based, permutation, data] [information, new, literature] [feature, different, used, capture, international, multiple, single, neural]
Fast ε-free Inference of Simulation Models with Bayesian Conditional Density Estimation
George Papamakarios, Iain Murray
George Papamakarios, Iain Murray
[number, probability, enough] [learning, algorithm, will, set, conference, rejection, make, lower] [mdn, abc, proposal, posterior, bayesian, approximate, inference, log, monte, parameter, approximation, carlo, efficient, learnt, likelihood, supplementary, mdns, conventional, processing, stochastic] [can, one, observed, need, exact, vector, order] [density, true, gaussian, sample, data, conditional, parametric, mixture, two, estimation, section, estimate, refer, procedure, given, population] [prior, model, time, left, process, typically] [used, neural, training, use, using, figure, effective, trained, work, learn, train, single, hidden, generative, approach, three, network, recognition, generated, propose, tanh, layer, calculated, per]
Reshaped Wirtinger Flow for Solving Quadratic System of Equations
Huishuai Zhang, Yingbin Liang
Huishuai Zhang, Yingbin Liang
[number, chen] [loss, algorithm, function, complexity, problem, cost, case, set, proof, guarantee, expected, achieves, show, rate] [gradient, convergence, nonsmooth, nonconvex, step, faster, quadratic, log, method, converges, update, processing, although] [signal, phase, can, initialization, ati, via, retrieval, truncation, candes, computational, paper, also, global, wirtinger, real, solving, relative, noise, error, cdp, sign, comparison, magnitude, altminphase, recovery, much, spectral, reshaped, truncated, following, small] [sample, two, section, random, gaussian, based, given, geometric] [time, complex, poisson, direction, next, design, information] [figure, performance, neural, proposed, shown, flow, compare, applied]
Learning Infinite RBMs with Frank-Wolfe
Wei Ping, Qiang Liu, Alexander T. Ihler
Wei Ping, Qiang Liu, Alexander T. Ihler
[number, probability, average, sum, find] [algorithm, learning, convex, function, set, greedy, relaxation, general, best, since, max, stopping] [gradient, standard, likelihood, optimization, log, rbms, update, size, parameter, term, solve, inference, softplus, step, efficient, method, machine, boosting, latent, iteration, draw, initialize, marginal] [can, one, restricted, also, initialization, sparse, linear, following, proposition, error] [rbm, test, boltzmann, infinite, random, gibbs, distribution, provides, functional, exp, two, sampler, selected, bias, empirical, mixture] [model, directly, new] [hidden, using, training, fractional, neural, validation, use, unit, propose, figure, learned, used, mnist, deep, layer, better, part, weight]
Joint Line Segmentation and Transcription for End-to-End Handwritten Paragraph Recognition
Theodore Bluche
Theodore Bluche
[weighted, vertical, complete] [may, conference, offline, workshop, set] [line, processing, full, standard, automatic, machine, step, method] [analysis, error, one, transcription, can, also, explicit] [document, section, word] [model, system, simple, implicit, state, followed, module, baseline, time, information, encoded] [recognition, neural, segmentation, text, handwriting, attention, collapse, handwritten, network, image, proposed, international, applied, layer, feature, recurrent, language, mdlstm, dpi, trained, preprint, arxiv, blstm, character, sequence, paragraph, presented, table, iam, decoder, architecture, softmax, whole, figure, validation, training, input, different, without, convolutional, transcribe]
Feature-distributed sparse regression: a screen-and-clean approach
Jiyan Yang, Michael W. Mahoney, Michael Saunders, Yuekai Sun
Jiyan Yang, Michael W. Mahoney, Michael Saunders, Yuekai Sun
[number, possible, total, thus, enough] [algorithm, theorem, problem, cost, least, show, round, learning, absolute, set] [log, size, screening, distributed, machine, stage, convergence, reduce, deco, constrained, datasets, step, method, michael, ensures, iterates, fit, central] [sparse, error, sketch, lasso, regression, sent, matrix, sketching, linear, high, creena, sure, cleaning, martin, can, sparsity, concerned, order, contraction, main, condition, lean, cone, pilanci] [statistical, data, two, exp, estimator, independent, random, selected, selection, refer, theoretical] [communication, model] [approach, prediction, amount, figure, preprint, arxiv, similar, single, performance, network, using]
CMA-ES with Optimal Covariance Update and Storage Complexity
Oswin Krause, Dídac Rodríguez Arbonès, Christian Igel
Oswin Krause, Dídac Rodríguez Arbonès, Christian Igel
[runtime, root, average, genetic] [algorithm, function, complexity, conference, learning, problem, every, krause, consider, annual] [update, cholesky, factor, evolutionary, triangular, optimization, objective, machine, rotation, standard, large, hansen, coordinate, rosenbrock, igel, discus, step, updating, term, denmark, ajj, cigar, diffpowers, efficient] [matrix, can, square, computation, one, decomposition, linear, also, sampled, error, note, global] [covariance, given, theoretical, sample, distribution, sphere, space, refer, normal] [time, evolution, benchmark, search, required, change, long, new, policy, value] [using, figure, adaptation, compared, approach, memory, original, propose, instead]
Unsupervised Learning of 3D Structure from Images
Danilo Jimenez Rezende, S. M. Ali Eslami, Shakir Mohamed, Peter Battaglia, Max Jaderberg, Nicolas Heess
Danilo Jimenez Rezende, S. M. Ali Eslami, Shakir Mohamed, Peter Battaglia, Max Jaderberg, Nicolas Heess
[number, probabilistic, structure] [context, learning, show, appendix, consider, class, function, best, bound, set, conference] [inference, latent, operator, variational] [can, projection, observed, first, via, also, one, missing, recover] [data, conditional, given, provided] [model, directly, form, abstract, complex, inferred, infer] [representation, figure, generative, volume, training, image, object, volumetric, mesh, using, capture, trained, arxiv, different, network, dataset, shapenet, neural, opengl, scene, deep, generation, camera, vision, seen, use, train, performance, computer, renderer, shown, multiple, hidden, jimenez, conditioned, rendered, unsupervised, used, work, learn, generate]
Budgeted stream-based active learning via adaptive submodular maximization
Kaito Fujii, Hisashi Kashima
Kaito Fujii, Hisashi Kashima
[many, number, obtained, partial] [adaptive, submodular, active, secretary, learning, algorithm, submodularity, set, setting, problem, expected, utility, maximization, selecting, conference, class, function, budget, selects, let, optimal, general, constant, give, rate, adaptivestream, adaptively, instance, observe, monotone, maximizing, budgeted, consider, golovin, assume, bound, return] [select, marginal, stochastic, sampling, objective, machine, step] [item, can, one, error, also, existing, first, real, analysis, denote] [theoretical, based, given, random, two, data] [policy, uncertainty, gain, realization, information, pool, state, world, prior, new, another] [stream, several, proposed, entire, international, propose, used, performance, labeled, figure]
Exact Recovery of Hard Thresholding Pursuit
Xiaotong Yuan, Ping Li, Tong Zhang
Xiaotong Yuan, Ping Li, Tong Zhang
[strong, number, level, larger, implied] [theorem, will, minimizer, arbitrary, learning, set, loss, conference, algorithm, problem, bound, function, bounded, appendix, guarantee, greedy, let] [logistic, parameter, log, large, free] [recovery, htp, support, condition, can, regression, sparse, analysis, exact, restricted, success, sparsity, relaxed, linear, also, recover, following, vector, error, hard, exactly, proposition, global, thresholding, certain, jain, high, arg, yuan, sparsistency, nonzero, min, recovering, sufficient, significantly, sensing] [estimation, result, data, provided, statistical, two, given, numerical, journal, well, finite] [model, target, information, pursuit, supporting] [able, figure, performance, output]
Robustness of classifiers: from adversarial to random noise
Alhussein Fawzi, Seyed-Mohsen Moosavi-Dezfooli, Pascal Frossard
Alhussein Fawzi, Seyed-Mohsen Moosavi-Dezfooli, Pascal Frossard
[relation] [case, show, chosen, let, defined, conference, even, study, define, learning, general, close, label, class, binary, arbitrary, theorem, provide, now] [large, machine, factor, optimization] [noise, subspace, dimension, can, small, high, perturbation, assumption, analysis, noted, following, sufficiently, note, global, linear, fraction, support, condition, also, first, norm, minimal] [robustness, random, curvature, boundary, result, robust, regime, theoretical, two, datapoint, data, quantity, affine, precisely, dimensional, vulnerable, radius, empirical, nonlinear, behaves, estimated, given, provided] [decision, therefore, ball] [adversarial, classification, different, neural, deep, classifier, image, shown, figure, international, classified, arxiv, preprint]
Designing smoothing functions for improved worst-case competitive ratio in online optimization
Reza Eghbali, Maryam Fazel
Reza Eghbali, Maryam Fazel
[variable, piecewise] [competitive, ratio, algorithm, function, online, optimal, smoothing, problem, convex, bound, adwords, concave, since, nesterov, set, let, consider, umax, lemma, show, case, satisfies, lower, pseq, theorem, dseq, learning, derive, worst, monotone, satisfy, assume, regret, finding, duality, provide, defined, differentiable, equivalent, att, studied, covering, dsim, cost, orthant] [dual, optimization, separable, objective, primal, update, sequential, respect, conjugate, method] [can, assumption, note, linear, solution, cone, simultaneous, regularization, condition, solving, one, also, sufficient, optimality, order] [given, two, section, positive, provides, based, derived, point] [maximize, continuous, simplify, subject] [figure, preprint, arxiv, achieved, using]
Deep Learning for Predicting Human Strategic Behavior
Jason S. Hartford, James R. Wright, Kevin Leyton-Brown
Jason S. Hartford, James R. Wright, Kevin Leyton-Brown
[weighted, sum, number, theory, level] [utility, element, learning, depends, set, function, expected, every, max, best] [parameter, respect] [row, can, column, matrix, iterative, assumption, also, first] [distribution, corresponding, two, normal, section, given, test] [response, action, model, game, player, strategic, behavioral, scalar, human, form, behavior, respond, cognitive, allows, internal, experimental, optimize, payoff, modeling, sharpness, blue, quantal, arl, arj] [hidden, output, layer, input, using, architecture, network, pooling, performance, deep, feature, figure, final, training, unit, used, use, reasoning, representation, predicting, red, softmax, applying, neural, invariance, able, previous, original, learn]
Clustering with Bregman Divergences: an Asymptotic Analysis
Chaoyue Liu, Mikhail Belkin
Chaoyue Liu, Mikhail Belkin
[clustering, number, coefficient, probability] [set, function, loss, algorithm, optimal, will, case, convex, lemma, general, defined, setting, define, example, theorem, constant, show] [divergence, size, respect, converges, method, convergence, apply, large] [error, analysis, support, order, min, also, first, following, suppose, matrix, can, hessian, intuition, related, paper, power, hard, note, one] [bregman, quantization, distribution, euclidean, asymptotic, density, measure, squared, data, corresponding, point, theoretical, distance, limit, based, discrete, mahalanobis, finite, cube, limiting, well, centroid, graf, reduces] [continuous, cell, form, information, experimental] [figure, used, volume, shown, dataset]
Blind Attacks on Machine Learners
Alex Beatson, Zhaoran Wang, Han Liu
Alex Beatson, Zhaoran Wang, Han Liu
[thus, knowledge, either] [minimax, attacker, learner, risk, lower, bound, setting, blind, consider, set, informed, proof, problem, cam, learning, bounded, theorem, privacy, may, fano, corollary, upper, chooses, loss, drawn, service, arbitrary, assume, rate, optimal, provide] [dkl, parameter, standard, log, convergence, machine, divergence, size, university] [attack, malicious, injection, analysis, regression, linear, can, denote, observed] [data, distribution, uniform, section, estimation, true, family, density, estimator, estimating, given, provides, sample, differential, mean, robust, mutual, two, fixed, random, considering] [information, security, form, interest, value, framework, choice, ability] [training, work, effective, presented, network, used, dataset, aware]
Spatiotemporal Residual Networks for Video Action Recognition
Christoph Feichtenhofer, Axel Pinz, Richard Wildes
Christoph Feichtenhofer, Axel Pinz, Richard Wildes
[average] [learning, loss, best, show] [size, batch, large, method, normalization] [can, also, first] [two, dimensionality] [temporal, spatiotemporal, action, model, time, across, human] [residual, convolutional, motion, appearance, layer, spatial, stream, recognition, network, training, relu, video, approach, architecture, input, use, image, table, performance, convnet, using, deep, convnets, field, resnet, idt, stride, flow, several, feature, pooling, used, work, fully, receptive, visual, resnets, three, trained, learn, optical, output, accuracy, stacking, classification, shown, connection, neural, mapping, initialized]
Minimizing Quadratic Functions in Constant Time
Kohei Hayashi, Yuichi Yoshida
Kohei Hayashi, Yuichi Yoshida
[number, graph, probability, cut, independently, partition, whose, obtained, aij] [let, function, problem, lemma, every, set, define, algorithm, constant, optimal, uniformly, exists, theorem, least, bijection, pearson, inf, property, obtain, show, dikernels, minimization, assume, defined, prove, satisfies, want, analyze] [method, approximation, quadratic, optimization, machine, log, sampling, siam, divergence, requires] [can, dikernel, wpt, matrix, following, sampled, note, min, solution, vector, equipartition, error, real, one, norm, need, linear, small, via] [journal, kernel, measure, independent, denotes, distribution, random, measurable, two, data, numerical] [time, value] [sequence, dense, proposed, different, using, used]
Tensor Switching Networks
Chuan-Yung Tsai, Andrew M. Saxe, Andrew M. Saxe, David Cox
Chuan-Yung Tsai, Andrew M. Saxe, Andrew M. Saxe, David Cox
[thus] [learning, equivalent, switching, even, may, function, since, scheme, consider, problem, rate, provide, define] [standard, gradient, expressive] [linear, can, analysis, also, tensor, error, still, vector, one, contraction, first, hence, much, via, following] [kernel, random, asymptotic, ridge, expansion, denotes, based, data, two, given] [backpropagation, inverted, time, learns, decision, activity, scalar, experimental, across] [deep, network, hidden, representation, activation, input, relu, using, unit, expanded, layer, cnn, readout, neural, nonlinearities, depth, convolutional, figure, entire, mlp, vec, different, lrc, deeper, vanishing, rnl, training, back, used, learn, activate, better, mnist, conveys]
Barzilai-Borwein Step Size for Stochastic Gradient Descent
Conghui Tan, Shiqian Ma, Yu-Hong Dai, Yuqiu Qian
Conghui Tan, Shiqian Ma, Yu-Hong Dai, Yuqiu Qian
[kong] [algorithm, function, convex, smoothing, learning, assume, may, prove, sensitive, show, idea, chosen, problem, unknown] [step, size, gradient, sgd, stochastic, method, svrg, convergence, technique, variance, line, diminishing, initial, compute, descent, strongly, computed, end, processing, large, dashed, full, usually, optimization, barzilai, ima, parameter, machine, dai] [can, also, following, linear, one, solving, see, first, need, comparison, running, min] [numerical, two, given, journal, fixed, section, correspond, denotes] [information, choice, search, solid, research, option] [use, different, propose, used, figure, neural, better, performance, using]
Variance Reduction in Stochastic Gradient Langevin Dynamics
Kumar Avinava Dubey, Sashank J. Reddi, Sinead A. Williamson, Barnabas Poczos, Alexander J. Smola, Eric P. Xing
Kumar Avinava Dubey, Sashank J. Reddi, Sinead A. Williamson, Barnabas Poczos, Alexander J. Smola, Eric P. Xing
[number, average, subset] [algorithm, set, cost, theorem, let, learning, function, show, since, constant, assume] [stochastic, aga, gradient, variance, log, langevin, convergence, gld, posterior, vrg, pass, monte, sgld, bayesian, step, mse, reduction, datasets, due, carlo, approximate, approximation, size, optimization, reduce, large, update, method, machine, term, standard, faster, however, smooth, sampling] [can, computational, regression, also, analysis, high, following, note, noisy, noise, much, component] [data, test, true, given, equation, independent, theoretical, mixture, two, distribution, empirical, section, based] [] [using, performance, use, figure, used, memory, dataset, approach, reducing, proposed, shown, similar]
Unified Methods for Exploiting Piecewise Linear Structure in Convex Optimization
Tyler B. Johnson, Carlos Guestrin
Tyler B. Johnson, Carlos Guestrin
[piecewise, many, structure, include, resulting, relation, number] [algorithm, set, theorem, problem, since, convex, define, consider, function, lemma, lower, learning, appendix, minimizer, optimal, general, choose, minimizing, conference, svm, may, bound, let, case, toward, choosing, loss] [screening, working, optimization, dca, objective, suboptimality, progress, exploiting, machine, xik, dual, constrained, minimizes, subproblem, convergence, applies, scalability, litz, select, ggl] [can, solving, linear, also, group, lasso, one, gap, sparse, principled, solution, suppose, existing, via, support] [test, result, important, point, simpler, theoretical] [time, prior, choice, upon, simple, safe] [using, international, training, feature, used]
Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering
Michaël Defferrard, Xavier Bresson, Pierre Vandergheynst
Michaël Defferrard, Xavier Bresson, Pierre Vandergheynst
[graph, localized, regular, coarsening, clustering, chebyshev, number, thus, vertex, fake, node, introduced, quality, level, normalized] [learning, complexity, may, set, conference, fast, defined, algorithm, loss] [size, efficient, processing, however, requires] [spectral, can, signal, local, computational, formulation, one, linear, matrix, vector, order, filtering, analysis, support, spline, ieee, via] [data, two, fourier, classical, basis, polynomial, mathematical, word, euclidean] [model, operation, time, future] [pooling, proposed, convolutional, cnns, table, neural, cnn, spatial, filter, figure, architecture, classification, use, text, input, mnist, accuracy, learn, learned, applied, convolution, deep, extract, domain, performance, meaningful, training, using, layer]
Mistake Bounds for Binary Matrix Completion
Mark Herbster, Stephen Pasteris, Massimiliano Pontil
Mark Herbster, Stephen Pasteris, Massimiliano Pontil
[number, graph, made, thus, science] [bound, algorithm, learning, margin, online, binary, theorem, complexity, upper, conference, every, problem, set, mistake, max, setting, bounded, may, will, case, regret, lemma, best, corollary, define, optimal, london, uij, let, annual, example, observe, unknown, label, appendix, exists, learner, loss, inf, assume, smaller, kjj, make] [machine, log, although] [matrix, row, norm, via, see, following, analysis, denote, completion, entry, hence, vector, assumption] [kernel, perceptron, given, quantity, underlying, section, consistent, finite] [trace, trial, nature] [predicting, using, sequence, prediction, use, task, international, different, per]
Optimizing affinity-based binary hashing using auxiliary coordinates
Ramin Raziperchikolaei, Miguel A. Carreira-Perpinan
Ramin Raziperchikolaei, Miguel A. Carreira-Perpinan
[cut, number, many, resulting, bit] [hash, binary, function, loss, hashing, learning, algorithm, mac, problem, ynm, ksh, precision, maccut, macquad, since, quad, will, finding, graphcut, set, fast, minimizer, esplh] [optimization, objective, free, step, term, quadratic, method, optimizes, apply, large, although, involves, auxiliary] [can, linear, one, first, vector, still, much, alternating, also, following] [given, nonlinear, kernel, two, embedding, data, space, point, section] [optimizing, continuous, optimize, framework, form, simply, value, directly] [using, use, training, original, image, better, learn, supervised, approach, work, proposed, different, cifar]
Operator Variational Inference
Rajesh Ranganath, Dustin Tran, Jaan Altosaar, David Blei
Rajesh Ranganath, Dustin Tran, Jaan Altosaar, David Blei
[program, many, find] [function, class, algorithm, consider, satisfies, learning, will, conference, sup, set, known] [variational, operator, objective, inference, posterior, log, divergence, respect, approximating, latent, stochastic, approximate, expectation, tractable, gradient, black, optimization, machine, approximation, develop, bayesian, requires, unbiased, method, calculate, factor, opvi, tractability, logistic, university] [can, require, zero, analysis, second, linear, one, note] [data, family, equation, test, distribution, two, density, mixture, distance, gaussian, given, statistical, positive, construct] [model, box, new, optimizing, design, continuous, typically, value, rich] [use, using, score, generative, neural, international, arxiv, different, preprint, truth]
Dueling Bandits: Beyond Condorcet Winners to General Tournament Solutions
Siddartha Y. Ramamohan, Arun Rajkumar, Shivani Agarwal, Shivani Agarwal
Siddartha Y. Ramamohan, Arun Rajkumar, Shivani Agarwal, Shivani Agarwal
[pairwise, anytime, definition, nij, cycle, probability] [regret, tournament, set, dueling, elect, roc, algorithm, pij, winner, arm, bandit, copeland, condorcet, uij, let, uncovered, bound, satisfies, appendix, will, general, ucbs, conference, winning, upper, maximal, define, defined, pjk, ucb, else, theorem, uji, selects, return, max, borda, cumulative, always, confidence, case] [select, stochastic, end] [preference, can, condition, matrix, also, solution, following, relative, one, see] [selection, procedure, based] [safe, target, feedback, individual, trial, design, exploration] [top, figure, used, work, international, natural, three, without, pair]
Learning brain regions via large-scale online structured sparse dictionary learning
Elvis DOHMATOB, Arthur Mensch, Gael Varoquaux, Bertrand Thirion
Elvis DOHMATOB, Arthur Mensch, Gael Varoquaux, Bertrand Thirion
[number, constraint, thus, represent, often] [online, problem, learning, algorithm, defined, set] [variance, parameter, update, method, large] [dictionary, sodl, sparse, explained, can, sparsity, regularization, canica, via, one, penalty, analysis, impose, component, bcd, small, linear, onto, see, vanilla, also, imaging, hcp, good, matrix, decomposition, neuroimage, tcanica, overall, varoquaux] [data, functional, laplacian, mean, statistical, test, sample, estimated, corresponding, ridge, major] [model, brain, across, fmri, variability, current, raw, behavioral, cognitive] [proposed, structured, spatial, like, different, compared, use, performance, propose, training, using, work]
Noise-Tolerant Life-Long Matrix Completion via Adaptive Sampling
Maria-Florina F. Balcan, Hongyang Zhang
Maria-Florina F. Balcan, Hongyang Zhang
[probability, represent, many, number] [algorithm, complexity, online, bound, theorem, bounded, lower, least, problem, learning, assume, setting, let, study, best, passive, drawn, uniformly, conference, constant, adaptive, upper, guarantee] [sampling, deterministic, size, machine, log, supplementary] [matrix, noise, subspace, column, completion, can, error, sparse, recovery, noisy, dictionary, one, exact, rank, small, linear, arriving, norm, ieee, exactly, vector, noiseless, recover, outlier, assumption, global, via, denote, dimension, corrupted] [sample, underlying, random, space, two, result, mixture, data, estimated, given, robust, estimate] [information, prior, another, goal] [figure, hidden, clean, add, layer, without]
The Product Cut
Thomas Laurent, James von Brecht, Xavier Bresson, arthur szlam
Thomas Laurent, James von Brecht, Xavier Bresson, arthur szlam
[cut, graph, normalized, partition, pcut, ncut, cluster, vertex, number, purity, thus, partitioning, connected, quality, nmfr, obtained, weighted] [algorithm, set, theorem, will, relaxation, convex, problem, bound, lower, rate, let, general, may, define, version, provide] [objective, energy, optimization, large, supplementary, variant, iterates] [product, linear, denote, matrix, can, following, perturbation, convexity, relies, exact, stability, solution, conductance, algorithmic, small, quite] [data, random, two, provides, given, denotes, classical, theoretical, point, mathematical] [model, continuous, maximize, therefore, simple, time, experimental, optimize] [use, using, table, figure, sequence]
Learning Supervised PageRank with Gradient-Based and Gradient-Free Optimization Methods
Lev Bogolubsky, Pavel Dvurechensky, Alexander Gasnikov, Gleb Gusev, Yurii Nesterov, Andrei M. Raigorodskii, Aleksey Tikhonov, Maksim Zhukovskii
Lev Bogolubsky, Pavel Dvurechensky, Alexander Gasnikov, Gleb Gusev, Yurii Nesterov, Andrei M. Raigorodskii, Aleksey Tikhonov, Maksim Zhukovskii
[number, level, graph, obtained, page, walk, probability, gfn, introduced] [set, algorithm, oracle, problem, function, lemma, choose, loss, general, learning, assume, convex, consider, theorem, complexity, gbn, inequality, obtain, upper, let, proof, lower, considered] [method, optimization, inexact, calculate, gradient, convergence, apply, supplementary, gbp, step, solve, parameter, restart] [vector, ranking, pagerank, following, can, also, first, power, surfer, iterative, solving] [random, stationary, test, point, distribution, calculation] [value, web, framework, required, unlike, allows, markov] [accuracy, supervised, using, used, use, different, figure]
Theoretical Comparisons of Positive-Unlabeled Learning against Positive-Negative Learning
Gang Niu, Marthinus Christoffel du Plessis, Tomoya Sakai, Yao Ma, Masashi Sugiyama
Gang Niu, Marthinus Christoffel du Plessis, Tomoya Sakai, Yao Ma, Masashi Sugiyama
[probability, either, many] [risk, learning, misclassification, rate, theorem, bound, loss, rpu, will, let, tighter, rnu, function, assume, drawn, minimizers, rademacher, problem, surrogate, set, case, class, upper, lemma, least, may, smaller, since, defined, label, proof, remaining, even, satisfies] [size, machine, sampling, marginal, unbiased] [error, can, order, one, also, denote, analysis] [data, estimation, based, theoretical, given, positive, random, estimator, fixed, empirical, kernel, two, statistical, limit, density, journal, ordinary] [experimental, benchmark, artificial] [figure, three, gpu, unlabeled, table, using, compare, supervised, rpn, training, improve, classifier]
A Simple Practical Accelerated Method for Finite Sums
Aaron Defazio
Aaron Defazio
[theory, relation, strong, number] [function, algorithm, rate, convex, theorem, chosen, known, set, general, loss, define, now, notation] [step, proximal, gik, accelerated, method, gradient, zjk, epoch, operator, suboptimality, sdca, saga, inner, gjk, stochastic, size, convergence, processing, incremental, catalyst, dual, recently, descent, log, primal, however, datasets, tong, shai, term, iterate, supplementary, efficiently, curran, lyapunov] [can, also, product, linear, note, main, see, following, condition, convexity, much] [finite, two, point, given] [fig, information, form, whereas] [used, using, neural, use, instead, applied, able, approach]
Feature selection in functional data classification with recursive maxima hunting
José L. Torrecilla, Alberto Suárez
José L. Torrecilla, Alberto Suárez
[variable, recursive, number, base, rule, average, complete, whose, introduced] [optimal, function, set, class, relevance, close, problem, max, depends, considered] [standard, reduction, method, end, size] [error, can, pca, one, local, analysis, also, second, first, linear, regression, vector, plot] [functional, rmh, selection, data, selected, hunting, distance, pls, brownian, bayes, dimensionality, berrendero, maximum, corresponding, two, tmax, redundancy, important, space, correction, journal, empirical, tinf, test, statistical, based, section, procedure] [correlation, information, process, goal, value] [classification, feature, used, different, figure, relevant, using, accuracy, performance, training, approach, similar]
Kernel Observers: Systems-Theoretic Modeling and Inference of Spatiotemporally Evolving Processes
Hassan A. Kingravi, Harshal R. Maske, Girish Chowdhary
Hassan A. Kingravi, Harshal R. Maske, Girish Chowdhary
[number] [let, set, problem, lower, learning, function, index, bound, show, general, property, provide, may, will] [sampling, inference, latent, machine, supplementary, approximate, large, auto] [can, matrix, error, measurement, proposition, linear, note, monitoring, sensing, condition, sufficient] [kernel, nonstationary, gaussian, given, observability, random, covariance, functional, space, data, differential, two, spatiotemporally, section, wind, rkhs] [model, spatiotemporal, evolution, observer, state, system, process, modeling, autonomous, temperature, rms, shaded, observation, cyclic, time, required, design, temporal, pclsk, sensor, form, heuristic, modeled] [approach, using, figure, map, input, spatial, original, training, work, feature, presented, domain, use, prediction]
Linear Contextual Bandits with Knapsacks
Shipra Agrawal, Nikhil Devanur
Shipra Agrawal, Nikhil Devanur
[probability, total, present, definition, relation] [algorithm, opt, contextual, online, bound, regret, arm, problem, every, confidence, bandit, learning, context, lemma, budget, round, let, case, optimal, set, special, optimistic, ellipsoid, corollary, get, define, unknown, outcome, lincbwk, played, proof, defined, setting, consider, theorem, general, since, may, lower, conference, bounded] [consumption, stochastic, parameter, log, optimization, depend, update] [linear, can, vector, first, assumption, matrix, column, following, arg, also, denote, one, high] [given, estimate, section, distribution] [time, reward, policy, value, resource, required, maximize, dynamic, along, choice] [using, use, static, used, weight, work, similar]
Achieving budget-optimality with adaptive schemes in crowdsourcing
Ashish Khetan, Sewoong Oh
Ashish Khetan, Sewoong Oh
[assignment, worker, probability, average, number, fundamental, crowdsourcing, assigned, total, level, threshold, quality, assign, requester, made, majority, introduced, enough] [algorithm, adaptive, budget, difficulty, rate, theorem, bound, set, scheme, lower, achieve, round, constant, minimax, get, assume, make, since, best, conference, provide, choose, let, case, achieves, bounded] [inference, end, supplementary, log, processing, large, standard] [error, can, one, arriving, fraction, denote, following, sufficient, first, also, note] [distribution, given, random, limit, mean, based, true, section] [model, gain, choice, difficult, information, next, scaling, take] [task, using, per, neural, approach, performance, multiple, figure, classify, classification]
Search Improves Label for Active Learning
Alina Beygelzimer, Daniel J. Hsu, John Langford, Chicheng Zhang
Alina Beygelzimer, Daniel J. Hsu, John Langford, Chicheng Zhang
[disagreement, probability, number, many, rune, loop] [earch, abel, learning, hypothesis, active, version, arch, counterexample, algorithm, set, label, complexity, example, let, class, query, theorem, least, may, rate, cal, ersion, cost, assume, lemma, oracle, agnostic, setting, consider, proof, case, realizable, appendix, return, earchh, learner, hanneke, call, balcan, ccq, just, now] [log, end, nested, iteration, step, processing] [can, error, union, practice, contain] [space, section, positive, consistent, provides, powerful] [search, target, information, current, substantially, new] [labeled, region, unlabeled, using, use, natural, neural, daniel, sequence]
A Credit Assignment Compiler for Joint Prediction
Kai-Wei Chang, He He, Stephane Ross, Hal Daume III, John Langford
Kai-Wei Chang, He He, Stephane Ross, Hal Daume III, John Langford
[assignment, probabilistic, programming, dependency, program, number, many, cache, variable] [learning, loss, function, algorithm, define, best, online, may, show, make, set, svm, complexity, essentially, defined] [run, machine, factor, hyperparameters, end, optimization] [can, one, also, much, tagging] [joint, space, two, perceptron, test, reference, based, underlying, statistical, point] [search, policy, credit, state, tdolr, compiler, time, rollin, action, rollout, system, parser, current, ctb, ner, speed, markov, effect, code, making, complex, simple] [prediction, training, structured, sequence, performance, figure, use, different, output, input, approach, predict, library, neural, crf, using, language, classifier]
A Bio-inspired Redundant Sensing Architecture
Anh Tuan Nguyen, Jian Xu, Zhi Yang
Anh Tuan Nguyen, Jian Xu, Zhi Yang
[number, level, many, total, larger, degree, thus, theory, calibration, probabilistic, precise] [set, precision, known, even, may, ratio, optimal, defined, binary] [increase, processing, secondary, large, due, implementation] [error, can, component, sensing, high, power, computational] [mismatch, distribution, data, quantization, shannon, redundant, reference, redundancy, conversion, adc, two, maximum, binocular, limit, journal, digital, quantizer, differential, integrated, assembly, result] [information, human, allows, design, biological, intrinsic, artificial, heuristic, primary, simulation, eye, geometrical, nonlinearity] [figure, resolution, visual, proposed, without, unit, vision, effective, approach, different, physical, architecture, similar, amount, million, generate, field, neural, unsupervised, using]
Spatio-Temporal Hilbert Maps for Continuous Occupancy Representation in Dynamic Environments
Ransalu Senanayake, Lionel Ott, Simon O'Callaghan, Fabio T. Ramos
Ransalu Senanayake, Lionel Ott, Simon O'Callaghan, Fabio T. Ramos
[obtained, number, probability, probabilistic, collected, regular] [learning, query, problem, assume, conference] [method, discussed, due, processing] [can, regression, main, also, computational] [section, point, kernel, hilbert, data, space, embedding, based, given, gaussian, distribution, two, random, equation] [dynamic, occupancy, model, time, new, laser, future, uncertainty, shm, process, grid, past, continuous, robotics, information, location, sensor, nll, state, raw, environment, dgm, occupied, hinged, ahead, merely] [motion, using, map, spatial, figure, used, mapping, predict, static, feature, dataset, object, predicting, approach, moving, international, frame, table, field, neural, similar]
Deconvolving Feedback Loops in Recommender Systems
Ayan Sinha, David F. Gleich, Karthik Ramani
Ayan Sinha, David F. Gleich, Karthik Ramani
[number, probability, influence, identify, likely, possible, induced] [algorithm, show, set, consider, considered, make, now] [line, datasets, parameter, supplementary, method, compute, due, validate] [matrix, recommender, rtrue, rating, observed, robs, user, collaborative, item, deconvolved, filtering, deconvolving, singular, assumption, ttrue, recommendation, plot, netflix, recommended, synthetic, can, see, also, first, hyperbola, recover, jester, rrecom, progressively, preference] [true, data, metric, based, given, density, equation] [feedback, model, system, effect, time, state, information, value, varying, straight, future, implicit] [score, figure, similarity, use, using, approach, dataset, higher, without, able, per, final]
Bi-Objective Online Matching and Submodular Allocations
Hossein Esfandiari, Nitish Korula, Vahab Mirrokni
Hossein Esfandiari, Nitish Korula, Vahab Mirrokni
[probability, balance, assign, weighted, wij, present, edge, total] [online, algorithm, allocation, submodular, set, budgeted, problem, super, greedy, competitive, theorem, ratio, lemma, welfare, let, hardness, swm, show, vahab, proof, upper, virtual, optimal, bij, provide, allocate, consider, maximization, offline, budget, almost, maximizing, since, tight, function, grp, bound, expected, nitish, special, max, define, may, optj] [objective, optimization, stochastic, run, approximation, dual, marginal, factor] [item, second, one, first, optimum, can, following, fraction, need, analysis] [two, based, result, exponential, curve] [agent, value, goal, gain, resource, maximize, model, assuming] [matching, weight, figure, adversarial, using, previous]
Scaling Factorial Hidden Markov Models: Stochastic Variational Inference without Messages
Yin Cheng Ng, Pawel M. Chilinski, Ricardo Silva
Yin Cheng Ng, Pawel M. Chilinski, Ricardo Silva
[number, message, passing, dependency, david, structure] [algorithm, learning, budget, set, binary, show, class] [variational, inference, fhmm, stochastic, computing, posterior, latent, bivariate, copula, large, elbo, fhmms, scalability, subchains, computed, approximate, factorial, respect, smf, marginal, machine, gradient, size] [can, also, computational, observed] [data, gaussian, distribution, length, chain, emission, given, random, space, equation, based, test, family, two] [markov, long, state, model, simulated, time, allows, information, correlation] [proposed, recognition, hidden, neural, network, sequence, using, learned, approach, different, compared, scale, training, use, structured, figure, per, table, propose, preprint, arxiv, validation]
Probing the Compositionality of Intuitive Functions
Eric Schulz, Josh Tenenbaum, David K. Duvenaud, Maarten Speekenbrink, Samuel J. Gershman
Eric Schulz, Josh Tenenbaum, David K. Duvenaud, Maarten Speekenbrink, Samuel J. Gershman
[structure, number, many, mechanical, average, base] [function, learning, set, since, chosen, best] [extrapolation, experiment, standard, showed, interpolation, university] [spectral, can, sampled, error, first, significantly, linear, via, regression, real] [compositional, kernel, mean, mixture, people, lxp, pxr, lxr, inductive, asked, basis, data, gaussian, radial, intuitive, lin, proportion, functional, space, grammar, predictability, two, given, rbf, perceived, well, chain, turk, expressed, distance, recruited, periodic, simpler, distribution, accepted, assessed, exp, pxlxr] [human, model, cognitive, complex, continuous, world, received, choice, markov, information] [figure, different, approach, input, used, per, shown, generate, using, structured, prefer, last, thought, ground]
Learnable Visual Markers
Oleg Grinchuk, Vadim Lebedev, Victor Lempitsky
Oleg Grinchuk, Vadim Lebedev, Victor Lempitsky
[bit, thus] [learning, loss, conference, might, show, make, defined] [capacity, transforms, augmented, machine, size, stochastic] [can, certain, also, one, sampled, matrix, recover, gram, related, suitable] [two, robust, random, geometric, based, affine, well, associated, corresponding] [information, design, process, system, code, encoding, adapt, maximize, environment] [network, marker, recognizer, synthesizer, texture, visual, deep, approach, using, convolutional, computer, recognition, neural, image, vision, rendering, trained, used, figure, color, input, layer, use, architecture, implemented, spatial, different, sequence, pattern, learned, renderer, single, international, style, training, natural, work, printing, fiducial, blur, output, generate, pretrained, performance]
Improved Regret Bounds for Oracle-Based Adversarial Contextual Bandits
Vasilis Syrgkanis, Haipeng Luo, Akshay Krishnamurthy, Robert E. Schapire
Vasilis Syrgkanis, Haipeng Luo, Akshay Krishnamurthy, Robert E. Schapire
[probability, admissible, number, thus, variable, definition] [max, algorithm, regret, contextual, bound, relaxation, cost, let, will, now, observe, upper, learner, learning, set, strategy, oracle, bandit, setting, minimax, rakhlin, online, lemma, rademacher, drawn, robert, problem, sridharan, optimal, access, bounded, show, defined, best, chooses, since, proof, assume, cumulative, consider, theorem, maxy] [optimization, coordinate, efficient, argmin, unbiased, term, computed, end, compute, standard] [can, min, also, first, denote, vector, one, following] [random, distribution, equal, based, maximum, quantity, given, mass, equation, section] [value, policy, framework, information, action, time, goal] [adversarial, sequence, using, use, recent, work]
A posteriori error bounds for joint matrix decomposition problems
Nicolo Colombo, Nikos Vlassis
Nicolo Colombo, Nikos Vlassis
[obtained, strictly] [problem, bound, defined, upper, set, conference, theorem, inequality, since, algorithm, proof, lower, feasible, show, case, learning, let, implies, smallest, function] [optimization, machine, approximate, operator, parameter, triangular, siam, processing, objective, due] [matrix, low, plow, decomposition, can, schur, analysis, triangularizer, orthogonal, signal, noise, skew, perturbation, exact, tensor, simultaneous, error, posteriori, triangularization, one, diagonalization, eigenvalue, observed, via, following, triangularizers, first, closest, ieee, norm, nonsymmetric, diagonalizable, synthetic, hence, also, frobenius, side] [joint, journal, two, empirical, estimation, manifold, distance, canonical, uinit] [] [used, international, jointly, use, using, approach, figure, ground]
LightRNN: Memory and Computation-Efficient Recurrent Neural Networks
Xiang Li, Tao Qin, Jian Yang, Xiaolin Hu, Tieyan Liu
Xiang Li, Tao Qin, Jian Yang, Xiaolin Hu, Tieyan Liu
[probability, node, number, unique, total, represent, larger] [will, algorithm, allocation, set, problem, achieves, loss, since, complexity, allocate, cost, learning, best] [size, large, log, processing, factor, calculate, datasets, key] [row, can, column, vector, also, one, computational, still, need, technical, matrix, see] [word, embedding, vocabulary, two, associated, based, test, distribution, given] [model, next, modeling, time, benchmark, hierarchical, state, share, information] [lightrnn, language, training, neural, table, rnn, recurrent, figure, perplexity, preprint, arxiv, shared, network, use, memory, aclw, compared, several, using, position, used, gpu, dataset, hidden, billionw, input, proposed, reducing, matching, natural, softmax]
On Mixtures of Markov Chains
Rishi Gupta, Ravi Kumar, Sergei Vassilvitskii
Rishi Gupta, Ravi Kumar, Sergei Vassilvitskii
[number, probability, definition, grows, block] [let, algorithm, since, set, show, consider, learning, problem, every, lemma, will, exists, now, setting, case, make, defined] [full, compute, step] [matrix, error, can, first, rank, diagonal, note, one, denote, see, decomposition, also, recovery, shuffle, condition, column, vector, app, recover, recovering, spectral, real, entry] [mixture, chain, given, section, two, distribution, length, underlying, together] [markov, state, transition, starting, form, jth, research] [use, using, different, figure, reconstruction, hidden, recall, work, performance, approach, generated, reconstructing]
Understanding Probabilistic Sparse Gaussian Process Approximations
Matthias Bauer, Mark van der Wilk, Carl Edward Rasmussen
Matthias Bauer, Mark van der Wilk, Carl Edward Rasmussen
[adding, number, many, obtained] [function, show, learning, will, complexity, bound, always, conference, lower] [inducing, fitc, vfe, full, objective, marginal, qff, likelihood, fit, term, variance, posterior, heteroscedastic, variational, optimisation, kff, optimised, log, approximation, additional, lengthscales, extra, inference, snelson, approximate, method, processing, away, intelligence, university, optimiser, machine] [can, noise, sparse, penalty, good, remark, reduced, solution, also, still, see] [data, gaussian, true, covariance, section, mean, practical, two, common, conditional] [behaviour, process, model, predictive, artificial, change, information, placed, trace, uncertainty] [input, training, using, neural, dataset, without, figure, top, investigate, like, fully, improves]
Understanding the Effective Receptive Field in Deep Convolutional Neural Networks
Wenjie Luo, Yujia Li, Raquel Urtasun, Richard Zemel
Wenjie Luo, Yujia Li, Raquel Urtasun, Richard Zemel
[many, number, larger] [case, get, consider, study, make, may, show, loss] [size, gradient, variance, standard, central, large] [can, analysis, also, note, one, initialization, still, much, see, following, relative, assumption, linear, first, signal] [random, gaussian, theoretical, kernel, distribution, center, nonlinear, well, section, uniform, empirical, fourier] [effect, information, change, within, therefore, binomial] [receptive, field, erf, deep, effective, convolution, convolutional, output, image, input, training, use, network, impact, neural, pixel, unit, layer, weight, relu, cnn, dilated, cnns, different, semantic, arxiv, preprint, like, used, activation, understanding, figure, residual, single, object, classification, using, shown, trained]
Backprop KF: Learning Discriminative Deterministic State Estimators
Tuomas Haarnoja, Anurag Ajay, Sergey Levine, Pieter Abbeel
Tuomas Haarnoja, Anurag Ajay, Sergey Levine, Pieter Abbeel
[probabilistic, piecewise, graph, knowledge, number] [since, conference, set, function, difficulty] [deterministic, standard, latent, method, gradient, normalization, inference] [can, computation, error, also, filtering, require, need] [estimation, test, two, based, nonlinear, distribution, space, well, corresponding, estimate, conditional] [state, model, observation, kalman, simple, raw, tracking, complex, time, directly, typically, optimize, backpropagation, design, corresponds] [neural, network, training, discriminative, recurrent, filter, lstm, generative, trained, feedforward, approach, using, bkf, figure, task, use, convolutional, image, visual, sequence, kitti, domain, international, dst, predict, used, train, camera, hidden, performance, entire, computer, shown, preprint, including, vision]
Recovery Guarantee of Non-negative Matrix Factorization via Alternating Updates
Yuanzhi Li, Yingyu Liang, Andrej Risteski
Yuanzhi Li, Yingyu Liang, Andrej Risteski
[potential, larger, level, negative, present, thus] [algorithm, case, will, theorem, even, general, since, learning, problem, function, show, proof, bound, consider, constant, max, lemma, guarantee, assume, let, upper, induction, assumed, heavy, focus, unknown] [large, unbiased, update, requires, provably, step, updating] [matrix, noise, can, decoding, small, also, factorization, order, assumption, one, much, interesting, note, nmf, still, related, signal, provable, recover, practice, sparse, popular, fraction, analysis, norm, see, alternating, denote] [data, topic, two, mild, theoretical, positive, robust, section] [model, potentially, simplified] [feature, adversarial, relu, use, work, different, intermediate, generative, used, similar]
Ancestral Causal Inference
Sara Magliacane, Tom Claassen, Joris M. Mooij
Sara Magliacane, Tom Claassen, Joris M. Mooij
[causal, ancestral, aci, independence, mek, fci, hej, pkc, erk, jnk, plcg, pka, raf, cfci, akt, discovery, possible, weighted, anytime, number, acyclic, directed, probability, scoring, frequentist, reliability, relation, represents] [confidence, show, obtain, will, set, function, loss, precision] [standard, method, supplementary, latent, bayesian, optimization, inference, log] [can, order, also, one, synthetic] [test, data, observational, conditional, given, statistical, two, distribution, interventional] [bootstrapped, execution, direct, experimental, encoding, time] [using, use, figure, input, score, several, different, like, used, weight, propose, recall, approach]
SEBOOST - Boosting Stochastic Learning Using Subspace Optimization Techniques
Elad Richardson, Rom Herskovitz, Boris Ginsburg, Michael Zibulevsky
Elad Richardson, Rom Herskovitz, Boris Ginsburg, Michael Zibulevsky
[number, added, many, balance] [algorithm, learning, set, problem, function, will, rate] [optimization, stochastic, method, sgd, seboost, step, descent, boosting, secondary, experiment, nag, gradient, size, sesop, adagrad, large, momentum, michael, although, sequential, usually, mse, processing] [subspace, can, one, error, anchor, overall, note, following, small, vanilla, regression] [test, based, section, two, point, spanned] [baseline, direction, current, process, effect, taking, time, change, significant, information] [different, figure, mnist, autoencoder, previous, original, deep, applied, used, neural, last, train, training, using, achieved, applying, classification, composed, better]
Learning Bound for Parameter Transfer Learning
Wataru Kumagai
Wataru Kumagai
[probability, definition, knowledge, often, number] [learning, bound, theorem, learnability, algorithm, show, set, satisfies, setting, consider, inequality, let, since, assume, conference, assumed, hypothesis, case, provide, gray, arbitrary, label, zhao, margin, expected, permissible, bounded, kind, exists, known] [parameter, machine, although, processing, large] [sparse, dictionary, stability, following, local, assumption, can, first, note, denote, matrix, perturbation, also, suppose, norm, regularization, paper, noise] [section, parametric, sample, based, space, theoretical, data, radius, estimator, exp, refer, useful] [target, coding, model, corresponds] [transfer, source, region, feature, representation, mapping, unlabeled, neural, used, performance, domain, effective, international, approach, using, applying, unsupervised, task, yang]
Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling
Jiajun Wu, Chengkai Zhang, Tianfan Xue, Bill Freeman, Josh Tenenbaum
Jiajun Wu, Chengkai Zhang, Tianfan Xue, Bill Freeman, Josh Tenenbaum
[probabilistic, structure] [learning, show, loss, observe] [latent] [can, vector, also, following] [space, distribution, two, based, sample, test, data] [model, modeling, framework, demonstrate] [object, generative, training, adversarial, image, figure, shape, discriminator, representation, generator, generated, learned, use, network, single, convolutional, deep, without, volumetric, reconstruction, classification, ikea, used, table, sharma, girdhar, synthesis, synthesize, generate, voxel, discriminative, consists, using, unsupervised, neural, input, previous, thomas, evaluate, performance, semantic, radford, preprint, recent, synthesized, visualize, novel, supervised, generating, proposed, arxiv, trained, able, different, dataset, supervision, higher, last]
An urn model for majority voting in classification ensembles
Victor Soto, Alberto Suárez, Gonzalo Martinez-Muñoz
Victor Soto, Alberto Suárez, Gonzalo Martinez-Muñoz
[ensemble, number, siba, disagreement, voting, complete, queried, urn, probability, average, pruning, lookup, possible, partial, vote, knowledge, majority, halted, made, querying, sonar, hypergeometric] [class, learning, expected, conference, instance, set, confidence, equivalent, label, remaining, case, problem, rate, considered, upper, algorithm, will] [hyper, machine, compute, size, method, faster, full] [can, one, error, analysis, proposition, following] [distribution, uniform, given, estimate, random, statistical, specified, closer, test, estimated, based, fixed, data] [prior, process, decision, dynamic, individual, assuming, new] [using, used, table, prediction, classification, training, different, international, color, proposed, final, use, classifier]
Large Margin Discriminant Dimensionality Reduction in Prediction Space
Mohammad Saberian, Jose Costa Pereira, Nuno Nvasconcelos, Can Xu
Mohammad Saberian, Jose Costa Pereira, Nuno Nvasconcelos, Can Xu
[number, combination, possible] [learning, multiclass, algorithm, set, svm, margin, duality, class, risk, loss, example, defined, hashing, max, optimal, binary] [boosting, codewords, discriminant, codeword, mcboost, large, reduction, method, iteration, update, gradient, optimization] [can, linear, dimension, note, arg, sign, ieee, pca, error] [predictor, dimensionality, data, space, embedding, kernel, based, two, given, dimensional, basis, wzi, preserving, procedure] [decision, current, learns, traffic, new] [ladder, using, learn, mapping, use, figure, prediction, classifier, classification, learned, neural, propose, deep, feature, implemented, intermediate, proposed, compared, image, multiple, different, performance, scene, weak, training, dataset, jointly, table, approach, computer, used, sij]
Deep Alternative Neural Network: Exploring Contexts as Early as Possible for Action Recognition
Jinzhuo Wang, Wenmin Wang, xiongtao Chen, Ronggang Wang, Wen Gao
Jinzhuo Wang, Wenmin Wang, xiongtao Chen, Ronggang Wang, Wen Gao
[number, connected, larger, often, split, possible, contains] [adaptive, arbitrary, context, class, learning, competitive] [size, method, preserve, standard, clipping] [can, local, first, one, also, comparison] [length, data, kernel, sample, two] [temporal, action, alternative, early, time, current, determine, model, human, followed] [layer, recurrent, input, dann, volumetric, video, convolutional, deep, pooling, network, neural, table, flow, figure, pyramid, optical, performance, recognition, feature, spatial, visual, impact, fully, training, using, semantic, architecture, use, unit, similar, preprint, arxiv, used, motion, clip, output, accuracy, different, previous, investigate, applied, three, fusion]
Latent Attention For If-Then Program Synthesis
Chang Liu, Xinyun Chen, Eui Chul Shin, Mingcheng Chen, Dawn Song
Chang Liu, Xinyun Chen, Eui Chul Shin, Mingcheng Chen, Dawn Song
[program, either, sum] [function, learning, set, active, observe, will, achieve, problem, best, show, consider, example] [latent, standard, size, supplementary] [can, also, first, one, existing, dictionary] [embedding, two, token, refer, test, data] [action, model, new, determine, code, take, simple, prior] [attention, trigger, training, using, figure, language, natural, accuracy, neural, prediction, channel, output, softmax, semantic, description, input, better, work, network, task, predicting, approach, instagram, parsing, weight, use, three, ensembling, rebalanced, sequence, trained, dataset, synthesis, bdlstm, different, preprint, train, arxiv, performance, recipe, improve, presented, previous, similar]
End-to-End Goal-Driven Web Navigation
Rodrigo Nogueira, Kyunghyun Cho
Rodrigo Nogueira, Kyunghyun Cho
[node, number, graph, page, evaluating, allowed, many, find, probability, thus, edge, average] [query, set, learning, example, make] [step, challenging, away] [one, first, vector, wikipedia, can, also] [based, test, two, given, maximum, selected, word] [agent, web, target, navigation, neuagent, wikinav, starting, search, information, world, action, state, current, artificial, human, focused, benchmark, website, webnav, model, outgoing, kentuchy, stop, time, next, software, making, frec, english, navigates, neuagents] [task, proposed, natural, content, neural, language, table, use, whole, description, training, trained, hidden, pretrained, using, evaluate, consists, representation, derby, used, understanding, beam]
A Non-convex One-Pass Framework for Generalized Factorization Machine and Rank-One Matrix Sensing
Ming Lin, Jieping Ye
Ming Lin, Jieping Ye
[probability, recursive, called, chen, whose] [lemma, theorem, algorithm, will, learning, convex, since, define, rate, bound, constant, complexity, bounded, show, hold, least, proof, version] [operator, efficient, step, convergence, standard, sampling, machine, key, method, requires] [matrix, sensing, gfm, rank, order, symmetric, recovery, rip, can, alternating, suppose, via, condition, low, norm, factorization, provable, noise, noisy, generalized, perturbation, following, denote, first, linear, one, need, second, completion, main, singular, prateek, sampled, error, high, cai] [estimation, provided, independent, theoretical, random, construct, gaussian, section, fixed, construction, based] [framework, target, therefore, trace, value, information] [sequence, training, several, proposed, using, feature]
Regularized Nonlinear Acceleration
Damien Scieur, Alexandre d'Aspremont, Francis Bach
Damien Scieur, Alexandre d'Aspremont, Francis Bach
[thus, number, degree, chebyshev] [problem, algorithm, max, convex, bound, will, let, now, optimal, function, assume, case, since, rate, oracle, defined, scheme] [gradient, method, acceleration, optimization, convergence, iterates, ampe, extrapolation, rmpe, parameter, conjugate, computed, solve, siam, accelerated, approximate, smooth, respect, compute, strongly, solves] [can, solution, linear, matrix, minimal, min, regularized, vector, condition, following, proposition, small, order, iterative, also, solving, norm, need, eigenvalue, note, error, regularization, perturbation, explicit, optimum, formulation, generic] [polynomial, nonlinear, numerical, estimate, point, equal, section, journal, fixed, equation, given, classical, mathematical, result] [information, control] [using, use, figure, without, similar, produced, performance]
Optimal Learning for Multi-pass Stochastic Gradient Methods
Junhong Lin, Lorenzo Rosasco
Junhong Lin, Lorenzo Rosasco
[number, probability, larger, total] [learning, optimal, inf, corollary, let, show, consider, least, setting, since, function, proof, derive, algorithm, theorem, rate, case, set, defined, stopping, latter, unknown, studied] [sgm, stochastic, gradient, batch, pass, convergence, capacity, machine, processing, size, variance, parameter, term, large, lead, distributed, optimization] [error, can, one, assumption, computational, regularization, also, see, main, regression, first, following, order, note, related] [sample, bias, fixed, considering, independent, given, data, measure, generalization, two, space, empirical, result, well, random, finite] [simple, information, choice, effect] [using, neural, three, multiple, sequence, different, figure]
Learning Structured Sparsity in Deep Neural Networks
Wei Wen, Chunpeng Wu, Yandan Wang, Yiran Chen, Hai Li
Wei Wen, Chunpeng Wu, Yandan Wang, Yiran Chen, Hai Li
[structure, number, average] [learning, achieve, show, achieves, even, conference] [speedup, method, reduce, efficient, approximation, university] [sparsity, can, error, regularization, group, computation, lasso, also, rank, low, matrix, zero, sparse, order, note, first] [] [neuron, model, baseline, within] [ssl, convolutional, figure, dnn, structured, deep, layer, filter, accuracy, lenet, alexnet, neural, depth, table, weight, cpu, learn, gpu, convnet, network, preprint, arxiv, without, flop, input, dnns, regularize, shape, learned, classification, feature, different, resnet, proposed, compared, higher, use, using, xeon, achieved, original, trained, work, connection, zeroed, residual]
Estimating the Size of a Large Network and its Communities from a Random Sample
Lin Chen, Amin Karbasi, Forrest W. Crawford
Lin Chen, Amin Karbasi, Forrest W. Crawford
[number, ulse, nsum, vertex, total, probability, edge, graph, pendant, social, larger, degree, thus, blockmodel, becomes, vary, community, block, denoted, subgraph, induced, men] [let, observe, algorithm, set, pij, smaller, study, theorem, deviation, assume, general, problem] [size, posterior, sampling, likelihood, stochastic, parameter, proposal, large, variance, compute, method] [relative, error, also, sampled, can] [sample, estimation, random, given, distribution, population, estimating, estimate, type, people, mean, estimator, joint, data, bias, two] [effect, model, must, value, prior] [network, performance, propose, randomly, different, better, using, shown]
Asynchronous Parallel Greedy Coordinate Descent
Yang You, Xiangru Lian, Ji Liu, Hsiang-Fu Yu, Inderjit S. Dhillon, James Demmel, Cho-Jui Hsieh
Yang You, Xiangru Lian, Ji Liu, Hsiang-Fu Yu, Inderjit S. Dhillon, James Demmel, Cho-Jui Hsieh
[variable, number, often, rule, block, solver, partition] [algorithm, svm, greedy, will, learning, rate, best, set, function, bound, since, define, assume, problem, selecting] [coordinate, descent, asynchronous, convergence, parallel, stochastic, step, gradient, update, machine, gcd, faster, speedup, objective, covtype, dual, method, optimization, webspam, lres, select, conduct, iteration, libsvm, distributed, asynchronously, rik, large, parallelizing] [can, following, linear, solving, note, matrix, first, also, vector, much, one, projected, analysis, error, global, column] [kernel, section, theoretical] [time, scaling, cyclic] [memory, using, used, training, use, figure, approach, proposed, randomly, work, multiple, scale, implemented, shared, stored]
Strategic Attentive Writer for Learning Macro-Actions
Alexander Vezhnevets, Volodymyr Mnih, Simon Osindero, Alex Graves, Oriol Vinyals, John Agapiou, koray kavukcuoglu
Alexander Vezhnevets, Volodymyr Mnih, Simon Osindero, Alex Graves, Oriol Vinyals, John Agapiou, koray kavukcuoglu
[possible, notice] [learning, defined, general, every, function, set] [update, step, experiment, stochastic] [can, one, first, matrix, also] [section, two, gaussian, random, distribution, useful, given] [straw, action, plan, reinforcement, time, temporal, agent, strawe, state, maze, planning, next, goal, learns, atari, commitment, exploration, model, value, attentive, policy, environment, reward, module, observation, demonstrate, pacman, temporally, thereby, frostbite, followed, corresponds] [network, neural, figure, attention, lstm, using, deep, use, prediction, training, feature, sequence, architecture, representation, trained, score, patch, used, learned, character, convolutional, recurrent, structured, text, proposed, layer, without, learn, different, generate, preprint]
Dynamic Mode Decomposition with Reproducing Kernels for Koopman Spectral Analysis
Yoshinobu Kawahara
Yoshinobu Kawahara
[obtained, number, describe] [set, algorithm, let, consider, learning, known, since, defined, define, considered] [operator, method, approximation, although, calculate, machine] [can, decomposition, matrix, analysis, subspace, spectral, linear, orthogonal, principal, eigenvalue, krylov, one, onto, note, denote, via, vector, gram] [koopman, nonlinear, dmd, kernel, data, eigenfunctions, given, empirical, corresponding, based, journal, procedure, estimated, section, spanned, fluid, reproducing, eigendecomposition, pod, ritz, true, space, basis, distance, toy, principle, observables, two, finite, embedding, calculation] [dynamical, system, mode, dynamic, change, value] [using, feature, proposed, sequence, perform, several, applied, neural, extracted, presented, used, map, prediction, approach]
Computational and Statistical Tradeoffs in Learning to Rank
Ashish Khetan, Sewoong Oh
Ashish Khetan, Sewoong Oh
[pairwise, ordered, subset, graph, number, ordering, among, grb, topology, resulting, partition, lrb, partial, prb, poset, paired, edge, sum] [bound, lower, oracle, set, dependence, provide, complexity, learning, concave, max, upper, utility, let, problem, fix, theorem, choose, general, since, show, chosen, define] [size, log, optimization, large, processing] [computational, order, error, can, user, ranking, inconsistent, generalized, rank, following, one, spectral, analysis, denote, item, first, ordinal, also, remark] [sample, data, estimator, provides, consistent, statistical, mle, canonical, estimate, finite, random, provided] [model, choice, information, simple, significant, preferred] [figure, effective, extracted, proposed, use]
Flexible Models for Microclustering with Application to Entity Resolution
Brenda Betancourt, Giacomo Zanella, Jeffrey W. Miller, Hanna Wallach, Abbas Zaidi, Beka Steorts
Brenda Betancourt, Giacomo Zanella, Jeffrey W. Miller, Hanna Wallach, Abbas Zaidi, Beka Steorts
[cluster, nbd, nbnb, entity, number, pyp, microclustering, clustering, partition, probability, negative, record, exchangeable, false, grow, create, singleton, linkage, italy, date, science, obtained, categorical, implicitly] [set, property, class, appendix, algorithm, will, define, assume] [size, posterior, sampling, bayesian, inference, large, university, negbin, variant] [can, one, analysis, require, via, noisy, assumption] [data, mixture, distribution, random, equation, two, statistical, maximum, mean, finite, dirichlet, percentile, fixed, conditional, infinitely, section, true, survey, empirical, journal, address, yield] [model, prior, process, exhibit, value, belief] [resolution, used, using, four, sequence, approach, generated, three, figure, perform, compare]
Object based Scene Representations using Fisher Scores of Local Subspace Projections
Mandar D. Dixit, Nuno Vasconcelos
Mandar D. Dixit, Nuno Vasconcelos
[combination, obtained] [class, problem, conference, show, function, best, set] [log, factor, variance, respect, large, due, gradient] [vector, can, sparse, local, second, order, linear, ieee, comparison, much] [fisher, gmm, covariance, mixture, based, gaussian, mean, dimensional, two] [model, information, observation, modeling, coding, state] [scene, cnn, object, image, classification, transfer, cnns, recognition, mfa, computer, shown, deep, performance, using, alexnet, vision, table, representation, score, mit, vgg, sun, indoor, used, feature, trained, better, dataset, task, hidden, neural, holistic, accuracy, use, similar, pooling, pattern, patch, training, outperforms, extracted]
Balancing Suspense and Surprise: Timely Decision Making with Endogenous Information Acquisition
Ahmed Ibrahim, Mihaela Van Der Schaar
Ahmed Ibrahim, Mihaela Van Der Schaar
[path, probability, partition, structure, tuple] [optimal, stopping, every, theorem, since, risk, cost, function, will, show, whenever, set, observe, interval, instance, context, observes, assume] [processing, posterior, bayesian, sequential, stochastic] [can, following, hence, order, observed, sensing, denote] [survival, sample, two, random, given, estimate, space, based, journal, type] [time, information, policy, process, belief, adverse, decision, event, realization, new, series, model, rendezvous, next, gain, current, predictable, action, gathering, decides, continuation, unlike, deadline, sensory, stop, whether, occurrence, whereas, surprise, care, acquiring, characterize, exemplary, timely, observing] [neural, figure, generated, captured]
Dialog-based Language Learning
Jason E. Weston
Jason E. Weston
[thus, negative, partial, external] [learning, learner, set, setting, will, case, consider] [machine, datasets] [can, also, one, signal, see, first] [positive, given, described, fixed, test, provided, useful, well] [feedback, answer, teacher, forward, imitation, model, dialog, supplied, supporting, asking, goal, correct, reinforcement, expert, babi, information, reward, response, another, rbi, policy, imitating, student, read, movieqa, state] [task, supervision, language, use, memory, prediction, using, learn, natural, dataset, question, training, output, neural, input, different, network, supervised, predict, used, work, help, text, table, preprint, answering, arxiv, incorrect, relevant, accuracy]
A Bandit Framework for Strategic Regression
Yang Liu, Yiling Chen
Yang Liu, Yiling Chen
[effort, worker, level, payment, exerting, partial, quality, number, collected, exert, sum, acm, exertion, scoring, rule, bne] [will, online, privacy, bandit, learner, index, set, learning, idea, consider, every, function, assume, case, bounded, show, least, competitive, best, theorem, conference, setting, incentivize] [log, update, term, convergence, variance, step] [can, regression, linear, noise, also, one, suppose, following, order, denote, much, solution, note, need, first] [data, two, selection, selected, estimator, bias, ridge, well, sample, differential] [time, model, acquisition, strategic, framework, design, information, target, future, change] [different, using, work, training, task, instead, trained, add, top, performance, propose, mechanism]
Doubly Convolutional Neural Networks
Shuangfei Zhai, Yu Cheng, Zhongfei (Mark) Zhang, Weining Lu
Shuangfei Zhai, Yu Cheng, Zhongfei (Mark) Zhang, Weining Lu
[number, dcnns, consistently, larger, denoted] [set, learning, show, define, conference] [size, standard, parameter, datasets, efficient, processing] [also, can, first, one] [two, data, maximum, section, together] [model, along, correlation, corresponds, information] [convolutional, layer, convolution, dcnn, doubly, neural, deep, filter, double, pooling, cnn, output, meta, image, arxiv, preprint, spatial, table, effective, network, performance, shown, maxout, three, shape, architecture, used, figure, without, different, learned, several, training, compared, translation, vggnet, maxoutcnn, cnns, augmentation, translated, computer, classification, input, alexnet, memory, dataset, improves, sharing, channel]
Deep Exploration via Bootstrapped DQN
Ian Osband, Charles Blundell, Alexander Pritzel, Benjamin Van Roy
Ian Osband, Charles Blundell, Alexander Pritzel, Benjamin Van Roy
[present, many, find, ensemble, number, ian] [learning, algorithm, may, function, optimal, appendix, even, thompson, cumulative, improved, consider, every, strategy, randomized] [efficient, bootstrap, approximate, large, posterior, faster, sampling, maintain] [can, via, require, also, small, benjamin, one, computational, order, initialization] [sample, data, generalization, distribution, random, computationally] [bootstrapped, dqn, exploration, value, agent, reinforcement, state, uncertainty, action, time, across, policy, head, unlike, atari, reward, upon, target, van, dithering, human, planning, extended, complex, temporally, simple] [deep, network, neural, figure, performance, several, approach, single, arxiv, work, preprint, shared, trained, learn, similar, effective]
Relevant sparse codes with variational information bottleneck
Matthew Chalk, Olivier Marre, Gasper Tkacik
Matthew Chalk, Olivier Marre, Gasper Tkacik
[obtained, variable, vertical, relation, subset, thus, find] [algorithm, bound, function, lower, maximizing, just, consider, set] [kib, log, variational, bottleneck, processing, respect, objective, additional, occlusion, variance, standard, latent, likelihood, derivative] [sparse, can, decoding, one, linear, noise, side, iterative, occluded, also, small, zero] [gaussian, kernel, distribution, data, fixed, expansion, way, selection, two, test, closely] [information, encoding, response, coding, model, bar, encoded, unlike, sensory, behaviour, rni] [figure, input, image, neural, used, presented, using, training, approach, relevant, shape, natural, use, original, performance, representation, learned, handwritten]
Optimal Black-Box Reductions Between Optimization Objectives
Zeyuan Allen-Zhu, Elad Hazan
Zeyuan Allen-Zhu, Elad Hazan
[community, probability, detection, graph, threshold, vertex, number, conjecture, nonbacktracking, abp, propagation, block, definition, adjacent, average, achieving, enough, degree, likely, assign, expect, sum, achievability, detect, cycle, presence, linearized, subtract] [algorithm, let, will, set, general, prove, proof, drawn, define, version, return, exists, least] [stochastic, large, solves, step, compute, initial] [one, matrix, also, approximately, can, spectral, symmetric, eigenvector, order, following, vector, eigenvalue, sparse, high, dominant, first] [length, two, random, equal, distribution, described, bias, positive, way, based] [model, belief, value, simply, snr, new] [different, instead, approach, part, use, multiple]
A Constant-Factor Bi-Criteria Approximation Guarantee for k-means++
Dennis Wei
Dennis Wei
[cluster, potential, clustering, base, resulting, adding, number, probability, report] [lemma, algorithm, optimal, theorem, bound, case, proof, constant, uncovered, set, since, corollary, implies, guarantee, ratio, known, let, bounded, selecting, satisfy, expected, june, problem, contribution, march, cost, function, assume] [approximation, sampling, factor, expectation, step, respect, term, supplementary, size] [also, first, can, one, technical, running, local, paper, sufficient, denote, main, minimum, existing, following, min] [given, section, center, two, inductive, result, data, euclidean, provided, sense, metric] [covered, search, taking, new, upon, therefore] [using, shown, used, work, randomly, improves]
Data driven estimation of Laplace-Beltrami operator
Frederic Chazal, Ilaria Giulini, Bertrand Michel
Frederic Chazal, Ilaria Giulini, Bertrand Michel
[graph, probability, calibration, laplacians, introduced, theory, according, many] [function, theorem, defined, let, consider, max, oracle, inequality, problem, learning, risk, bound, adaptive, lemma, selecting, known, assume, least, instance, constant, pointwise] [method, variance, operator, convergence, smooth, term, machine, standard, large, approximation] [can, following, paper, spectral, first, proposition, see, analysis, sampled] [laplacian, given, bandwidth, section, data, estimator, bias, exp, procedure, estimation, riemannian, manifold, sphere, mikhail, selected, family, mathematical, partha, kernel, quantity, belkin, finite, sample, gaussian, two, metric, jump, density, geometric] [model, driven] [approach, proposed, various, pascal, figure, previous]
Causal Bandits: Learning Good Interventions via Causal Inference
Finnian Lattimore, Tor Lattimore, Mark D. Reid
Finnian Lattimore, Tor Lattimore, Mark D. Reid
[causal, graph, variable, intervention, number, weighted, identification, knowledge, thus, possible] [bandit, algorithm, regret, problem, optimal, case, general, learning, set, arm, may, will, depends, focus, best, learner, consider, cumulative, show, theorem, selecting, known, assume, class, unknown, observe, max, contextual] [parallel, sampling, extra, importance, sequential, stochastic, standard, large, variance, log] [can, also, observed, low, one, side, via, truncated, analysis] [observational, distribution, estimate, fixed, given, estimator, conditional] [reward, simple, model, action, value, feedback, information, choice, simultaneously, framework, experimental, design, future] [used, use, figure, improve, work, applying]
Structured Prediction Theory Based on Factor Graph Complexity
Corinna Cortes, Vitaly Kuznetsov, Mehryar Mohri, Scott Yang
Corinna Cortes, Vitaly Kuznetsov, Mehryar Mohri, Scott Yang
[graph, present, number, theory, possible, many, dependency] [loss, learning, general, complexity, margin, hypothesis, upper, will, set, function, may, bound, rademacher, voted, theorem, defined, surrogate, lemma, arbitrary, max, convex, special, proof, bounded, show, appendix, vcrf, example, assume, case, consider, hold, known] [factor, log, size, standard, term, processing, machine] [can, also, following, analysis, explicit, linear, sparsity, first] [section, empirical, based, random, generalization, family, conditional, result, space, given, principle, sample, theoretical, additive] [new, complex, markov, along] [structured, prediction, used, using, output, use, field, natural, language, several]
An Architecture for Deep, Hierarchical Generative Models
Philip Bachman
Philip Bachman
[merge, indicates, structure, many] [learning, conference, bound, show, provide, set] [latent, inference, stochastic, machine, log, variational, deterministic, processing, sequential, stage, posterior, lsun, sampling] [can, one, also, first, see, local, second] [conditional, distribution, data, gaussian, mixture, described, section, two] [model, module, state, information, hierarchical, prior, measured, current, ability, trainable] [network, generative, using, performance, matnets, used, use, matnet, international, deep, architecture, neural, three, image, output, cifar, depth, training, modelling, figure, trained, final, meta, quantitative, omniglot, residual, convolutional, updated, generated, ladder, back, input, work, generation, reconstruction]
Fast recovery from a union of subspaces
Chinmay Hegde, Piotr Indyk, Ludwig Schmidt
Chinmay Hegde, Piotr Indyk, Ludwig Schmidt
[number, block, structure] [complexity, algorithm, let, theorem, problem, constant, appendix, general, show, set, guarantee, arbitrary, achieve, case, give, achieves] [approximate, approximation, gradient, faster, size, descent, apply, large, log, requires] [matrix, recovery, projection, subspace, linear, can, also, exact, tail, running, svd, singular, vector, order, following, svp, note, krylov, small, svds, sparse, sufficiently, rip, compressive, algorithmic, sparsity, recovering, hence, orthogonal, union, projected] [sample, two, section, important, corresponding, statistical, result, empirical] [time, head, framework, value, model, design, prior, new] [use, structured, using, work, approach, several, better, already, similar]
Synthesizing the preferred inputs for neurons in neural networks via deep generator networks
Anh Nguyen, Alexey Dosovitskiy, Jason Yosinski, Thomas Brox, Jeff Clune
Anh Nguyen, Alexey Dosovitskiy, Jason Yosinski, Thomas Brox, Jeff Clune
[quality, many, thus] [learning, set, modified, best, show, maximization] [method, invert] [can, one, highly, global, synthetic, real, also] [space, two, well] [neuron, prior, preferred, code, model, information, human, whether, target, found] [image, trained, dnn, network, different, deep, learned, activation, neural, feature, generator, imagenet, layer, caffenet, architecture, convolutional, hidden, visualizing, natural, training, visualize, generative, dataset, shown, generalizes, recognition, output, encoder, synthesizing, input, dnns, activates, figure, classify, vision, train, visualized, visualization, previous, dgn, produce, computer, pattern, activate, generalize, alexnet, without, qualitatively, understanding, improve, produced, mit, synthesized]
Local Minimax Complexity of Stochastic Convex Optimization
sabyasachi chatterjee, John C. Duchi, John Lafferty, Yuancheng Zhu
sabyasachi chatterjee, John C. Duchi, John Lafferty, Yuancheng Zhu
[department, number, flat, david, larger] [minimax, convex, complexity, algorithm, function, modulus, inf, risk, continuity, binary, optimal, oracle, sup, set, consider, class, show, define, lower, let, superefficiency, difficulty, subgradient, constant, interval, will, now, rate, case, adaptive, general, problem, defined, achieves, example, satisfies, lipschitz, query, proof, logarithmic, setting] [stochastic, optimization, gradient, descent, log, university, stepsize, derivative] [local, error, analysis, can, computational, minimum, suppose, also, noise, note, following, min, main, arg] [given, result, statistical, point, estimation, section, analogue, polynomial, random] [search, information, benchmark, current, optimizing, alternative, form] [traditional, work, figure]
Measuring Neural Net Robustness with Constraints
Osbert Bastani, Yani Ioannou, Leonidas Lampropoulos, Dimitrios Vytiniotis, Aditya Nori, Antonio Criminisi
Osbert Bastani, Yani Ioannou, Leonidas Lampropoulos, Dimitrios Vytiniotis, Aditya Nori, Antonio Criminisi
[find, constraint, according, signed, number, describe] [algorithm, set, show, example, since, convex, label, may, learning, pointwise, feasible, function, finding, consider, notion, problem, conference] [optimization, gradient, approximate, compute, computing, parameter, improving] [can, linear, frequency, also, def, perturbation, solving, one, high] [robustness, robust, test, given, distance, two, based, nearest] [baseline, substantially, found, search] [adversarial, neural, net, using, figure, training, use, input, used, work, accuracy, severity, deep, original, network, lenet, mnist, approach, improve, nin, arxiv, relu, layer, generated, region, preprint, compared, trained, image, labeled, fails, evaluate]
Visual Dynamics: Probabilistic Future Frame Synthesis via Cross Convolutional Networks
Tianfan Xue, Jiajun Wu, Katherine Bouman, Bill Freeman
Tianfan Xue, Jiajun Wu, Katherine Bouman, Bill Freeman
[possible, probabilistic, often] [algorithm, show, problem, set, learning, function] [variational, latent, method, variance, deterministic, michael] [also, can, observed, real, first] [distribution, conditional, difference, sample, mean, test, given, two, kernel, toy, reference] [model, future, learns, movement, next, simple] [motion, image, network, figure, input, frame, convolutional, using, video, training, cross, visual, different, prediction, generative, feature, representation, encoder, single, proposed, layer, deep, flow, learned, without, dataset, synthesize, map, synthesis, field, unsupervised, propose, neural, shown, learn, multiple, decoder, autoencoder, scale, rgb, predict, use, novel, work, able, optical, recognition]
Efficient and Robust Spiking Neural Circuit for Navigation Inspired by Echolocating Bats
Bipin Rajendran, Pulkit Tandon, Yash H. Malviya
Bipin Rajendran, Pulkit Tandon, Yash H. Malviya
[detection, average, obtained] [rate, even, function, constant, will, chosen] [sampling, processing, increase, aim] [noise, signal, can, error, frequency, success, also, observed] [two, uniform, based, additive, difference, inspired, section, fixed] [spike, azimuth, model, snn, poisson, rms, bat, spiking, head, intensity, angle, encoding, target, time, left, neuron, tracking, pid, information, particle, ear, right, response, lso, system, echolocation, avcn, varying, turn, navigation, khz, dnll, received, biological, superior, effect, arrival, interaural, echo, aor, receiver, synaptic, within, angular] [network, input, source, sound, figure, performance, neural, higher, output, layer, detected, using, compared, use, different]
Short-Dot: Computing Large Linear Transforms Distributedly Using Coded Short Dot Products
Sanghamitra Dutta, Viveck Cadambe, Pulkit Grover
Sanghamitra Dutta, Viveck Cadambe, Pulkit Grover
[block, number, node, combination, subset] [strategy, set, expected, now, let, case, provide, complexity, exists, problem, learning, since] [parallel, computing, compute, transforms, processing, distributed, computed, parameter, faster, requires, size, reduce, university] [matrix, computation, can, row, linear, dot, processor, sparsity, uncoded, column, vector, one, repetition, also, short, analysis, sufficient, wait, indexed, bcols, square] [length, given, maximum, data, redundancy, exponential, mean] [time, required, experimental, encoding, divide, choice] [using, generate, fusion, different, figure, table, compare, shown, pattern, single, task, entire, instead]
VIME: Variational Information Maximizing Exploration
Rein Houthooft, Xi Chen, Xi Chen, Yan Duan, John Schulman, Filip De Turck, Pieter Abbeel
Rein Houthooft, Xi Chen, Xi Chen, Yan Duan, John Schulman, Filip De Turck, Pieter Abbeel
[possible, according, adding] [learning, compression, lower, bound, make, improvement, maximizing, strategy, parametrized, show, assume, supported, defined] [variational, dkl, divergence, posterior, bayesian, gradient, method, bnn, approximation, approximate, inference, term, compute, sampling, log, berkeley, performs] [can, following, noise, sparse, also] [distribution, gaussian, section, practical, given, random, polynomial] [exploration, information, vime, reinforcement, continuous, reward, state, model, agent, control, intrinsic, gain, curiosity, policy, action, environment, value, simple, allows, trpo, optimizing, surprise, heuristic, history, mountaincar, research, behavior] [using, performance, figure, neural, use, deep, fully, proposed, without, better, different, prediction, task, propose]
Neural Universal Discrete Denoiser
Taesup Moon, Seonwoo Min, Byunghan Lee, Sungroh Yoon
Taesup Moon, Seonwoo Min, Byunghan Lee, Sungroh Yoon
[probability, rule, obtained, average] [function, binary, context, algorithm, will, loss, learning, since, define, set, show, concentration, lemma, always, setting, obtain, obtaining, cost] [size, objective, large] [can, noisy, note, error, arg, also, order, real, significantly, matrix, see, ieee, vector, linear, much, following, denote] [data, discrete, universal, estimated, true, given, estimate, underlying, result, based] [model, location, framework, future, information] [dude, neural, denoising, sequence, image, ber, clean, denoiser, window, nanopore, dna, used, performance, training, symbol, figure, using, deep, supervised, source, network, sliding, dnn, shown, lnew, different, generated, use, several, minion, channel, single, grayscale, mentioned]
Learning from Rational Behavior: Predicting Solutions to Unknown Linear Programs
Shahin Jabbari, Ryan M. Rogers, Aaron Roth, Steven Z. Wu
Shahin Jabbari, Ryan M. Rogers, Aaron Roth, Steven Z. Wu
[number, constraint, intersection, subset, vertex, contains] [learnedge, problem, unknown, learning, bound, set, known, mistake, feasible, algorithm, day, polytope, learner, show, ellipsoid, lemma, will, theorem, changing, study, learnhull, optimal, make, function, revealed, precision, questionable, conference, convex, interval, infeasible, learnellipsoid, let, consider, give, may, bounded, upper, defined, pac, whenever, always, arbitrary, know] [objective, update, additional, optimization] [can, assumption, solution, linear, observed, also, following, denote, dimension] [given, point, finite, section, specified, fixed, polynomial] [information, written, time, goal, new, must] [prediction, predict, region, predicting, sequence, natural, learn]
Hypothesis Testing in Unsupervised Domain Adaptation with Applications in Alzheimer's Disease
Hao Zhou, Vamsi K. Ithapu, Sathya Narayanan Ravi, Vikas Singh, Grace Wahba, Sterling C. Johnson
Hao Zhou, Vamsi K. Ithapu, Sathya Narayanan Ravi, Vikas Singh, Grace Wahba, Sterling C. Johnson
[consistency] [will, hypothesis, may, observe, class, whenever, set, problem, show, assume, convex, provide, algorithm, even, theorem, known, consider, learning, since, function] [objective, log, size, optimization, due] [can, minimal, power, linear, min, one, first, error, related, relative, need, ieee, technical, analysis] [mmd, data, sample, estimation, testing, statistical, test, kernel, based, transformation, procedure, section, statistic, estimate, csf, distance, two, null, discrepancy, given, correspond, mean, exp, well, space, biomarkers, corresponding, normal] [model, target, transformed, simply, directly, disease] [domain, source, adaptation, feature, different, using, used, unsupervised, performance, use, multiple, work, recall]
Unsupervised Risk Estimation Using Only Conditional Independence Structure
Jacob Steinhardt, Percy S. Liang
Jacob Steinhardt, Percy S. Liang
[thus, variable, likely, seed, structure, independence] [risk, learning, loss, theorem, algorithm, even, class, show, will, obtain, make, problem, hinge, complexity, case, close] [log, method, machine, compute, large, full, term] [can, assumption, error, tensor, suppose, also, def, matrix, need, see, recover, hence, small, decomposition] [estimate, conditional, estimation, test, given, data, estimating, section, independent, sample, random, distribution, true, estimated, family, result, two, journal, needed] [model, shift, framework] [unsupervised, unlabeled, using, domain, training, figure, use, without, perform, approach, labeled, adaptation, shown, different, prediction, train, structured, conditioned, single, classification, used, work]
Crowdsourced Clustering: Querying Edges vs Triangles
Ramya Korlakai Vinayak, Babak Hassibi
Ramya Korlakai Vinayak, Babak Hassibi
[triangle, edge, clustering, block, adjacency, probability, graph, cluster, filled, number, obtained, inside, querying, crowdsourcing, reliable, confusion, possible, denoted, belong, program, hit, crowd, total] [query, consider, cost, conference, since, learning, convex, label, will, provide, make, budget, give, revealed, let] [processing, size] [matrix, spectral, can, via, note, following, comparison, error, partially, real, significantly, relative, recovery, reveal, one, condition, hence, also] [random, true, two, data, independent, fixed, based, conditional, density, empirical] [model, information, observing, observation, answer, time] [figure, use, different, neural, table, performance, using, compare, dataset, compared, work]
A Multi-Batch L-BFGS Method for Machine Learning
Albert S. Berahas, Jorge Nocedal, Martin Takac
Albert S. Berahas, Jorge Nocedal, Martin Takac
[node, number] [algorithm, learning, function, set, every, show, problem, convex, let, strategy, case] [gradient, method, batch, compute, convergence, stochastic, step, webspam, sgd, descent, iteration, bfgs, distributed, updating, gksk, optimization, implementation, computing, objective, large, approximation, size, machine, nonconvex, employ, variant, mpi, evaluation] [can, one, hessian, computational, small, also, analysis, computation, paper, fraction] [robust, sample, data, length, based, numerical, illustrates, given, correction, journal, two] [time, communication, scaling, direction, new] [using, different, figure, approach, training, overlap, used, proposed, use, similar, generated, without, consecutive, performance]
Fast and accurate spike sorting of high-channel count probes with KiloSort
Marius Pachitariu, Nicholas A. Steinmetz, Shabnam N. Kadir, Matteo Carandini, Kenneth D. Harris
Marius Pachitariu, Nicholas A. Steinmetz, Shabnam N. Kadir, Matteo Carandini, Kenneth D. Harris
[number, average, cluster, threshold, thus, sorted, total, identified, false, overlapping, find, obtained] [best, algorithm, cost, function, set, show] [optimization, large, variance, parallel, standard] [can, matrix, noise, local, running, filtering, also, small, svd, amplitude] [data, two, mean, positive, based, covariance, common, automated] [spike, template, kilosort, sorting, klustakwik, model, temporal, waveform, raw, recording, across, required, pursuit, electrode, time, recorded, typically, electrical, found, prototypical, voltage, miss, decision, new, spatiotemporal, framework, manual, allows, location] [matching, channel, spatial, score, previous, using, single, ground, several, different, reconstruction, figure, feature, three, used]
Linear-Memory and Decomposition-Invariant Linearly Convergent Conditional Gradient Algorithm for Structured Polytopes
Dan Garber, Dan Garber, Ofer Meshi
Dan Garber, Dan Garber, Ofer Meshi
[number, pairwise, vertex, definition] [algorithm, convex, rate, let, problem, set, since, follows, case, optimal, conference, function, setting, depends, polytope, feasible, lemma, theorem, worst, learning, consider] [convergence, gradient, method, optimization, iteration, machine, pcg, variant, standard, dicg, polytopes, iterate, converges, iterates, acg, away, showed, faster, linearly] [linear, decomposition, also, can, denote, garber, certain, analysis, require, solution, gap, dimension, note, following, arg, first, explicit, running, one, suppose] [conditional, section, two, given, point, specific, important] [new, option, current, observation, time, beck] [memory, previous, using, international, structured, figure, performance, several, explicitly, use]
Mutual information for symmetric rank-one matrix estimation: A proof of the replica formula
jean barbier, Mohamad Dia, Nicolas Macris, Florent Krzakala, Thibault Lesieur, Lenka Zdeborová
jean barbier, Mohamad Dia, Nicolas Macris, Florent Krzakala, Thibault Lesieur, Lenka Zdeborová
[formula, possible, community, find, detection, potential, theory, threshold, clique] [proof, theorem, show, problem, function, algorithm, corollary, let, case, will, since, define, inequality, now, optimal, prove, consider] [large] [irs, amp, can, one, coupled, matrix, egood, phase, also, symmetric, mmse, gap, noise, minimum, sparse, algorithmic, global, ieee, first, small, ebad, note, computational, order, spectral, signal, interesting, wigner, spiked, analysis, side, vmmsen, local, second, explicit, replica, mmmsen] [point, statistical, fixed, stationary, theoretic, analytic, asymptotic, equal, mutual, corresponding, estimation, result, limit, gaussian, two] [information, model, system, range, transition, must] [using, region, three, hand, figure]
Variational Information Maximization for Feature Selection
Shuyang Gao, Greg Ver Steeg, Aram Galstyan
Shuyang Gao, Greg Ver Steeg, Aram Galstyan
[pairwise, number, graphical, subset, average, according, becomes, tree, scoring, independence] [lower, bound, learning, general, will, class, max, complexity, maximizing, maximization, greedy, algorithm, function, may] [variational, vmi, naive, method, machine, step, supplementary, mrmr, iblb, ilb, tractable, hln, term, rely, speccmi, xft, datasets, large, due] [can, assumption, also, existing, following, error, arg, decomposition, first, one, see, ieee] [selection, mutual, data, two, selected, based, estimating, bayes, conditional, distribution, journal, denotes] [information, model, forward, framework, time, demonstrate] [feature, using, shown, table, use, neural, proposed, filter, previous, figure, approach, three]
Observational-Interventional Priors for Dose-Response Learning
Ricardo Silva
Ricardo Silva
[causal, according, level, many, variable, confounding, intervention, sometimes] [outcome, learning, function, might, will, set, study, provide, problem] [posterior, method, bayesian, inference, variance, marginal, hyperparameters, supplementary, sampling] [can, one, analysis, also, synthetic] [observational, data, treatment, gaussian, interventional, distribution, given, mean, covariance, sample, curve, section, provides, competitor, covariates, estimating, true, estimate, confounders, sensitivity, nonparametric, common, two, distortion, affine, journal, difference, practical, andom, done, regime, dobs] [model, prior, process, uncertainty, information, simulated, adjustment, controlled, range, design] [figure, different, generated, computer, using, generate, learned, learn, per, used, use, shown, combined]
Optimal Sparse Linear Encoders and Sparse PCA
Malik Magdon-Ismail, Christos Boutsidis
Malik Magdon-Ismail, Christos Boutsidis
[number, whose] [algorithm, loss, theorem, optimal, bound, lemma, learning, give, set, minimizing, lower, let, every, maximizing, guarantee, problem, achieves, show, define, now, obtain, prove, since] [variance, batch, approximation, compute, log, machine, parameter, method, term, step] [sparse, linear, sparsity, matrix, iterative, can, pca, symmetric, principal, good, explained, component, error, first, singular, rank, column, encoders, one, eout, running, tpower, eir, also, analysis, much, generalized, power, second, minimum, relative] [data, construct, given, journal, theoretical, selection] [information, time] [encoder, using, use, used, reconstruction, neural, better, feature, residual, traditional, original]
Supervised Word Mover's Distance
Gao Huang, Chuan Guo, Matt J. Kusner, Yu Sun, Fei Sha, Kilian Q. Weinberger
Gao Huang, Chuan Guo, Matt J. Kusner, Yu Sun, Fei Sha, Kilian Q. Weinberger
[number, neighborhood, sentiment, unique, introduced] [learning, may, classic, algorithm, fast, optimal, set, problem, complexity, loss, lsi] [gradient, datasets, large, batch, compute, latent, efficient, approximation, dual, stochastic, university, optimization] [can, linear, error, matrix, computation, also, vector, initialization, generalized, first, relaxed] [word, distance, document, metric, wmd, embedding, transport, nca, bow, tij, wcd, section, neighbor, euclidean, described, two, news, nearest, knn, given, bbcsport, wasserstein, lda, space, twitter] [time, information, learns] [training, supervised, learn, use, text, work, using, representation, propose, similar, unsupervised, classification, recipe, table, neural, used, different, dataset, learned, proposed, labeled, semantic, shown]
Finding significant combinations of features in the presence of categorical covariates
Laetitia Papaxanthos, Felipe Llinares-Lopez, Dean Bodenham, Karsten Borgwardt
Laetitia Papaxanthos, Felipe Llinares-Lopez, Dean Bodenham, Karsten Borgwardt
[cmh, tar, facs, criterion, testability, association, number, subset, categorical, pruning, mining, itemset, attainable, covariate, combination, false, genetic, runtime, confounding, many, prunable, enumeration, threshold, resulting, contingency, testable, fwer, corrected] [algorithm, will, function, set, class, problem, binary, lemma, lower, let, combinatorial, precision, hypothesis, define] [line, method, datasets, efficient, log, large, apply, key] [can, one, minimum, condition, computational, power, first, main] [test, section, statistical, two, significance, based, covariates, testing, correction, conditional, data, associated, envelope, given] [significant, correct, potentially, search, allows, ability, value, current] [feature, discriminative, using, approach, novel, work, multiple, table, figure]
The Forget-me-not Process
Kieran Milan, Joel Veness, James Kirkpatrick, Michael Bowling, Anna Koop, Demis Hassabis
Kieran Milan, Joel Veness, James Kirkpatrick, Michael Bowling, Anna Koop, Demis Hassabis
[base, fmn, piecewise, partition, ptw, probabilistic, tree, number, probability, bag, repeating, made, weighting, level, mboc, many, fmnd, completed, obtained] [set, now, binary, will, learning, online, regret, complexity, algorithm, loss, class, defined, bound, best, define, conference, upper, bounded, notation, provide] [log, bayesian, averaging, performing, method] [can, first, denote, also, one, following, ieee, note] [data, given, segment, stationary, equation, provided, distribution, introduce, space, specific] [model, temporal, time, process, atari, information, game, within] [sequence, generating, task, using, used, figure, performance, source, prediction, previous, multiple, depth, perform, similar, generated, domain, digit, trained]
Average-case hardness of RIP certification
Tengyao Wang, Quentin Berthet, Yaniv Plan
Tengyao Wang, Quentin Berthet, Yaniv Plan
[graph, clique, definition, probability, number, detection, subgraph] [property, problem, hardness, lower, algorithm, known, satisfy, show, randomized, even, uniformly, define, will, complexity, satisfies, theorem, set, bound] [parameter, efficient, large, size] [certifier, rip, restricted, planted, matrix, isometry, computational, sparse, can, proposition, compressed, assumption, high, following, recovery, also, certification, signal, order, hard, ieee, note, one, sensing, linear, certifiers, sparsity, incoherence, submatrix, great, detecting] [random, distribution, regime, statistical, polynomial, asymptotic, result, based, given, computationally, testing, two, construction, section, specific, mathematical, sample, well] [design, time, write, whether] [dense, sequence, used, using, use, work]
Interaction Screening: Efficient and Sample-Optimal Learning of Ising Models
Marc Vuffray, Sidhant Misra, Andrey Lokhov, Michael Chertkov
Marc Vuffray, Sidhant Misra, Andrey Lokhov, Michael Chertkov
[ising, number, coupling, structure, probability, degree, node, graph, interaction, iso, rise, nmin, spin, xul, greater, graphical, los, national, reconstructs, according, strong] [theorem, lemma, algorithm, learning, function, case, set, least, bound, every, prove, consider, max, bounded, even, complexity, proof, logarithmic, convex] [objective, screening, respect, parameter, gradient, around] [condition, following, error, paper, regularized, can, zero, minimal, main, restricted, second, observed, also, order, perfect, ieee, one, penalty, vector] [maximum, estimator, exp, random, empirical, statistical, based, exponential, uniform, test, two] [model, intensity, information, value, observation, prior, choice] [reconstruction, using, neural, performance]
Interpretable Distribution Features with Maximum Testing Power
Wittawat Jitkrittum, Zoltán Szabó, Kacper P. Chwialkowski, Arthur Gretton
Wittawat Jitkrittum, Zoltán Szabó, Kacper P. Chwialkowski, Arthur Gretton
[probability, negative] [set, problem, function, bound, let, consider, lower, learning, hypothesis, class, since, bounded, theorem, may, uniformly] [parameter, optimization, large, size, full] [power, can, high, one, frequency, linear, error, see, following] [test, two, gaussian, scf, statistic, kernel, difference, sample, distribution, empirical, null, testing, positive, multivariate, mean, data, analytic, statistical, chwialkowski, given, witness, maximum, distinguishing, distance, toy, specified, gretton, dtr, gmd, journal, estimate, nonparametric, mmd] [maximize, characteristic, interpretable, model] [use, using, randomly, performance, learned, shown, table, perform, used, figure, facial, discriminative, spatial]
Efficient state-space modularization for planning: theory, behavioral and neural signatures
Daniel McNamee, Daniel M. Wolpert, Mate Lengyel
Daniel McNamee, Daniel M. Wolpert, Mate Lengyel
[theory, degree, number, resulting, thus, level] [optimal, compression, show, will, learning, algorithm, may, complexity, modular, define, function] [efficient, factor, processing, supplementary] [can, local, global, also, matrix, via, note, require] [length, data, entropy, based, given, empirical, corresponding, specific] [planning, state, modularization, module, modularized, start, centrality, soho, trajectory, activity, within, goal, information, across, lever, route, human, behavioral, entropic, hierarchical, markov, policy, simulated, transition, time, representational, firing, temporal, framework, artificial, process, navigation, search, environment, right, decision, action, stop, found] [neural, description, task, used, representation, use, using, compared, different, higher, approach]
Exploiting the Structure: Stochastic Gradient Methods Using Raw Clusters
Zeyuan Allen-Zhu, Yang Yuan, Karthik Sridharan
Zeyuan Allen-Zhu, Yang Yuan, Karthik Sridharan
[clustering, cluster, average, definition, probability, proportional] [algorithm, loss, convex, every, best, smaller, since, known, erm, consider, focus, learning, appendix, choose] [gradient, stochastic, clusteracdm, acdm, svrg, dual, coordinate, accelerated, method, clustersvrg, faster, haar, objective, descent, saga, covtype, machine, iteration, tong, step, approximate, large, due, proximal, apcg, auxiliary, datasets, epoch, ahclt] [can, one, running, vector, def, also, following, regularizer, much, see, regularized, comparison, optimum, first, small] [data, ridge, transformation, estimator, two, given, random, length, space] [time, raw, new, option, information] [using, use, training, dataset, feature, without, performance, improve, better, different, figure]
Global Analysis of Expectation Maximization for Mixtures of Two Gaussians
Ji Xu, Daniel J. Hsu, Arian Maleki
Ji Xu, Daniel J. Hsu, Arian Maleki
[many, either, according] [theorem, learning, may, algorithm, annual, expected, let, show, assume, corollary, lemma, maximization, known, provide, general, proof, conference] [convergence, iterates, parameter, likelihood, initial, log, converges, large, size, expectation, university, latent, royal, iteration] [global, following, denote, analysis, local, also, symposium, can, ieee, suppose, paper, certain, one, still, sufficiently, spectral] [population, mixture, stationary, sample, two, gaussian, point, statistical, fixed, distribution, result, dasgupta, given, balakrishnan, estimate, data, maximum, specific, well, mle, finite, true, chaudhuri, separation, theoretical, journal] [model, characterize, information, starting] [sequence, computer, arxiv, fully, preprint, using, work, performance]
Dual Space Gradient Descent for Online Learning
Trung Le, Tu Nguyen, Vu Nguyen, Dinh Phung
Trung Le, Tu Nguyen, Vu Nguyen, Dinh Phung
[number] [budget, online, learning, loss, function, mistake, budgeted, algorithm, strategy, rate, set, theorem, hinge, problem, best, achieve, assume] [dualsgd, size, maintenance, gradient, fogd, merging, descent, datasets, fth, dual, removed, hyperparameters, convergence, approximate, machine, provision, nogd, conduct, key, extensive] [can, computational, regression, vector, analysis, support, dimension, running, projection, following, xij, linear, high] [random, space, kernel, data, two, procedure, section, perceptron, address, denotes] [time, model, information, execution, effect, decision, whilst] [proposed, feature, classification, approach, performance, using, use, table, input, part, shown, work, different, dataset]
On Graph Reconstruction via Empirical Risk Minimization: Fast Learning Rates and Scalability
Guillaume Papa, Aurélien Bellet, Stephan Clémençon
Guillaume Papa, Aurélien Bellet, Stephan Clémençon
[graph, biau, bleakley, rule, probability, incomplete, average, preferential, theory, established, node, possible, thus, obtained, involving, edge, sum] [risk, learning, theorem, rate, problem, fast, set, minimization, class, complexity, consider, excess, minimizers, constant, may, inf, bound, minimizer, always] [sampling, supplementary, large, variance, material, var, additional] [can, order, analysis, one, error, computational, note, symmetric, also, related, remark, decomposition, following, stated] [empirical, statistical, random, distribution, based, given, section, distance, finite, conditional, numerical, estimate, independent, universal, bias, result, two] [form] [reconstruction, training, used, performance, prediction, representation, referred, approach, pair, similar]
Improving Variational Autoencoders with Inverse Autoregressive Flow
Diederik P. Kingma, Tim Salimans, Rafal Jozefowicz, Xi Chen, Xi Chen, Ilya Sutskever, Max Welling
Diederik P. Kingma, Tim Salimans, Rafal Jozefowicz, Xi Chen, Xi Chen, Ilya Sutskever, Max Welling
[den, variable, made, number] [learning, lower, context, will, conference, general, best, bound, since] [autoregressive, variational, inference, iaf, latent, posterior, log, flexible, method, vae, oord, stochastic, normalizing, masked, standard, approximate, parameterized, step, jacobian, improving, dkl, compute, elu, marginal, auxiliary, likelihood, sampling] [can, vector, still, see, diagonal, much, one, also, order] [transformation, gaussian, powerful, distribution, density, type, data, sample, computationally, nonlinear, mean] [inverse, model, simple, van, information, long] [flow, arxiv, preprint, deep, generative, neural, using, use, resnet, used, convolutional, pixelcnn, previous, figure, layer, natural, mnist, residual, single, international, autoencoders]
Towards Conceptual Compression
Karol Gregor, Frederic Besse, Danilo Jimenez Rezende, Ivo Danihelka, Daan Wierstra
Karol Gregor, Frederic Besse, Danilo Jimenez Rezende, Ivo Danihelka, Daan Wierstra
[level, number, variable, quality, structure, introduced, den, present] [compression, cost, learning, algorithm, will, might, show, conference, lower] [latent, draw, variational, variance, posterior, approximate] [can, one, first, high, also, low, global, see, good] [distribution, given, two, data] [information, model, prior, time, coding, abstract] [input, image, network, convolutional, figure, different, generative, layer, deep, pixel, amount, shown, higher, neural, per, lossy, compress, using, used, imagenet, omniglot, store, use, storing, conceptual, generated, recurrent, scale, generate, lzt, rnn, international, deepmind, representation, encode, stored, trained, original, google, produce, architecture]
Integrated perception with recurrent multi-task neural networks
Hakan Bilen, Andrea Vedaldi
Hakan Bilen, Andrea Vedaldi
[detection, number, possible] [learning, function, label, consider, class, max] [method, update, efficient] [can, one, first, det, also, iterative, order] [ordinary, data, integrated, corresponding, test, common, two, based, given, integrator, independent, well, space] [model, information, individual, box] [object, part, task, image, representation, multinet, shared, enc, different, neural, recurrent, classification, network, convolutional, used, output, architecture, multiple, using, feature, sharing, decoder, three, prediction, several, computer, initialized, improve, deep, train, propose, instead, encoder, training, input, back, spatial, layer, rcls, ground, pascal, img, last, perceptual, visual, vision, use]
Coin Betting and Parameter-Free Online Learning
Francesco Orabona, David Pal
Francesco Orabona, David Pal
[potential, number, definition, total] [algorithm, betting, olo, coin, regret, learning, let, wealth, online, optimal, bound, lea, strategy, will, wealtht, adaptive, convex, defined, endowment, satisfies, theorem, show, appendix, set, guarantee, loss, function, regrett, bet, kelly, rewardt, problem, every, excellent, prove, round, upper, best, absolute, rate, gambler, advice, lemma, just, define, proof, lower, assume, case] [initial, processing, optimization] [can, also, see, fraction, vector, linear, good, analysis, following] [hilbert, space, based, section, estimator, normal, two, construction] [reward, information, expert, new, design, time, prior] [sequence, used, neural, prediction, table, use, previous]
Near-Optimal Smoothing of Structured Conditional Probability Matrices
Moein Falahatgar, Mesrob I. Ohannessian, Alon Orlitsky
Moein Falahatgar, Mesrob I. Ohannessian, Alon Orlitsky
[probability, thus, number, many, probabilistic, fact] [smoothing, algorithm, risk, bound, since, theorem, learning, problem, context, pij, may, uniformly, give, even, complexity, optimal, minimax, lemma] [log, latent, naive, size, stochastic] [matrix, can, cij, one, rank, first, sampled, also, note, min, following, arg, bigram, penalized, qij, analysis, global, related, factorization, moothed, row, order, main] [empirical, data, conditional, section, estimator, sample, word, uniform, given, test, point, stationary, result] [model, framework, choice, simple, implicit] [language, performance, using, figure, proposed, work, preprint, arxiv, natural, structured, neural, instead, like, used, original]
Consistent Estimation of Functions of Data Missing Non-Monotonically and Not at Random
Ilya Shpitser
Ilya Shpitser
[missingness, graph, law, identification, ipw, independence, status, undirected, graphical, mnar, identified, variable, complete, causal, edge, pag, directed, represented, allow, fairly, lcpl] [every, lemma, set, function, may, known, confidence, implies, define, case, class, since, example, follows, general, problem, give, dependence] [full, parameter, inference, size, log] [missing, observed, analysis, following, can, factorization, also] [data, conditional, chain, random, estimator, consistent, section, distribution, underlying, true, estimation, sample, independent, functional, given, statistical, practical, joint, discrete, particular] [model, simple, form, complex, markov, inverse, mar] [using, approach, shown, used, work]
Algorithms and matching lower bounds for approximately-convex optimization
Andrej Risteski, Yuanzhi Li
Andrej Risteski, Yuanzhi Li
[probability, number, definition] [convex, function, algorithm, will, lower, diameter, bound, show, problem, minx, set, exists, every, consider, theorem, proof, max, since, lip, poly, lipschitz, define, upper, optimal, let, arbitrary, query, defined, case, even, give, want, lemma, idea, bandit, prove, outside, notion, obtain, implies, hope] [log, gradient, optimization, sampling, standard, factor] [can, one, following, high, need, small, approximately, also, algorithmic, order, min, linear, real, still] [point, polynomial, radius, theoretic, section, two, construction, result, positive, random, sample, construct, given] [ball, time, information, value, simple, extend, optimizing] [body, unit, use, natural, work, neural]
A Pseudo-Bayesian Algorithm for Robust PCA
Tae-Hyun Oh, Yasuyuki Matsushita, In Kweon, David Wipf
Tae-Hyun Oh, Yasuyuki Matsushita, In Kweon, David Wipf
[represents, basic, number, knowledge, represent, either, probabilistic] [ratio, algorithm, will, function, convex, problem, may, set, bound, cost, even, least, assume, case, minimization, equivalent] [bayesian, log, objective, optimization, method, although, standard, iid] [can, rank, matrix, outlier, rpca, ieee, via, sparse, pcp, support, recovery, principal, also, relative, quite, global, subspace, penalty, success, zgt, solving, solution, singular, norm, exact, proposition, first, existing, phase, nuclear, note, analysis] [robust, empirical, described, data, section, symmetry, given, true, estimation, practical] [prior, model, form, transition, new, design] [using, pattern, original, segmentation, performance, applied, used, motion, adopt, figure]
Verification Based Solution for Structured MAB Problems
Zohar S. Karnin
Zohar S. Karnin
[probability, identification, number, graphical, structure, many] [arm, best, algorithm, problem, query, bandit, complexity, set, dueling, expected, mab, verification, optimal, least, learning, hexplore, hverify, confidence, conference, unimodal, advice, condorcet, general, regret, exist, obtain, round, provide, findbestarm, setting, verifybestarm, focus, known, since, function, let, yet, elaborate] [machine, log, stochastic, parameter, additional, pure] [min, can, one, linear, vector, following, sufficiently, high, first, indeed, assumption] [given, independent, section, associated, corresponding, two, based, random, result, provides] [exploration, reward, value, framework, failure, rather, information, extended] [candidate, using, identity, pair, international, several, different, output, consists, input]
Bayesian Optimization with Robust Bayesian Neural Networks
Jost Tobias Springenberg, Aaron Klein, Stefan Falkner, Frank Hutter
Jost Tobias Springenberg, Aaron Klein, Stefan Falkner, Frank Hutter
[number, many, thus, probabilistic, obtained] [function, learning, set, algorithm, will] [bayesian, optimization, gradient, stochastic, bohamiann, monte, sghmc, dngo, ddpg, hyperparameters, method, carlo, mcmc, large, parallel, machine, supplementary, additional, evaluation, standard, posterior, scalable, scalability, parameter, fit] [can, matrix, via, noise, following, related, regression, good] [hamiltonian, based, estimate, well, equation, mean, robust, hmc, given, section, sample, robustness, distribution, data, random, embedding, two] [model, hyperparameter, uncertainty, found, required, reinforcement, optimized, acquisition, reward] [neural, deep, used, using, performance, network, scale, use, different, figure, residual, performed, proposed, dataset, adaptation]
The Multi-fidelity Multi-armed Bandit
Kirthevasan Kandasamy, Gautam Dasarathy, Barnabas Poczos, Jeff Schneider
Kirthevasan Kandasamy, Gautam Dasarathy, Barnabas Poczos, Jeff Schneider
[number, many, mth, probability] [fidelity, arm, will, bound, regret, lower, ucb, bandit, algorithm, strategy, set, consider, play, upper, confidence, setting, since, cost, optimal, might, learning, highest, online, problem, capital, expected, playing, may, played, theorem, cheaper, advertising, let, suboptimal, eliminate, bernoulli, lemma, smaller, ignore, notion, obtain, displaying, analyse] [due, expensive, step, approximate] [can, following, high, first, one, small, assumption, denote, much, low, note, analysis, also] [mean, given, two] [time, reward, design, goal, information, within, choice] [using, work, performance, higher, use, better]
Mapping Estimation for Discrete Optimal Transport
Michaël Perrot, Nicolas Courty, Rémi Flamary, Amaury Habrard
Michaël Perrot, Nicolas Courty, Rémi Flamary, Amaury Habrard
[coupling, probability, probabilistic, quality] [problem, learning, optimal, set, function, convex, consider, example, algorithm, let, case, define, choose, relaxation, cost, theorem, obtain] [method, gradient, optimization, supplementary, approximation, solve, term, technique, respect] [can, one, linear, regularization, regularized, first, formulation, solution, good, matrix, paper, min, following, arg, also, computation] [transport, transformation, barycentric, two, theoretical, section, kernel, empirical, based, space, given, corresponding, wasserstein, discrete, true, way, fixed] [target, new, corresponds, framework] [source, image, domain, using, map, mapping, adaptation, propose, learned, approach, jointly, used, use, learn, several, proposed, different, seen, training, color, dataset, original]
Dual Decomposed Learning with Factorwise Oracle for Structural SVM of Large Output Domain
Ian En-Hsu Yen, Xiangru Huang, Kai Zhong, Ruohan Zhang, Pradeep K. Ravikumar, Inderjit S. Dhillon
Ian En-Hsu Yen, Xiangru Huang, Kai Zhong, Ruohan Zhang, Pradeep K. Ravikumar, Inderjit S. Dhillon
[structural, number, grows, labeling, many, find, graph, according] [algorithm, active, maximization, greedy, loss, learning, problem, complexity, set, oracle, max, since, constant, show, qmax, upper, binary, case] [factor, large, mjf, objective, method, dual, convergence, bcfw, step, size, factorwise, fmo, iteration, inference, gdmm, ssg, machine, approximate, requires, approximation, chineseocr, faster, augmented, descent, primal, sampling, efficient, processing, optimization] [can, one, via, linear, also, min, note, much, error, computational, arg, multiplier, solving, sublinear, support, analysis] [two, given, test, type] [time, direction, information] [structured, output, domain, training, prediction, figure, map, svms, approach, dataset]
Bayesian Optimization for Probabilistic Programs
Tom Rainforth, Tuan-Anh Le, Jan-Willem van de Meent, Michael A. Osborne, Frank Wood
Tom Rainforth, Tuan-Anh Le, Jan-Willem van de Meent, Michael A. Osborne, Frank Wood
[program, number, probabilistic, contains] [function, problem, query, unknown, surrogate, consider, observe, show, setting] [optimization, method, likelihood, inference, line, importance, sampling, additional] [error, first, comparison, optimum, note] [mean, maximum, corresponding, estimate, based, section, provides, transformation, distribution] [bopp, acquisition, demonstrate, target, potentially, optimizing, significant, implicit, take, adapting, spearmint, simply, prior, smac, automatically, tpe, unbounded, scaling, must, new, optimize, evaluated, assigns, alternative, transformed, metropolis, directly, hastings, therefore, giving, form, identical, prominent, statement, continuous, corresponds, along, previously, solid, subject, next, annealed] [using, figure, used, shown, approach, region, optimizer, top, fully]
A Non-generative Framework and Convex Relaxations for Unsupervised Learning
Elad Hazan, Tengyu Ma
Elad Hazan, Tengyu Ma
[improper, definition, theory, probability, average] [learning, hypothesis, class, convex, set, algorithm, theorem, give, complexity, loss, let, rademacher, consider, close, compression, function, define, general, since, will, show, problem, relaxation, proof, constant, unknown, pac, exists, appendix, known, best] [efficient, respect, sampling, optimization] [can, spectral, error, dictionary, pca, decoding, linear, vmax, much, norm, min, following, main, arg, sparse, group, algebraic, matrix, though, component, vector, one, analysis, suppose, condition] [data, length, based, given, euclidean, distribution, kernel, bias, polynomial, section, random, space, generalization] [encoding, framework, information, new, allows] [unsupervised, reconstruction, generative, approach, without, learn, previous, using, neural, learned, domain]
Value Iteration Networks
Aviv Tamar, Sergey Levine, Pieter Abbeel, YI WU, Garrett Thomas
Aviv Tamar, Sergey Levine, Pieter Abbeel, YI WU, Garrett Thomas
[graph] [learning, optimal, function, algorithm, show, rate, general, defined, may, loss, differentiable] [standard, iteration, additional, gradient, solve] [can, computation, also, success, note, vector] [based, section, random, test, true, particular, important] [policy, planning, vin, reactive, value, state, reward, goal, module, vins, continuous, plan, observation, mdp, reinforcement, model, action, obstacle, control, search, form, design, information, new, elevation, decision, within, trajectory] [network, image, trained, learn, using, domain, neural, training, cnn, layer, better, different, deep, task, figure, map, approach, similar, convolution, attention, mapping, performance, fully, convolutional, several, prediction, generalize, supervised]
Graph Clustering: Block-models and model free results
Yali Wan, Marina Meila
Yali Wan, Marina Meila
[clustering, pfm, graph, sbm, cluster, weighted, stable, node, indicator, community, political, block, ckk, detection, compatible, marina, lfr, edge] [theorem, let, will, bound, obtain, satisfy, set, since, whenever, proof, assume, show, algorithm, case, close, prove] [stochastic, sampling, large, processing] [matrix, assumption, spectral, can, also, proposition, recovery, perturbation, stability, error, existing, note, main, maxk, much, paper, frobenius, small] [data, result, well, construct, two, laplacian, distance, goodness, measure, based, given] [model, framework, information] [used, work, using, volume, neural, network]
Efficient Globally Convergent Stochastic Optimization for Canonical Correlation Analysis
Weiran Wang, Jialei Wang, Dan Garber, Dan Garber, Nati Srebro
Weiran Wang, Jialei Wang, Dan Garber, Dan Garber, Nati Srebro
[number, loop] [algorithm, least, complexity, obtain, appendix, theorem, show, set, problem, max, proof, dependence] [stochastic, log, svrg, appgrad, convergence, normalization, ccalin, gradient, optimization, cca, approximate, iterates, suboptimality, preconditioning, rdx, convergent, sgd, solve, step, batch, asvrg, objective, converges, variance, solved] [min, singular, phase, can, analysis, solution, condition, power, matrix, alternating, regularization, following, gap, linear, rdy, note, initialization, globally, solving, high] [canonical, two] [time, correlation, value, form] [using, use, training, work]
Deep Learning without Poor Local Minima
Kenji Kawaguchi
Kenji Kawaguchi
[conjecture, negative, number, theory, expression] [loss, theorem, function, learning, corollary, case, lemma, proof, let, every, assume, prove, set, conference, consider, show, general, problem, even, now, obtain] [saddle, optimization] [local, linear, can, global, hessian, critical, minimum, necessary, following, condition, matrix, semidefinite, choromanska, assumption, rdy, baldi, arbitrarily, unrealistic, conclude, entry, also, proposition, sketch, seven, note, one, noted, perturbation, first] [nonlinear, point, theoretical, section, corresponding, positive, practical, random] [model, information, write, value, statement, poor] [deep, neural, open, layer, previous, output, training, without, use, used, work, deeper, shallow]
Privacy Odometers and Filters: Pay-as-you-Go Composition
Ryan M. Rogers, Salil Vadhan, Aaron Roth, Jonathan Ullman
Ryan M. Rogers, Salil Vadhan, Aaron Roth, Jonathan Ullman
[valid, probability, definition, neighboring, basic, number, analyst] [privacy, composition, theorem, algorithm, loss, bound, adaptive, odometer, round, will, private, chosen, setting, give, randomized, differentially, define, show, adaptively, adversary, case, every, may, even, let, function, realized, advanced, prove, halt, satisfies, now, martingale, set, stopping, worse, upper, coin] [parameter, log, run, select, datasets] [can, note, following, also, one, high, view, global, first, computation, see] [differential, data, two, fixed, random, statistical, asymptotic, result, useful] [choice, time, must, next, response, future, allows] [filter, output, previous, sequence, use, used, without, similar, input]
On Regularizing Rademacher Observation Losses
Richard Nock
Richard Nock
[equivalence, number, definition, iff, fact] [oost, rado, loss, learning, rados, equivalent, learner, example, algorithm, theorem, slope, let, proportionate, set, max, may, show, now, rademacher, regularizing, fix, since, even, maxj, obtain, minimization, defined, provide, nock, best] [boosting, step, efficient, logistic, fit, log, large, key, size, depend, end, boost, optimization, datasets, update] [regularized, can, regularization, linear, one, suppose, sparsity, sufficient, via, following, also, first, significantly, order, regularizer, accurate, popular, good] [two, exponential, exp, data, given, corresponding] [take] [weak, classifier, using, table, feature, supervised, domain, shown, different, training]
Reward Augmented Maximum Likelihood for Neural Structured Prediction
Mohammad Norouzi, Samy Bengio, zhifeng Chen, Navdeep Jaitly, Mike Schuster, Yonghui Wu, Dale Schuurmans
Mohammad Norouzi, Samy Bengio, zhifeng Chen, Navdeep Jaitly, Mike Schuster, Yonghui Wu, Dale Schuurmans
[edit, according, negative, probability, number, average, probabilistic] [learning, expected, function, consider, best, make, set] [sampling, likelihood, machine, dkl, gradient, optimization, inference, augmented, divergence, large, objective, log, variance, standard, approximate] [can, one, regularized, also, first, note] [maximum, given, conditional, distance, distribution, difference, two, test, sample, entropy, length] [reward, model, policy, exponentiated, rml, reinforcement, lrl, lrml, payoff, baseline, temperature, optimizing, framework, alternative, optimize, form] [output, training, neural, task, use, sequence, using, prediction, work, translation, bleu, structured, speech, different, approach, ground, truth, recognition, score, recurrent, table, several, deep, achieved, network, improves, supervision, augmentation, category]
The Sound of APALM Clapping: Faster Nonsmooth Nonconvex Optimization with Stochastic Asynchronous PALM
Damek Davis, Brent Edmunds, Madeleine Udell
Damek Davis, Brent Edmunds, Madeleine Udell
[block, number, wjk, cache, linearized, many] [algorithm, problem, convex, function, let, theorem, will, class, provide, rate, expected, might, minimize, known, assume] [asynchronous, stochastic, sapalm, convergence, coordinate, gradient, nonconvex, parallel, optimization, method, nonsmooth, synchronous, proximal, iteration, distributed, large, speedup, firm, compute, counter, palm, size, xkj, lyapunov, objective, converges, iterates, update, stepsize, processing] [can, noise, assumption, linear, following, first, pca, processor, one, matrix, denote, low, sparse, global, see, alternating] [two] [time, read, model, delayed, write, information] [arxiv, preprint, using, use, sequence, neural, work, last, different]
Fast Mixing Markov Chains for Strongly Rayleigh Measures, DPPs, and Constrained Sampling
Chengtao Li, Suvrit Sra, Stefanie Jegelka
Chengtao Li, Suvrit Sra, Stefanie Jegelka
[probability, path, rayleigh, partition, constraint, graph, exchange, probabilistic, total, shortest, ising, include, many] [set, fast, matroid, algorithm, bound, consider, since, show, let, may, general, appendix, theorem, special, observe, max, combinatorial, will, uniformly, learning, dependence, bounded, cardinality] [sampling, strongly, inference, convergence, end, log, determinantal, size, factor, efficient, dpp, constrained, machine] [can, one, first, also, certain, hence] [mixing, chain, uniform, point, distribution, psrf, sampler, sample, discrete, result, mix, two, random, journal, fixed, theoretical, construct, empirically, length] [time, markov, transition, simple] [flow, use, iter, ground, using, shown]
A scaled Bregman theorem with applications
Richard Nock, Aditya Menon, Cheng Soon Ong
Richard Nock, Aditya Menon, Cheng Soon Ong
[clustering, fact, flat, potential] [theorem, convex, lemma, ratio, learning, may, problem, since, online, appendix, loss, adaptive, bound, multiclass, function, binary, let, case, general, improvement, beat, now, differentiable] [divergence, log, reduction, approximation, step, dual, mirror, university, optimisation, update] [can, one, norm, lifting, vector, related, signal, suitable, ieee, matrix, via] [bregman, density, estimation, scaled, skm, manifold, sphere, two, drec, curved, given, hyperboloid, seeding, estimate, euclidean, exponential, geodesic, normalisation, distance, spherical, section] [new, significant] [table, perspective, map, using, three, use, figure, used, different]
InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets
Xi Chen, Xi Chen, Yan Duan, Rein Houthooft, John Schulman, Ilya Sutskever, Pieter Abbeel
Xi Chen, Xi Chen, Yan Duan, Rein Houthooft, John Schulman, Ilya Sutskever, Pieter Abbeel
[categorical, variable, represent, present, regular] [learning, even, show, lower, bound, will, algorithm, maximizing, minimax, set] [latent, variational, posterior, method, auxiliary, easy] [can, one, noise, also, though] [mutual, distribution, data, two, random, discrete, true] [information, code, continuous, model, interpretable, learns, azimuth, research, prior, varying, simple, goal] [infogan, learn, generative, representation, disentangled, adversarial, generator, unsupervised, variation, face, digit, using, figure, use, network, gan, learned, neural, deep, able, preprint, pose, lighting, arxiv, different, generated, mnist, supervised, training, disentangle, dataset, explicitly, semantic, manipulating, work, identity, evaluate, visual, propose]
High Dimensional Structured Superposition Models
Qilong Gu, Arindam Banerjee
Qilong Gu, Arindam Banerjee
[sum, probability, number, structural, structure, interaction, present] [bound, let, theorem, general, lower, will, lemma, show, set, give, upper, complexity, satisfied, implies, appendix, inf, choose, constant, proof, convex, depends, consider, hold, defined, case, provide] [parameter, log, deterministic] [condition, error, can, matrix, component, following, superposition, width, high, norm, sparse, vector, recovery, analysis, coherence, noise, componentwise, also, one, cone, maxi, eigenvalue, dimension, suitable, characterization, restricted, linear, denote] [random, gaussian, given, estimation, section, sample, estimator, geometric, independent, two, based, result, statistical, geometry] [design, information, form, start, observation] [different, using, figure, structured, use]
Clustering Signed Networks with the Geometric Mean of Laplacians
Pedro Mercado, Francesco Tudisco, Matthias Hein
Pedro Mercado, Francesco Tudisco, Matthias Hein
[clustering, signed, laplacians, negative, graph, lgm, lbn, lsn, pin, arithmetic, normalized, sym, block, ordering, presence, number, qsym, present, cluster, lam, informative, lsr, definition, obtained] [smallest, let, algorithm, best, show, consider, observe, expected, defined] [method, stochastic, expectation, large, compute, technique, size] [matrix, eigenvectors, can, spectral, pout, solution, small, analysis, computation, linear, error, existing, one, krylov, condition, related, first, eigenvalue, subspace, eigenvector, sparse, fraction, note] [mean, positive, geometric, laplacian, section, corresponding, two, based, definite, given] [model, whereas, extended, next] [use, proposed, network, figure, ground, table, truth, different]
Confusions over Time: An Interpretable Bayesian Model to Characterize Trends in Decision Making
Himabindu Lakkaraju, Jure Leskovec
Himabindu Lakkaraju, Jure Leskovec
[cluster, confusion, cot, maker, asthma, indicator, purity, insurance, bail, diagnostic, made, evaluating, mth, eth, present, obtained, possible] [label, learning, set, element, will, provide] [latent, bayesian, inference, sampling] [item, can, subspace, also, sampled, vector, matrix, one, following, row, error] [true, prototype, corresponding, denotes, distribution, associated, important, data, dirichlet, metric, discrete, provides, gender, estimated, multinomial] [decision, model, time, making, value, prior, interpretable, modeling, framework, individual, process, collective, temporal, inverse, human, research] [feature, predicting, using, figure, performance, table, single, evaluate, generative, use]
Learning Additive Exponential Family Graphical Models via \ell_{2,1}-norm Regularized M-Estimation
Xiaotong Yuan, Ping Li, Tong Zhang, Qingshan Liu, Guangcan Liu
Xiaotong Yuan, Ping Li, Tong Zhang, Qingshan Liu, Guangcan Liu
[graphical, pairwise, structure, graph, among, probability, node] [learning, consider, assume, unknown, price, set, theorem, function, will] [sampling, parameter, convergence, approximation, size, likelihood, stock, approximate, method, gradient, log] [following, regularized, sufficient, assumption, can, high, recovery, error, analysis, sparse, order, via, computational, condition, restricted] [adefgm, estimator, estimation, joint, mle, exponential, data, nonparanormal, additive, random, exp, statistical, distribution, family, estimate, conditional, two, ugms, basis, ggm, dimensional, underlying, associated, sample, true, journal, gaussian, fst, result, annals, given, maximum] [model, information, complex] [proposed, using, learn, propose, performance]
Phased Exploration with Greedy Exploitation in Stochastic Combinatorial Partial Monitoring Games
Sougata Chaudhuri, Ambuj Tewari
Sougata Chaudhuri, Ambuj Tewari
[unique, partial, number, according] [algorithm, regret, learner, optimal, bound, set, gcb, online, problem, function, expected, pege, dependent, will, adversary, ranked, combinatorial, existence, exponentially, relevance, constant, defined, might, bandit, exploitation, since, oracle, even, theorem, achieve, dependence, let, get, confidence, conference, classic, strategy] [log, size, end, large, full, stochastic, requires] [can, assumption, global, gap, cpm, ranking, linear, second, one, paper, first, note, monitoring, user, vector, also, small] [distribution, observable, space, finite, estimation, independent, estimate, transformation, fixed, infinite, two, practical] [action, feedback, reward, exploration, model, framework, game, time, continuous, move, information, within, rmax] [top, work]
Achieving the KS threshold in the general stochastic block model with linearized acyclic belief propagation
Emmanuel Abbe, Colin Sandon
Emmanuel Abbe, Colin Sandon
[community, probability, detection, graph, threshold, vertex, number, conjecture, nonbacktracking, abp, propagation, block, definition, adjacent, average, achieving, enough, degree, likely, assign, expect, sum, achievability, detect, cycle, presence, linearized, subtract] [algorithm, let, will, set, general, prove, proof, drawn, define, version, return, exists, least] [stochastic, large, solves, step, compute, initial] [one, matrix, also, approximately, can, spectral, symmetric, eigenvector, order, following, vector, eigenvalue, sparse, high, dominant, first] [length, two, random, equal, distribution, described, bias, positive, way, based] [model, belief, value, simply, snr, new] [different, instead, approach, part, use, multiple]
Simple and Efficient Weighted Minwise Hashing
Anshumali Shrivastava
Anshumali Shrivastava
[number, weighted, average, many, consistency] [scheme, hashing, algorithm, hash, fast, will, theorem, return, binary, known, show, idea, constant, define, cost] [minwise, sampling, unweighted, requires, ioffe, end, faster, datasets, method, around, compute, unbiased, log, size, computing, green, oxford, approximation, costly] [can, vector, real, exact, one, existing, computation, sparsity, also, error, component, first, note, need, running] [random, data, given, independent, procedure, two, permutation, based, biased, associated, consistent, equation] [time, required, value, simple, range] [using, similarity, accuracy, proposed, generate, used, different, input, work, figure, vision, table, red, effective, computer, per, dataset]
Learning Sparse Gaussian Graphical Models with Overlapping Blocks
Seyed Mohammad Javad Hosseini, Su-In Lee
Seyed Mohammad Javad Hosseini, Su-In Lee
[grab, graphical, block, gene, overlapping, number, constraint, cancer, structure, clustering, expression, variable, average, assignment, connected, aml, ith, assigned, many, belong, present, called, densely] [learning, algorithm, problem, show, set, convex, known, consider, lemma, element] [method, parameter, log, optimization, standard, size, dual, coordinate, descent] [matrix, can, regularization, lasso, following, sparse, solution, first, one, sparsity, synthetic, group, det, significantly, existing, need, diagonal] [given, data, estimate, gaussian, based, covariance, two, associated, random, denotes, estimation, journal] [prior, maximize, subject, fig, model] [network, use, learn, novel, overlap, used, jointly, similar, learned, work, figure, multiple]
FPNN: Field Probing Neural Networks for 3D Data
Yangyan Li, Soeren Pirk, Hao Su, Charles R. Qi, Leonidas J. Guibas
Yangyan Li, Soeren Pirk, Hao Su, Charles R. Qi, Leonidas J. Guibas
[connected, represented, informative, thus, number, many] [learning, since, set, complexity, even] [batch, size, designed, michael, efficient] [can, note, one, also, computation, computational] [data, distance, point, two, gaussian, normal, random, important, associated, testing, well, space] [range, long, grid, directly, sensor, occupancy, advantage, information] [probing, field, input, shape, layer, fpnn, performance, object, convolutional, table, neural, fully, filter, figure, fpnns, resolution, deep, classification, cnns, used, output, task, feature, different, higher, representation, network, training, transform, understanding, dotproduct, approach, dataset, image, evaluate, voxel, trained, volumetric, recognition, cnn]
Homotopy Smoothing for Non-Smooth Problems with Lower Complexity than O(1/\epsilon)
Yi Xu, Yan Yan, Qihang Lin, Tianbao Yang
Yi Xu, Yan Yan, Qihang Lin, Tianbao Yang
[number, present, developed, strong] [smoothing, function, algorithm, bound, convex, complexity, homotopy, set, problem, let, optimal, considered, max, smoothed, constant, property, loss, lower, minimization, finding, will, learning, show, theorem, inequality, rate, make, logb, implies] [iteration, convergence, optimization, parameter, proximal, gradient, method, smooth, objective, siam, faster, accelerated, dual, term, large, strongly, machine, approximation, university] [error, can, local, solving, linear, solution, min, denote, condition, cone, assumption, also, sparse, norm, one, analysis, see, following, running, accurate, first] [result, family, given, point, two, smoothness] [sharpness, value] [proposed, work, different, relatively, yang, style]
DISCO Nets : DISsimilarity COefficients Networks
Diane Bouchacourt, Pawan K. Mudigonda, Sebastian Nowozin
Diane Bouchacourt, Pawan K. Mudigonda, Sebastian Nowozin
[probabilistic, scoring, coefficient, rule, strictly, represent] [loss, function, proper, example, best, learning, pointwise, diversity, consider, defined] [objective, supplementary, evaluation, posterior] [noise, can, require, one, order, first] [distribution, given, data, two, true, based, estimate, test, equation, sample, kernel, euclidean, random, conditional, mean, section] [model, uncertainty, value] [disco, pose, depth, training, hand, image, input, use, adversarial, convolutional, generative, output, dissimilarity, figure, deep, prediction, cgan, candidate, different, neural, using, gan, table, single, architecture, network, dataset, train, oberweger, fed, gneiting, referred, several, mirza, presented, used, similar, employed, dense, task]
Protein contact prediction from amino acid co-evolution using convolutional networks for graph-valued images
Vladimir Golkov, Marcin J. Skwark, Antonij Golkov, Alexey Dosovitskiy, Thomas Brox, Jens Meiler, Daniel Cremers
Vladimir Golkov, Marcin J. Skwark, Antonij Golkov, Alexey Dosovitskiy, Thomas Brox, Jens Meiler, Daniel Cremers
[structure, coupling, many, number, present] [function, learning, set, problem, conference, considered] [contact, protein, amino, method, plmconv, acid, metapsicov, evolutionary, plmdca, homologs, size, residue, inference, pressure, additional, homologous, tertiary, potts] [can, one, order, also, accurate, computational, matrix, regularization, related] [data, based, positive, way, length] [value, information, predictive, new, potentially] [prediction, sequence, network, convolutional, input, neural, different, map, position, multiple, using, window, used, predicted, performance, training, predict, feature, proposed, layer, use, able, deep, higher, alignment, figure, table, propose]
Fast Distributed Submodular Cover: Public-Private Data Summarization
Baharan Mirzasoleiman, Morteza Zadimoghaddam, Amin Karbasi
Baharan Mirzasoleiman, Morteza Zadimoghaddam, Amin Karbasi
[number, find, summary, many, centralized, subset, social, threshold, spark] [set, fast, submodular, algorithm, greedy, private, problem, cover, summarization, personalized, utility, function, smaller, fastcover, public, covering, least, dominating, friendster, massive, define, discover, coverage, mapreduce, even, smallest, finding, return, round, instance] [size, distributed, machine, marginal, large, central, run, factor, parameter, objective, gps] [solution, can, user, recommendation, one, movie, also, high, running, note, good, small, recommender] [data, provides, maximum] [value, location, information, time, desired, truly, gain] [performance, dataset, used, single, use, using, network, including, consists, similarity]
Tractable Operations for Arithmetic Circuits of Probabilistic Models
Yujia Shen, Arthur Choi, Adnan Darwiche
Yujia Shen, Arthur Choi, Adnan Darwiche
[psdd, probabilistic, circuit, arithmetic, psdds, vtree, variable, compilation, graphical, multiplication, polytime, instantiation, compiling, sdd, definition, decomposable, multiply, probability, sentential, summing, represent, iff, compile, node, darwiche, choi, called, boolean, obtained, compiled, whose, number, diagram] [learning, algorithm, theorem, show, every, class, will, known, consider] [factor, inference, size, tractable, bayesian, deterministic] [can, also, one, see, product, following, support, first, algebraic] [two, distribution, based, given, section] [decision, time, operation, markov, model, belief, tabular] [figure, using, input, used, reasoning, representation, network, including, available, proposed]
Variational Autoencoder for Deep Learning of Images, Labels and Captions
Yunchen Pu, Zhe Gan, Ricardo Henao, Xin Yuan, Chunyuan Li, Andrew Stevens, Lawrence Carin
Yunchen Pu, Zhe Gan, Ricardo Henao, Xin Yuan, Chunyuan Li, Andrew Stevens, Lawrence Carin
[number, developed] [learning, set, svm, element, lower, consider] [variational, latent, inference, bayesian, method, caltech, stochastic, posterior, sampling, datasets, vae, parameter, approximate] [error, also, vector, can, tensor, first] [test, distribution, joint, based, associated, proportion, provided, data, gibbs, section, two] [model, time, new, code] [image, cnn, labeled, training, generative, deep, used, caption, imagenet, neural, classification, using, pooling, dgdn, table, layer, encoder, use, recognition, decoder, trained, network, convolutional, unsupervised, validation, captioning, deconvolutional, unpooling, googlenet, performance, mcem, activation, novel, better, accuracy, figure, single, employed, feature, unlabeled, available, map]
Convergence guarantees for kernel-based quadrature rules in misspecified settings
Motonobu Kanagawa, Bharath K. Sriperumbudur, Kenji Fukumizu
Motonobu Kanagawa, Bharath K. Sriperumbudur, Kenji Fukumizu
[weighted, degree, lattice, knowledge, probability, describe] [rate, let, case, theorem, worst, setting, optimal, general, achieve, assumed, even, consider, defined, function, satisfies, known, adaptive, assume, provide, show, will] [convergence, smooth, carlo, deterministic, bayesian, monte] [assumption, can, error, order, one, also, see, certain, power, note, small, following, product, analysis] [kernel, integrand, space, korobov, section, quadrature, rkhs, distribution, wkor, smoothness, numerical, sobolev, random, uniform, given, qmc, bach, reproducing, result, rkhss, integral, construction, belongs, theoretical, mpn, denotes, statistical, important, hilbert, integrands] [integration, form, misspecified, therefore, state] [use]
Structure-Blind Signal Recovery
Dmitry Ostrovsky, Zaid Harchaoui, Anatoli Juditsky, Arkadi S. Nemirovski
Dmitry Ostrovsky, Zaid Harchaoui, Anatoli Juditsky, Arkadi S. Nemirovski
[structure, sum, present, constraint] [oracle, let, problem, set, theorem, consider, risk, appendix, price, may, exists, convex, bound, unknown, least, known, assume, adaptive, implies, close, function, pointwise, interval, minimax, satisfies] [constrained, optimization, computed, parameter, operator] [signal, can, recovery, linear, one, assumption, lasso, small, penalized, subspace, norm, dimension, noise, recover, see, recovering, harmonic, note, error, still, suppose, high, solution] [estimator, given, fourier, equation, polynomial, estimation, nonparametric, gaussian, difference, random, length, numerical, corresponding, family, theoretical, specified] [left, simple, right] [filter, denoising, use, using, image, proposed, adaptation, approach, transform, hidden]
Collaborative Recurrent Autoencoder: Recommend while Learning to Fill in the Blanks
Hao Wang, Xingjian SHI, Dit-Yan Yeung
Hao Wang, Xingjian SHI, Dit-Yan Yeung
[ctr, probabilistic, average] [learning, will, set, scheme, since, provide, show] [bayesian, rrn, datasets, supplementary, hyperparameters, boost, hybrid] [can, collaborative, recommendation, user, note, matrix, vector, also, recommender, one, rating, first, filtering, see] [section, two, word, distribution, equation, based, robust, topic, denotes] [model, information, modeling, implicit, directly] [crae, recurrent, use, denoising, deep, sequence, generation, cdl, used, different, wildcard, autoencoder, content, citeulike, pooling, input, ikw, training, using, representation, like, figure, performance, neural, svdfeature, table, rnn, score, deepmusic, bleu, generative, outperform, propose, jointly, able, recall, network, output, shown, gate, task]
Learning the Number of Neurons in Deep Networks
Jose M. Alvarez, Mathieu Salzmann
Jose M. Alvarez, Mathieu Salzmann
[number, thus, total, obtained, removing, influence] [learning, set, loss, general, algorithm, remaining] [initial, reduction, size, parameter, method, constructive, batch, gradient, proximal, operator, large, additional] [group, can, also, regularizer, note, sparsity, first, relative, gap, percentage, one, linear, sparse] [selection, test, two, generalization, reduces] [model, automatically, individual, effectively, saving, determine, behavior] [deep, network, approach, layer, using, memory, original, accuracy, convolutional, training, neural, used, bnetc, consists, architecture, three, compact, per, imagenet, figure, different, param, table, effective, trained, recognition, icdar, last, overcomplete, single, similar, shallow]
Online ICA: Understanding Global Dynamics of Nonconvex Optimization via Diffusion Processes
Chris Junchi Li, Zhaoran Wang, Han Liu
Chris Junchi Li, Zhaoran Wang, Han Liu
[diffusion, theory, number, negative, neighborhood] [algorithm, online, let, learning, consider, case, theorem, problem, implies, weakly, upper] [sgd, stochastic, nonconvex, convergence, optimization, gradient, ode, approximation, converges, log, processing, method, iteration, descent, initial, around, objective, iterates, appropriate, especially, sde] [local, tensor, phase, analysis, solution, unstable, global, can, following, via, sparse, matrix, desirable, first, component, conclude, second, wkk, solving, escaping, alternating, proposition, noise, escape] [statistical, exp, differential, based, stationary, section, equation, independent, random] [markov, information, process, time, within, characterize, towards, system, value] [arxiv, preprint, neural, weak, different, using, three, figure, spatial, understanding]
Coupled Generative Adversarial Networks
Ming-Yu Liu, Oncel Tuzel
Ming-Yu Liu, Oncel Tuzel
[number, constraint, contains, edge, attribute] [learning, drawn, let, function, set, problem] [marginal, showed] [can, also, one, note, via, first] [joint, distribution, corresponding, two, conditional, given, transformation, usps] [model, share, decode, framework] [cogan, image, generative, different, discriminative, training, used, domain, deep, depth, color, figure, without, generation, adversarial, pair, face, dataset, generated, network, performance, input, learn, digit, neural, task, using, trained, achieved, gan, unsupervised, several, last, mnist, correspondence, pixel, applied, work, convolutional, sharing, adaptation, semantics, including, learned, randomly, agreement, rgbd, generate, gans]
Incremental Variational Sparse Gaussian Process Regression
Ching-An Cheng, Byron Boots
Ching-An Cheng, Byron Boots
[probability, number, structure] [learning, function, max, problem, online, set, algorithm, general, conference, equivalent] [variational, inducing, stochastic, gpr, posterior, inference, gradient, ascent, approximate, log, hyperparameters, sarcos, mirror, update, approximation, incremental, full, ivsgprada, ivsgpr, batch, machine, subproblem, solved, solve, parametrization, datasets, large, divergence, processing, size, objective, performing, solves, step, dual, vsgprsvi] [sparse, can, regression, subspace, error, first] [gaussian, kernel, covariance, denotes, space, rkhs, basis, distribution, fixed, data, manifold, mean, finite, journal] [process, information, prior, therefore, change] [training, natural, approach, neural, used, several, using, use, propose, international, representation, better, dataset]
Hierarchical Clustering via Spreading Metrics
Aurko Roy, Sebastian Pokutta
Aurko Roy, Sebastian Pokutta
[clustering, xtij, ultrametric, graph, definition, constraint, ultrametrics, tree, linkage, spreading, induced, partitioning, hierarchy, introduced, cut, flat, vertex, equivalence, subset, thus, iff, joseph, possible] [cost, function, algorithm, let, lemma, set, every, feasible, combinatorial, theorem, relaxation, since, optimal, problem, studied, round, annual, will, arbitrary, defined, study, satisfies] [approximation, log, supplementary, size, factor, approximate, method] [following, solution, also, can, note, characterization, error, rounding, symposium, xij, one, denote, condition] [data, polynomial, given, corresponding, based, metric] [hierarchical, ball, time] [similarity, using, use, pair, work, approach, used, ground, several, sequence, natural]
Finite Sample Prediction and Recovery Bounds for Ordinal Embedding
Lalit Jain, Kevin G. Jamieson, Rob Nowak
Lalit Jain, Kevin G. Jamieson, Rob Nowak
[constraint, probability, pgd, possible, fact] [let, theorem, function, since, set, loss, define, consider, known, learning, show, will, sup, convex, problem, conference, defined, least, assume, case, observe] [log, logistic, gradient, optimization, operator, machine] [matrix, hlt, can, ordinal, norm, nuclear, gram, linear, dik, denote, link, paper, rank, kxi, see, noisy, error, centered, dij, solution, recover, following, orthogonal, magnitude, solving, recovery, snh, note, subspace, symmetric, item] [distance, embedding, given, euclidean, space, corresponding, kernel, data, two, random, section, result, multidimensional, based] [new, observation, information] [prediction, using, like, work, triplet, international]
Structured Matrix Recovery via the Generalized Dantzig Selector
Sheng Chen, Arindam Banerjee
Sheng Chen, Arindam Banerjee
[probability, present] [theorem, bound, general, constant, sup, assume, set, let, satisfies, define, defined, convex, upper, essentially, proof, lemma, will, absolute, implies, class, bounded] [stochastic, deterministic, dual] [norm, matrix, recovery, can, invariant, unitarily, measurement, restricted, width, analysis, error, following, spectral, compatibility, generic, one, chaining, completion, via, condition, noise, sparse, centered, symmetric, rest, seminorm, owl, regularized, high, denote, generalized, vector, dantzig, certain] [gaussian, geometric, random, given, exp, estimation, section, estimator, two, metric, associated, based, gauge, functional] [trace, model] [structured, using, used, unit, work, bounding, use, shown, different]
Online Bayesian Moment Matching for Topic Modeling with Unknown Number of Topics
Wei-Shou Hsu, Pascal Poupart
Wei-Shou Hsu, Pascal Poupart
[number, basic, according, david, discovered, actual, resulting, contains, total] [online, learning, algorithm, set, conference, will, since, exponentially] [bayesian, posterior, likelihood, variational, inference, latent, log, compute, sampling, approximate, update, approximating, expectation] [can, first, also, spectral, exact, denote, global] [dirichlet, distribution, ddm, moment, tdm, test, data, topic, two, hdp, estimate, degenerate, alpha, uniform, well, lda, ohdp, corpus, hdps, fixed, nonparametric, word] [model, prior, hierarchical, found, modeling, new, directly, process, whereas, experimental, inferred] [used, matching, figure, using, different, text, use, dataset, propose, proposed, similar, computer, shown, tested, able]
Efficient Second Order Online Learning by Sketching
Haipeng Luo, Alekh Agarwal, Nicolò Cesa-Bianchi, John Langford
Haipeng Luo, Alekh Agarwal, Nicolò Cesa-Bianchi, John Langford
[dag, number, represents] [algorithm, online, regret, set, learning, appendix, bound, setting, since, loss, return, bounded, even, show, example, theorem, rate, round, constant, lower, study] [update, rad, newton, stochastic, datasets, gradient, method, step, compute, quadratic, efficient, size, siam, stepsize, develop, adagrad, deterministic, due, convergence] [matrix, sketch, order, sparse, sketching, can, diagonal, first, linear, second, invariant, error, running, sketched, one, def, projection, condition, synthetic, son, also, require, computational, small, vector, row, assumption, still, note, computation] [random, journal, two, data, fixed] [time, frequent] [using, similar, performance, work, three]
Adaptive Skills Adaptive Partitions (ASAP)
Daniel J. Mankowitz, Timothy A. Mann, Shie Mannor
Daniel J. Mankowitz, Timothy A. Mann, Shie Mannor
[probability, number, definition] [learning, set, define, defined, adaptive, function, optimal, expected, now, may, algorithm, will, conference, derive, return, general, lifelong] [gradient, solve, objective, update, supplementary, parameterized, convergence] [can, vector, generalized, also, need, solution, matrix, necessary] [two, given, space, well, distribution, polynomial, refer, together] [skill, asap, policy, hyperplane, state, sps, framework, location, reward, action, mdp, automatically, trajectory, reinforcement, learns, misspecified, robocup, executing, striker, agent, hyperplanes, current, defender, ball, temporally, goal, reused, continuous, therefore, move, timothy, misspecification, flipped] [figure, learned, using, feature, domain, learn, different, multiple, task, shown, able, performance, single, seen, building, used]
Combining Fully Convolutional and Recurrent Neural Networks for 3D Biomedical Image Segmentation
Jianxu Chen, Lin Yang, Yizhe Zhang, Mark Alber, Danny Z. Chen
Jianxu Chen, Lin Yang, Yizhe Zhang, Mark Alber, Danny Z. Chen
[structure, hierarchy, level] [may, learning, known, will, context, contextual, strategy] [size, method, university, evaluation] [can, one, component, generalized, much, also, first] [two, slice, result, section, based, six] [information, framework, new, neuron, exploit, along, neuronal] [segmentation, image, deep, fcn, different, biomedical, rnn, feature, convolutional, neural, using, layer, recurrent, training, input, fungus, lstm, performance, combining, map, network, four, scale, preprint, submodule, arxiv, sequence, convolution, architecture, used, clstm, output, approach, extracted, applied, propose, region, work, fully, original, dataset, voxel, use, figure, dame]
Combining Adversarial Guarantees and Stochastic Fast Rates in Online Learning
Wouter M. Koolen, Peter Grünwald, Tim van Erven
Wouter M. Koolen, Peter Grünwald, Tim van Erven
[probability, many, theory] [regret, loss, bernstein, learning, online, theorem, bound, hedge, will, case, squint, setting, fast, excess, expected, show, convex, let, consider, erven, implies, best, koolen, rate, oco, set, risk, inequality, may, function, lemma, hinge, example, appendix, audibert, round, bounded, learner, now, metagrad, adaptive, make, constant, equivalent, vtf, absolute, get] [stochastic, machine, optimization, expectation] [condition, can, order, also, high, main, linear, one, following, assumption, see] [distribution, section, statistical, family, result, point, two, data, infinite] [expert, van, automatically, allows] [sequence, use, compared, classification, using, prediction, used]
beta-risk: a New Surrogate Risk for Learning from Weakly Labeled Data
Valentina Zantedeschi, Rémi Emonet, Marc Sebban
Valentina Zantedeschi, Rémi Emonet, Marc Sebban
[number, introduced] [learning, weakly, surrogate, label, risk, loss, algorithm, svm, instance, problem, function, set, wellsvm, show, will, convex, setting, class, defined, hypothesis, derive, optimal, depends, margin, context, permissible, minimizing, ell] [machine, standard] [can, one, matrix, iterative, formulation, linear, generic, first, vector, noisy, order, noise, min, initialization] [empirical, data, section, two, proportion, provided, given, positive, based, classical, corresponding, mean, gaussian, address] [new, information, deal, state] [labeled, different, using, supervised, training, unlabeled, used, learn, supervision, propose, fully, approach, figure, neural, weak]
Statistical Inference for Cluster Trees
Jisu KIM, Yen-Chi Chen, Sivaraman Balakrishnan, Alessandro Rinaldo, Larry Wasserman
Jisu KIM, Yen-Chi Chen, Sivaraman Balakrishnan, Alessandro Rinaldo, Larry Wasserman
[cluster, tree, merge, definition, pruning, tph, partial, dmm, clustering, valid, pruned, connected, gvhd, suitability, eldridge, find, height, topology, level, department, rule, contains] [confidence, set, appendix, consider, lemma, function, unknown, defined, will, focus, implies, exists, sup, define] [inference, bootstrap, convergence, university] [can, paper, order, see, first, also, mouse, computational, usa, denote] [density, metric, distortion, statistical, two, data, sample, empirical, construct, bandwidth, estimation, well, ring, kernel, positive, insignificant, estimated, functional, statistically, length, true, interleaving, mickey, estimate, distance, way, based, important, carnegie] [control, solid, information] [figure, several, use, three, propose, work, top]
Kronecker Determinantal Point Processes
Zelda E. Mariet, Suvrit Sra
Zelda E. Mariet, Suvrit Sra
[probability, partial, structure, obtained, monotonic, average] [learning, algorithm, set, cost, fast, problem, may, obtain, provide, show, will] [icard, dpp, kronecker, sampling, ron, determinantal, stochastic, large, dpps, efficient, iteration, updating, due, size, increase, faster, log, update, full, iterates, showed, gradient, approximate, key, mcmc, machine, initial] [can, matrix, also, product, one, exact, synthetic, real] [kernel, point, positive, definite, data, two, important, application, given] [time, model, pps, future] [training, using, dataset, ground, table, neural, evaluate, performance, shown, used, memory]
Learning feed-forward one-shot learners
Luca Bertinetto, João F. Henriques, Jack Valmadre, Philip Torr, Andrea Vedaldi
Luca Bertinetto, João F. Henriques, Jack Valmadre, Philip Torr, Andrea Vedaldi
[number, possible, contains] [learning, function, problem, consider, best, class, case] [parameter, large, size, method, evaluation, objective, university, usually, oxford] [can, one, second, order, also, factorized, first, linear, error, solving, small] [given, two, space, embedding, distance] [tracking, model, dynamic, new, simple, baseline] [siamese, learnet, using, exemplar, deep, single, object, convolutional, layer, different, network, discriminative, neural, predicted, architecture, filter, image, output, learn, prediction, visual, character, training, predict, recognition, generative, approach, input, use, shared, trained, work, predicts, stream, performance, applied, frame, classification, figure]
A Probabilistic Framework for Deep Learning
Ankit B. Patel, Minh Tan Nguyen, Richard Baraniuk
Ankit B. Patel, Minh Tan Nguyen, Richard Baraniuk
[probabilistic, number, many, passing, configuration, path] [algorithm, learning, class, max, will, appendix, set, since, focus] [inference, latent, factor, processing, machine, key] [can, also, via, linear, principled, order, formulation, sparse, one] [mixture, test, corresponding] [model, information, new, argmax, framework, hierarchical, template, activity] [drmm, deep, nuisance, training, rendering, generative, dcns, neural, layer, dcn, preprint, arxiv, image, rmm, object, variation, task, work, different, labeled, figure, unsupervised, discriminative, performance, several, using, rendered, multiple, classification, table, convolutional, recognition, rfm, supervised, shallow, reconstruction, network, approach, pixel, relu]
Linear Feature Encoding for Reinforcement Learning
Zhao Song, Ronald E. Parr, Xuejun Liao, Lawrence Carin
Zhao Song, Ronald E. Parr, Xuejun Liao, Lawrence Carin
[number, represented, theory, block, indicator, probability, represents] [function, algorithm, set, learning, problem, theorem, optimal, expected, since, may, provide, example] [approximation, approximate, method, additional] [linear, error, can, good, one, sufficient, following, also, sampled, projection, see, supplemental, matrix] [random, selection, two, fixed, point, based, given, space, result, dimensional, needed] [value, raw, policy, state, reinforcement, next, bellman, action, encoding, model, reward, encoded, pendulum, parr, mdp, angle, controlled, taking, inverted, extend, framework] [feature, encoder, using, training, deep, representation, approach, use, work, used, performance, decoder, figure, previous, prediction, recent, neural, single, learn]
Weight Normalization: A Simple Reparameterization to Accelerate Training of Deep Neural Networks
Tim Salimans, Diederik P. Kingma
Tim Salimans, Diederik P. Kingma
[average] [max] [] [global, dimension, noise, whitening] [gaussian, type] [pool, raw] [relu, leaky, conv, network, dropout, neural, input, layer, architecture, table, softmax, output, rgb]
Dual Learning for Machine Translation
Di He, Yingce Xia, Tao Qin, Liwei Wang, Nenghai Yu, Tieyan Liu, Wei-Ying Ma
Di He, Yingce Xia, Tao Qin, Liwei Wang, Nenghai Yu, Tieyan Liu, Wei-Ying Ma
[message, association, according, quality] [learning, will, set, algorithm, may, conference] [machine, dual, parallel, gradient, log, beginning, warm, method, processing] [can, first, second, computational, following, one] [two, data, based, section, statistical, versus, france] [model, reward, game, agent, reinforcement, search, feedback, pseudo, communication, policy, baseline] [translation, sentence, language, monolingual, bilingual, neural, using, nmt, used, training, trained, improve, natural, source, mechanism, outperforms, table, translate, beam, deux, original, bleu, performance, train, smid, recurrent, use, learn, middle, generate, back, proposed, work, generated, aligned]
Exponential expressivity in deep neural networks through transient chaos
Ben Poole, Subhaneil Lahiri, Maithreyi Raghu, Jascha Sohl-Dickstein, Surya Ganguli
Ben Poole, Subhaneil Lahiri, Maithreyi Raghu, Jascha Sohl-Dickstein, Surya Ganguli
[propagation, thus, number, grows, theory] [exponentially, function, consider, complexity, notion] [large, compute, processing] [can, vector, linear, one, principal, analysis, iterative, phase, high, highly, first, small] [length, curvature, fixed, point, random, curve, dimensional, manifold, chaotic, geometry, space, circle, tangent, boundary, metric, theoretical, exponential, euclidean, expressivity, gaussian, riemannian, distribution, mean, radius, curved, two, growth, gauss, chaos, nonlinear, hli, sigmoidal, quantitatively, measure] [extrinsic, across, information, decision, activity, neuron, simple, correlation, focused, evolution] [deep, neural, layer, input, map, network, depth, shallow, hidden, feedforward, understand, unit, figure, weight, preprint, nonlinearities]
Convex Two-Layer Modeling with Latent Structure
Vignesh Ganapathiraman, Xinhua Zhang, Yaoliang Yu, Junfeng Wen
Vignesh Ganapathiraman, Xinhua Zhang, Yaoliang Yu, Junfeng Wen
[negative, graph, structure, total, number] [convex, learning, max, set, problem, optimal, function, relaxation, will, assume, let, since, algorithm, hull, setting] [latent, optimization, inference, extreme, objective, quadratic, term, gradient, method, convy, operator, machine, recently] [can, min, linear, first, via, rank, error, cvx, mrr, one, occluded, local, vector, arg, also, small, inpainting, support, hard] [conditional, given, random, two, test, word, based] [model, framework, therefore, value] [structured, using, output, discriminative, letter, deep, training, matching, polar, layer, image, unsupervised, map, achieved, representation, weight, patch, performance, prediction, proposed, used, transliteration, higher]
Reconstructing Parameters of Spreading Models from Partial Observations
Andrey Lokhov
Andrey Lokhov
[dmp, psi, node, rec, number, spreading, probability, transmission, partial, diffusion, fact, hts, infection, directed, cascade, involving, tree, fdmp, propagation, infected] [case, problem, learning, algorithm, conference, complexity, even, set, general, defined, will, since] [likelihood, initial, marginal, optimization, large, step, free, energy, full, inference] [can, observed, missing, recovery, one, exact, assumption, small, accurate] [data, given, equation, section, corresponding, well, equal, based, fixed, important] [time, information, dynamic, model, state, complex, observation] [network, reconstruction, using, activation, hidden, figure, international, used, original, approach, use, part]
Minimizing Regret on Reflexive Banach Spaces and Nash Equilibria in Continuous Zero-Sum Games
Maximilian Balandat, Walid Krichene, Claire Tomlin, Alexandre Bayen
Maximilian Balandat, Walid Krichene, Claire Tomlin, Alexandre Bayen
[probability, definition, sum, strong, subset] [regret, set, convex, function, let, online, general, learning, bound, uniformly, theorem, play, nash, strategy, banach, consider, essentially, modulus, will, dom, problem, reflexive, algorithm, lower, example, case, obtain, upper, minimizing, show, proper, assume, repeated, duality, corollary, make, defined, class, continuity] [dual, averaging, strongly, log, semicontinuous, method, additional, supplementary] [assumption, suppose, regularizer, can, denote, sublinear, analysis, also, explicit, following, convexity] [space, metric, section, finite, measure, empirical, given] [player, reward, continuous, payoff, game, action, decision, corresponds, choice] [sequence, compact]
Agnostic Estimation for Misspecified Phase Retrieval Models
Matey Neykov, Zhaoran Wang, Han Liu
Matey Neykov, Zhaoran Wang, Han Liu
[variable, program, theory, knowledge] [algorithm, convex, least, class, function, optimal, case, constant, let, even, satisfies, considered, obtain, since, index, will, exists, assume, inequality, problem, implies, learning, lemma] [log, step, size, full, parameter, requires, refined, solve, standard, quadratic] [mpr, phase, can, retrieval, condition, vector, following, second, sparse, also, first, one, init, link, noisy, proposition, twf, tpm, principal, solving, high, regression, dimension, regularized, linear, semidefinite, sims, analysis, via, sufficiently, addition, satisfying] [estimate, sample, estimation, procedure, two, data, random, based, journal, given, section, gaussian, result, statistical] [model, direction, suggested, information] [proposed, approach, figure, using, single, propose, applied]
Yggdrasil: An Optimized System for Training Deep Decision Trees at Scale
Firas Abuzaid, Joseph K. Bradley, Feynman T. Liang, Andrew Feng, Lee Yang, Matei Zaharia, Ameet S. Talwalkar
Firas Abuzaid, Joseph K. Bradley, Feynman T. Liang, Andrew Feng, Lee Yang, Matei Zaharia, Ameet S. Talwalkar
[tree, ggdrasil, partitioning, split, worker, vertical, number, node, lanet, master, horizontal, xgb, sorted, uncompressed, yggdrasil, bitvector, spark, bitvectors, runtime, yahoo, assigned, ttruth, cache, leaf, total] [learning, cost, optimal, algorithm, set, oost, compression, best] [distributed, iteration, large, compute, requires, faster, machine, implementation, discretization, sequential] [can, sparse, one, via, order, arg, sufficient, compressed, linear, need] [data, two, well, random, section, scan] [communication, decision, time, encoding, continuous, optimized, across, rather, new] [training, feature, depth, memory, using, figure, single, mnist, candidate, without, better, deep, dataset, impact, performance, perform, several]
RETAIN: An Interpretable Predictive Model for Healthcare using Reverse Time Attention Mechanism
Edward Choi, Mohammad Taha Bahadori, Jimeng Sun, Joshua Kulas, Andy Schuetz, Walter Stewart
Edward Choi, Mohammad Taha Bahadori, Jimeng Sun, Joshua Kulas, Andy Schuetz, Walter Stewart
[reverse, number, represented] [learning, contribution, risk, will, case, index, context, appendix, set, label] [step, size, logistic, machine] [can, vector, one, order] [clinical, patient, two, given, embedding, interpretability, test, data, comparable, electronic, specific, word, dpm] [time, visit, retain, model, predictive, ehr, medical, information, health, heart, interpretable, esl, modeling, disease, medication, temporal, healthcare, failure, interpretation, wemb, skin, making] [attention, use, rnn, using, input, prediction, sequence, neural, accuracy, figure, generate, recurrent, hidden, generating, predict, layer, mechanism, single, training, preprint, arxiv, rnns, traditional, softmax, used, table, mlp, predicting, generation, different]
Mixed vine copulas as joint models of spike counts and local field potentials
Arno Onken, Stefano Panzeri
Arno Onken, Stefano Panzeri
[number, probability, tree, thus, obtained, normalized] [margin, function, algorithm, cumulative, will, complexity] [mixed, copula, likelihood, vine, method, lfps, apply, clayton, sampling, parameter, quadratic, full, inference, standard, efficient, fit, derivative] [can, one, denote, analysis, second, condition, need] [discrete, independent, multivariate, entropy, statistical, distribution, data, density, estimate, mutual, based, section, two, joint, gaussian, statistic, construct, estimation, corresponding, journal, sample, denotes, estimated, population] [model, continuous, information, spike, framework, activity, simulated, stimulus, concurrent, brain] [neural, network, different, fully, pair, used, using, field, approach, input]
Optimal Cluster Recovery in the Labeled Stochastic Block Model
Se-Young Yun, Alexandre Proutiere
Se-Young Yun, Alexandre Proutiere
[misclassified, number, sbm, probability, lsbm, cluster, community, block, detection, clustering, present, average, edge, graph, partition, obtained, independently, possible, degree, social] [algorithm, asymptotically, label, set, theorem, let, consider, assume, binary, optimal, general, exists, since, proof, problem, max, provide, expected, show, prove] [stochastic, log, supplementary] [can, recovery, condition, spectral, high, exact, observed, first, denote, item, necessary, matrix, also, minimal, accurate, column, sparse, sufficient, singular, one, symmetric] [random, two, corresponding, proportion, true, positive, reference, section, estimate, given] [model, establish, observation] [using, part, hidden, performance, pair, labeled, work, generated, consists, without]
Deep Learning Games
Dale Schuurmans, Martin A. Zinkevich
Dale Schuurmans, Martin A. Zinkevich
[constraint, vertex, number, theory] [learning, set, regret, nash, protagonist, antagonist, convex, problem, utility, algorithm, appendix, loss, since, function, equilibrium, online, chooses, will, dvt, defined, strategy, best, zannis, conference, zanni, define, consider, playing, appears, finding, even, differentiable, considered, competitive, folded] [gradient, optimization, stochastic, reduction, constrained, parameter, machine, standard, descent] [can, one, also, global, first, interesting, kkt, critical] [given, joint, point, two, data, affine] [action, game, expert, allows, model, simple, information, choice, alternative] [training, deep, supervised, neural, network, output, using, input, figure, activation, investigate, used, matching, conducted, international, feedforward]
Synthesis of MCMC and Belief Propagation
Sung-Soo Ahn, Michael Chertkov, Jinwoo Shin
Sung-Soo Ahn, Michael Chertkov, Jinwoo Shin
[loop, zloop, annealing, number, cycle, probability, worm, ising, partition, graph, bethe, describe, external, graphical, wmin, interaction, pairwise, propagation, science, called, strength, computable, adding] [algorithm, set, theorem, scheme, general, function, binary, case, consider, known, since, strategy, problem] [mcmc, approximation, log, full, inference, approximating, computing, experiment, efficient, sampling, end] [generalized, following, can, first, one, error, via, also, running, popular] [section, journal, estimating, provides, statistical, estimate, sample, given, random, basis, described, distribution, sampler, polynomial] [series, design, model, simulated, allows, belief, information, experimental] [using, generate, proposed, planar, use, approach, propose]
Disentangling factors of variation in deep representation using adversarial training
Michael F. Mathieu, Junbo Jake Zhao, Junbo Zhao, Aditya Ramesh, Pablo Sprechmann, Yann LeCun
Michael F. Mathieu, Junbo Jake Zhao, Junbo Zhao, Aditya Ramesh, Pablo Sprechmann, Yann LeCun
[obtained, variable, belonging, number] [learning, class, set, label, show, observe, access, loss] [approximate, posterior, interpolation, latent, evaluation, likelihood, variational, dec, vae, supplementary, log] [can, one, component, solving, still] [specified, given, conditional, data, two, test, described, procedure, corresponding] [model, information, goal] [unspecified, generative, training, variation, using, deep, able, different, figure, used, adversarial, neural, trained, image, norb, use, learned, disentangle, network, dataset, generated, arxiv, representation, swapping, preprint, unseen, approach, proposed, learn, generate, hidden, object, disentanglement, sprite, visualization, work, yann, supervision, gan, separate, style, without, shown, identity]
Adaptive Neural Compilation
Rudy R. Bunel, Alban Desmaison, Pawan K. Mudigonda, Pushmeet Kohli, Philip Torr
Rudy R. Bunel, Alban Desmaison, Pawan K. Mudigonda, Pushmeet Kohli, Philip Torr
[program, probability, number, present, possible, array, find, compilation] [will, learning, algorithm, set, problem, differentiable, make, element, loss, show, example, version, learnable] [optimisation, machine, efficient, gradient, solve, initial, descent, step, supplementary] [can, first, one, also, generic, matrix, success, good, need] [given, two, data, associated, based, section, biased, type, way] [value, model, controller, instruction, compiler, simple, register, linked, irt, written, required, tape, code, stop, jez, read, execution, design, target, head, correct, found] [neural, use, output, input, memory, figure, learned, using, perform, representation, able, different, work, original, computer, presented, without, learn]
A state-space model of cross-region dynamic connectivity in MEG/EEG
Ying Yang, Elissa Aminoff, Michael Tarr, Kass E. Robert
Ying Yang, Elissa Aminoff, Michael Tarr, Kass E. Robert
[describe, indicates, among] [dependence, set, may, assume] [log, step, method, standard, distributed, processing] [can, linear, connectivity, also, one, error, noise, leading, localization, relative, norm, denote, real, matrix, first, solution, diagonal] [given, estimate, mean, data, two, space, independent, point, estimation, covariance, gaussian, estimating] [model, activity, time, roi, dynamic, mne, sensor, meg, evc, across, within, ppa, information, state, along, lagged, trial, cortical, right, brain, feedback, current, simulated, effect, cortex, interest] [source, using, figure, use, visual, neural, applied, flow, scene, previous, different, multiple]
Fast and Provably Good Seedings for k-Means
Olivier Bachem, Mario Lucic, Hamed Hassani, Andreas Krause
Olivier Bachem, Mario Lucic, Hamed Hassani, Andreas Krause
[quality, clustering, probability, total, cluster, number, science] [algorithm, set, complexity, expected, fast, even, guarantee, theorem, let, consider, provide, uniformly, bound, optimal, competitive, since] [free, proposal, step, log, full, requires, provably, kdd, key, approximation, initial, sampling, rna] [solution, error, computational, can, preprocessing, first, sampled, one, main, good, denote, also] [data, chain, ssumption, seeding, distance, length, distribution, random, theoretical, quantization, two, bachem, center, sample, afk, section, based] [markov, web] [using, table, propose, compared, without, different, outperforms, used]
Single-Image Depth Perception in the Wild
Weifeng Chen, Zhao Fu, Dawei Yang, Jia Deng
Weifeng Chen, Zhao Fu, Dawei Yang, Jia Deng
[agree, many, possible] [algorithm, will, query, loss, set, yet, learning, function] [method, datasets, large, full] [relative, ordinal, can, error, one, symmetric, still, existing, sampled, first, significantly] [metric, point, two, random, closer, data, estimation, sample, estimate, test, estimating, well] [new, unconstrained, human, system, intrinsic] [depth, network, image, using, zoran, dataset, nyu, single, training, trained, deep, figure, wild, eigen, train, consists, work, per, predicted, use, used, perception, superpixel, learn, performance, segmentation, approach, input, pair, convolutional, kinect, prediction, semantic, recent, preprint, arxiv, neural, outperforms, predicts]
Bayesian optimization for automated model selection
Gustavo Malkomes, Charles Schaff, Roman Garnett
Gustavo Malkomes, Charles Schaff, Roman Garnett
[base, number, probabilistic, potential] [function, set, will, learning, conference, consider, define, active, expected, case, problem, focus, defined, best, arbitrary] [bayesian, machine, log, hyperparameters, optimization, compute, method, select, datasets, latent, processing, appropriate, approximation, iteration, especially] [can, also, via, note, regression, observed, one] [kernel, given, space, gaussian, two, distance, mean, covariance, infinite, hellinger, data, described, selection, distribution, automated, construct, squared, associated, random, grammar, fixed, well] [model, evidence, prior, search, process, hyperparameter, explain, information, complex, acquisition, simple, observation, modeling, value, framework] [dataset, training, used, neural, use, using, work, per, candidate, perform, similarity, approach, proposed, novel]
Learning Influence Functions from Incomplete Observations
Xinran He, Ke Xu, David Kempe, Yan Liu
Xinran He, Ke Xu, David Kempe, Yan Liu
[influence, diffusion, incomplete, node, dic, retention, seed, dlt, edge, social, number, probability, cascade, cic, complete, graph, activated, improper, mae, notice, wuv, independently] [learning, function, set, pac, algorithm, learnability, let, problem, loss, rate, active, proper, theorem, class, will, focus, even, appendix, provide, learnable, consider] [parameter, efficient, likelihood, log] [can, missing, observed, also, error, following, one, synthetic, linear] [distribution, data, estimation, independent, random, true, result, sample, two] [model, information, goal, process, time, observation] [activation, use, using, learned, learn, network, dataset, figure, training, approach, three]
MoCap-guided Data Augmentation for 3D Pose Estimation in the Wild
Gregory Rogez, Cordelia Schmid
Gregory Rogez, Cordelia Schmid
[number, normalized, average, probability, resulting] [learning, algorithm, show, set] [large, method, constrained, compute, blending, requires, size] [synthetic, error, real, can, also, cmu, existing, first] [estimation, data, joint, corresponding, two, given, considering, associated, distance] [human, model, new, information] [pose, training, image, using, different, dataset, trained, approach, synthesis, used, lsp, classifier, generate, annotated, cnn, use, mocap, deep, rendering, pixel, body, convolutional, motion, capture, train, similar, classification, engine, table, performance, single, compare, better, camera, figure, cnns, neural, source, captured, outperforms, recent, work, evaluate, map, aligned, text, mosaic]
Local Maxima in the Likelihood of Gaussian Mixture Models: Structural Results and Algorithmic Consequences
Chi Jin, Yuchen Zhang, Sivaraman Balakrishnan, Martin J. Wainwright, Michael I. Jordan
Chi Jin, Yuchen Zhang, Sivaraman Balakrishnan, Martin J. Wainwright, Michael I. Jordan
[probability, neighborhood, number, many] [algorithm, function, theorem, learning, least, will, problem, existence, scheme, constant, even, annual, show, conference, setting, consider, case, uniformly, optimal, provide, general, focus] [likelihood, gradient, convergence, saddle, converges, method] [local, initialization, can, global, one, critical, first, order, main, vector, arbitrarily, component, also, small] [mixture, gaussian, population, random, bad, true, two, maximum, result, point, mean, spherical, finite, gaussians, sample, converge, annals, section, statistical, srebro, distribution, gmms, favorable, estimation, practical, cgap, gmm, given, data, density] [model, form, search] [initialized, open, recent, work]
Correlated-PCA: Principal Components' Analysis when Data and Noise are Correlated
Namrata Vaswani, Han Guo
Namrata Vaswani, Han Guo
[number, enough, possible, block] [set, bound, let, problem, assume, algorithm, will, define, complexity, theorem, upper, example, least, lower, smaller, index, study, consider, since, hold, get, proof, interval] [log, large, compute, supplementary] [assumption, matrix, pca, one, can, small, also, first, evd, itt, subspace, noise, following, eigenvectors, principal, condition, denote, eigenvalue, pcp, solution, need, satisfying, sparse, error, much, singular, observed, solving, incoherence, mutually, completion, zero, nonzero] [data, given, sample, robust, result, covariance, true, mean, refer, generalization] [time] [video, using, use, top, used, sequence, moving, neural]
Sample Complexity of Automated Mechanism Design
Maria-Florina F. Balcan, Tuomas Sandholm, Ellen Vitercik
Maria-Florina F. Balcan, Tuomas Sandholm, Ellen Vitercik
[hierarchy, payment, split, describe, theory, greater] [revenue, complexity, auction, reserve, bidder, function, optimal, class, combinatorial, set, valuation, allocation, bundle, show, learning, conference, prove, vcg, ama, theorem, proof, bound, consider, tuomas, grand, bundling, rademacher, every, let, may, price, annual, designer, study, mbarps, reva, mbarp, upper, defined, interval, general, derive, expected, analyze, morgenstern, problem] [log, deterministic, convergence, machine, mixed, supplementary, optimization] [can, item, one, much, following, vector, symposium] [sample, automated, two, section, uniform, distribution, space, fixed] [design, determine, simple, must] [mechanism, used, work]
Multimodal Residual Learning for Visual QA
Jin-Hwa Kim, Sang-Woo Lee, Donghyun Kwak, Min-Oh Heo, Jeonghee Kim, Jung-Woo Ha, Byoung-Tak Zhang
Jin-Hwa Kim, Sang-Woo Lee, Donghyun Kwak, Min-Oh Heo, Jeonghee Kim, Jung-Woo Ha, Byoung-Tak Zhang
[number, block, multiplication] [learning, conference, since, may, function, idea] [method, extra, machine, processing] [linear, vector, explicit, first, can, overall, dimension] [joint, based, given, embedding, section] [model, information, effect, alternative, effectively] [visual, residual, attention, question, deep, tanh, neural, using, feature, mapping, multimodal, image, preprint, arxiv, table, used, figure, identity, shortcut, mrn, output, pretrained, attentional, vqa, input, use, learn, dataset, international, representation, language, visualization, shown, various, spatial, multiple, vision, accuracy, novel, recent, propose, answering, rnn, mechanism, performance]
Gaussian Processes for Survival Analysis
Tamara Fernandez, Nicolas Rivera, Yee Whye Teh
Tamara Fernandez, Nicolas Rivera, Yee Whye Teh
[covariate, probability, number, proportional, knowledge, variable] [function, set, algorithm, observe, scheme, interval, choose, consider, since, just, let] [inference, bayesian, method, parameter, approximation, sampling, line, tractable, processing] [can, first, following, proposition, analysis, one, need, second, noisy, overall, denote, synthetic] [survival, gaussian, data, hazard, random, covariates, sample, given, kernel, nonparametric, point, distribution, treatment, cox, censoring, weibull, rejected, centred, exponential, independent, type, accepted, jump, way, length, estimator, parametric, associated, gamma, patient, annals, two] [model, process, time, prior, poisson, baseline, intensity, right, expert] [score, use, figure, using, scale, neural, perform, used]
Image Restoration Using Very Deep Convolutional Encoder-Decoder Networks with Symmetric Skip Connections
Xiaojiao Mao, Chunhua Shen, Yu-Bin Yang
Xiaojiao Mao, Chunhua Shen, Yu-Bin Yang
[average, block, contains] [learning, achieve, show, observe, achieves, obtain, best, even] [size, evaluation, gradient, large, method] [can, ieee, recover, noise, much, recovering, existing, second, sparse, one, corrupted, symmetric, also, first, handle] [two, based, testing] [model, framework, passed] [image, network, convolutional, skip, denoising, deep, deconvolution, deconvolutional, using, training, performance, psnr, different, better, ssim, table, neural, single, restoration, shown, convolution, feature, proposed, fully, input, layer, use, figure, propose, clean, deeper, used, multiple, residual, output, cscn, slightly, without, trained, dnn, train, work]
Long-term Causal Effects via Behavioral Game Theory
Panagiotis Toulis, David C. Parkes
Panagiotis Toulis, David C. Parkes
[causal, assignment, assigned, thus, definition, possible, according, theory, structural] [algorithm, set, every, assume, let, defined, revenue, strategy, reserve, expected, price, known, auction, play] [experiment, step, inference, objective, latent, method, university] [assumption, can, observed, also, denote, vector, one, matrix] [population, two, data, treatment, denotes, estimate, sample, given, estimation, conditional, space, journal, statistical, estimated, random] [policy, agent, model, behavioral, game, effect, time, behavior, experimental, new, temporal, across, action, payoff, economic, multiagent, information, economy, value, baseline, assuming, framework, therefore, evolve, american, infer, rapoport, dynamical] [figure, approach, different, used, evaluate]
Hardness of Online Sleeping Combinatorial Optimization Problems
Satyen Kale, Chansoo Lee, David Pal
Satyen Kale, Chansoo Lee, David Pal
[graph, path, called, theory] [set, online, sleeping, regret, problem, combinatorial, learning, algorithm, nline, loss, instance, round, hardness, learner, awake, adversary, agnostic, dnf, algosco, hortest, since, now, pac, every, let, chooses, disjunction, least, define, property, inimum, show, satisfies, prove, algdisj, bipartite, index, richness, maximal, heaviness, label, implies, known, upper, lower, kanade, consider, sla] [optimization, efficient, size, reduction, stochastic, parameter] [hard, ranking, exactly, note, running, one] [given, two, result, fixed, computationally] [action, time, decision, policy] [ground, figure, shown, open, input, available, matching, labeled]
A forward model at Purkinje cell synapses facilitates cerebellar anticipatory control
Ivan Herreros, Xerxes Arsiwalla, Paul Verschure
Ivan Herreros, Xerxes Arsiwalla, Paul Verschure
[rule, contains, level, interaction, actual, basic, transmission] [learning, adaptive, optimal, will, even, scheme, algorithm] [gradient, smooth, descent, update, parallel] [error, can, signal, linear, note, one, matrix, computational, also, min] [reference, basis, given, conditioning, based, transport, difference, provides] [control, cerebellar, eligibility, feedback, model, anticipatory, trace, system, time, cfpc, motor, cerebellum, reactive, controller, purkinje, trial, forward, current, plant, response, pursuit, cell, module, target, counterfactual, delay, fiber, impulse, synaptic, predictive, anatomical, within, receives, indicate, simulation, behavior, eye] [output, architecture, use, layer, performance, generate, using, input, filter, weight, neural, task, figure, position]
Infinite Hidden Semi-Markov Modulated Interaction Point Process
matt zhang, Peng Lin, Peng Lin, Ting Guo, Yang Wang, Yang Wang, Fang Chen
matt zhang, Peng Lin, Peng Lin, Ting Guo, Yang Wang, Yang Wang, Fang Chen
[probability, ancestor, interaction, represent, number, variable, developed] [set, function, defined, drawn, consider] [latent, inference, energy, sampling, bayesian, stochastic, step, consumption] [can, via, following, first, also, synthetic, one] [point, hmm, emission, kernel, conditional, infinite, distribution, data, measure, hdp, sample, resampling, space, associated, sampler, distance, hamming, dirichlet, two, given, random, section] [state, process, event, triggering, model, arrival, intensity, hawkes, transition, time, temporal, particle, correlation, failure, poisson, new, markov, emitted, smc, observation, trading, prior, duration, information, pipe] [proposed, hidden, used, use, sequence, background, part, performance, compared]
Improved Techniques for Training GANs
Tim Salimans, Ian Goodfellow, Wojciech Zaremba, Vicki Cheung, Alec Radford, Xi Chen, Xi Chen
Tim Salimans, Ian Goodfellow, Wojciech Zaremba, Vicki Cheung, Alec Radford, Xi Chen, Xi Chen
[quality, find, number, thus, ian, becomes] [learning, loss, nash, may, cost, virtual, function, equilibrium, distinguish, class, label] [minibatch, batch, normalization, gradient, objective, descent, standard, large] [can, real, one] [data, point, metric, well, discrimination, described, sample, distribution, given, test] [model, human, new, player] [training, generative, using, discriminator, generator, generated, arxiv, adversarial, preprint, inception, deep, gan, pmodel, gans, use, feature, classifier, neural, score, matching, figure, layer, image, output, network, generate, approach, table, proposed, work, dataset, lunsupervised, single, trained, learn, able, labeled, train, unsupervised, collapse, supervised, used, shown]
Showing versus doing: Teaching by demonstration
Mark K. Ho, Michael Littman, James MacGlashan, Fiery Cushman, Joe Austerweil, Joseph L. Austerweil
Mark K. Ho, Michael Littman, James MacGlashan, Fiery Cushman, Joe Austerweil, Joseph L. Austerweil
[probability, number, possible, department] [learning, show, function, even, concept, optimal, will, best, example, learner, algorithm, conference, choose, set, chooses] [experiment, bayesian, standard, university, parameter] [can, one, signal, also, condition, row, first] [people, two, given, well, true, distribution, maximum] [reward, model, teaching, showing, pedagogical, goal, irl, safe, behavior, reinforcement, tile, teacher, inverse, agent, observer, pdoing, state, expert, human, teach, trajectory, simply, colored, policy, action, won, value, pobserving, planning, demonstration, told, location, infer, grid, worth] [better, shown, figure, task, learned, international, three, learn, calculated]
A Communication-Efficient Parallel Algorithm for Decision Tree
Qi Meng, Guolin Ke, Taifeng Wang, Wei Chen, Qiwei Ye, Zhi-Ming Ma, Tieyan Liu
Qi Meng, Guolin Ke, Taifeng Wang, Wei Chen, Qiwei Ye, Zhi-Ming Ma, Tieyan Liu
[number, tree, attribute, voting, according, split, histogram, thus, probability, ctr, node, ltr, find, parallelize, larger, total, bestsplit, leftsum, gbdt, informativeness, becomes] [algorithm, best, will, since, cost, lower, theorem, learning, bound, achieve, choose, even, smaller] [parallel, machine, size, select, large, sequential, big, boosting] [local, can, global, one, globally, also, order, need, comparison, small, analysis, see, following, rank, good, note, much, high] [data, point, theoretical, based, quantized, two] [decision, communication, information, gain, new, process, communicate, speed] [training, accuracy, used, different, table, classification, using, volume, figure, proposed, similar, conducted]
Unsupervised Learning from Noisy Networks with Applications to Hi-C Data
Bo Wang, Junjie Zhu, Armin Pourshafeie, Oana Ursu, Serafim Batzoglou, Anshul Kundaje
Bo Wang, Junjie Zhu, Armin Pourshafeie, Oana Ursu, Serafim Batzoglou, Anshul Kundaje
[number, community, genomic, genome, interaction, detection, clustering, node, structure, denoise, chromosome, chromatin, ctcf, incorporate, subset] [problem, may, minimize, set, function, algorithm, confidence, general, learning, appendix, will, cost] [optimization, method, respect, objective, solve, protein, due] [can, matrix, noisy, noise, denoised, order, one, also, link, local, small, addition, confident, missing, observed, regulatory, high, alternating, highly, detecting, histone] [data, important] [framework, information, biological, baseline, subject, determine, value, human] [network, using, used, resolution, different, sij, figure, use, map, denoising, approach, multiple, ground, performance, weight, stanford, capture, unsupervised, single, binding]
Identification and Overidentification of Linear Structural Equation Models
Bryant Chen
Bryant Chen
[identification, identified, identify, causal, graph, structural, identifiable, directed, pearl, node, admissible, recursive, definition, coefficient, earl, criterion, bidirected, independence, removing, constraint, tian, allowed, edge, descendant, developed, chen, ian, discovery, dormant, path, shpitser, recursively, testable, overidentifying, identifying, resulting, allow, brito, discovered, many, connected, verma, ezy] [set, will, may, algorithm, conference, theorem, show, let, since, satisfies, equivalent, function, exists, obtain] [method, additional] [linear, can, decomposition, error, also, matrix, first] [equation, two, given, covariance, distribution] [model, artificial, uncertainty, head, direct] [using, figure, able, applied, use]
Equality of Opportunity in Supervised Learning
Moritz Hardt, Eric Price, None, Nati Srebro
Moritz Hardt, Eric Price, None, Nati Srebro
[threshold, false, attribute, definition, parity, possible, intersection, horizontal, thus, getting] [optimal, roc, notion, loss, will, might, binary, function, fairness, rate, learning, outcome, always, max, convex, consider, since, satisfies, case, depends, satisfy] [oblivious, requires, big] [can, group, also, fraction, linear, profit, accurate, one, require, def] [predictor, equalized, odds, protected, equal, opportunity, positive, derived, fico, demographic, true, two, data, bayes, curve, joint, loan, distribution, conditional, point, well, people, based, deriving, result, construct, given, random, specified] [target, white, within, goal, information] [score, different, figure, training, supervised, using, used, prediction, use, better, single]
Combinatorial semi-bandit with known covariance
Rémy Degenne, Vianney Perchet
Rémy Degenne, Vianney Perchet
[possible, number, structure, coefficient, find] [regret, bound, algorithm, let, case, problem, lemma, combinatorial, bounded, bandit, subgaussian, general, lower, setting, will, arm, upper, confidence, appendix, ivt, show, theorem, combes, proof, smaller, consider, least, get, pulled, choosing, set, online, unknown, learning, kveton] [log, stage, term, supplementary, parameter, stochastic, processing, variance, due] [matrix, linear, can, one, analysis, suppose, regression, also, diagonal, maxi, union, following] [independent, covariance, positive, maximum, equal, given, two, finite, mean, estimator] [action, information, event, reward, design, exploration, goal, form] [using, used, use, neural, different]
Iterative Refinement of the Approximate Posterior for Directed Belief Networks
Devon Hjelm, Ruslan R. Salakhutdinov, Kyunghyun Cho, Nebojsa Jojic, Vince Calhoun, Junyoung Chung
Devon Hjelm, Ruslan R. Salakhutdinov, Kyunghyun Cho, Nebojsa Jojic, Vince Calhoun, Junyoung Chung
[directed, graphical, many, average, number] [learning, adaptive, algorithm, conference, general] [refinement, posterior, approximate, variational, importance, air, inference, monte, sampling, carlo, gradient, variance, lowerbound, stochastic, log, irvi, refined, step, latent, likelihood, additional, initial, sequential, sbn, reweighted, rws, university, sigmoid, machine, autoregressive, darn, unbiased] [can, iterative, also] [estimate, test, sample, discrete, well, conditional, true, density, provided, procedure] [model, belief, demonstrate] [recognition, network, generative, using, trained, training, used, neural, better, deep, effective, arxiv, preprint, approach, use, available, improve, mnist, train, international, final, layer, work, improves, figure]
Unifying Count-Based Exploration and Intrinsic Motivation
Marc Bellemare, Sriram Srinivasan, Georg Ostrovski, Tom Schaul, David Saxton, Remi Munos
Marc Bellemare, Sriram Srinivasan, Georg Ostrovski, Tom Schaul, David Saxton, Remi Munos
[probability, number, proportional, den] [learning, may, conference, consider, problem, theorem, notion, general, will, ratio, now, define, maximizing, since, defined, provide, optimistic] [progress, posterior, approximation, machine] [also, assumption, related, can, one, zero, error, note] [density, empirical, distribution, theoretical, result, derived, particular, two, lim] [model, exploration, information, intrinsic, motivation, gain, atari, reinforcement, state, agent, bonus, salient, visit, event, across, policy, environment, uncertainty, must, tabular, artificial, arcade, game, reward, transition, van, recoding, motivated] [prediction, sequence, used, score, use, figure, neural, approach, international, deep, training, without, performance]
Bayesian Optimization with a Finite Budget: An Approximate Dynamic Programming Approach
Remi Lam, Karen Willcox, David H. Wolpert
Remi Lam, Karen Willcox, David H. Wolpert
[programming, number, configuration, possible] [function, algorithm, optimal, budget, set, utility, problem, greedy, defined, expected, best, define, worst, strategy, discount, improvement, consider, maximizes, conference] [optimization, objective, bayesian, iteration, posterior, approximate, nested, initial, fmin, evaluation, lookahead, expensive, variance, stage, rolling, computed, characterized, machine, approximation, respect, auxiliary] [can, global, gap, note, following, solving, also, formulation, one, existing] [mean, finite, given, gaussian, space, journal, statistical, distribution] [rollout, design, reward, dynamic, state, control, value, heuristic, median, system, policy, next, simulated, information, model, new] [using, use, training, used, performance, propose, several, evaluate, proposed, achieved, table, approach]
Linear Relaxations for Finding Diverse Elements in Metric Spaces
Aditya Bhaskara, Mehrdad Ghadiri, Vahab Mirrokni, Ola Svensson
Aditya Bhaskara, Mehrdad Ghadiri, Vahab Mirrokni, Ola Svensson
[subset, number, quality, many, incorporate, clique, obtained, total] [diversity, set, algorithm, let, matroid, will, consider, xir, maximization, randomized, show, chosen, now, greedy, problem, constant, implies, lemma, cardinality, known, bound, every, study, theorem, relevance, general, since] [approximation, objective, factor, size, step, large, additional, picked] [can, rounding, also, one, solution, linear, note, ieee, local, suppose, denote, planted, much, main] [metric, two, section, well, distance, data, space, result, selection] [search, value, maximize, goal, new, web] [figure, used, better, different, dataset, feature, output, computer, pair]
f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization
Sebastian Nowozin, Botond Cseke, Ryota Tomioka
Sebastian Nowozin, Botond Cseke, Ryota Tomioka
[reverse, probabilistic] [function, learning, class, pearson, provide, algorithm, show, bound, general, set, lower, convex, case, minimization, defined, special] [variational, divergence, objective, log, method, saddle, approximate, gradient, supplementary, conjugate, likelihood] [can, one, also, linear, see, note] [sample, distribution, estimation, two, true, given, density, squared, kde, hellinger, kernel, mixture, point, random, data, estimate, nguyen, difference, test] [model, information, framework] [generative, neural, gan, training, using, table, use, generator, mnist, output, train, used, network, work, activation, trained, different, figure, proposed, deep, approach, input, natural, final, representation, learned, goodfellow, three, layer]
Avoiding Imposters and Delinquents: Adversarial Crowdsourcing and Peer Prediction
Jacob Steinhardt, Gregory Valiant, Moses Charikar
Jacob Steinhardt, Gregory Valiant, Moses Charikar
[reliable, constraint, probability, block, number, crowdsourcing, detection, influence, community, quality, graph, according, actual, worker, among, possible] [algorithm, let, set, will, obtain, lemma, setting, online, even, general, may, show, assume, concentration, problem, remaining, least, proof, rate, consider, conference] [stochastic, key, large, involves] [raters, matrix, can, item, rating, nuclear, norm, noisy, rater, peer, mij, also, row, recover, vector, assumption, note, suppose, denote, semirandom, proposition, fraction, one, via, following, sparse, good, much, technical, arbitrarily, necessary, recovering, main] [section, given, selected, random, true, robust, result, independent] [model, information, goal] [using, work, randomly, output, amount, evaluate]
Stein Variational Gradient Descent: A General Purpose Bayesian Inference Algorithm
Qiang Liu, Dilin Wang
Qiang Liu, Dilin Wang
[find] [set, algorithm, function, learning, general, since, let, consider, case, problem] [variational, gradient, inference, descent, stein, bayesian, log, method, monte, transforms, carlo, smooth, size, iteration, stochastic, divergence, posterior, expectation, sgld, parallel, sequential, optimization, large, initial, jacobian, requires, step, approximate, derivative, full, calculate] [can, also, matrix, perturbation, zero, one, see, small, via, need, kernelized] [distribution, kernel, mean, density, gaussian, random, functional, data, test, rbf, pbp, reduces, empirical, theoretical, procedure, discrepancy, independent] [particle, target, form, simple, new, except, purpose, take, prior, inverse] [using, use, identity, map, figure, neural, different, transform, preprint, arxiv, used, single]
Batched Gaussian Process Bandit Optimization via Determinantal Point Processes
Tarun Kathuria, Amit Deshpande, Pushmeet Kohli
Tarun Kathuria, Amit Deshpande, Pushmeet Kohli
[probability, subset] [regret, function, algorithm, est, ucb, choose, maximization, set, chosen, bound, bandit, theorem, consider, problem, will, appendix, proof, defined, improvement, cumulative, show, just, confidence, greedy, case, upper, expected, choosing, let] [batch, optimization, sampling, dpp, bucb, size, bayesian, posterior, determinantal, iteration, fastxml, large, method, batched, machine, compute, extreme, due, experiment] [via, can, one, also, first, arg, much, popular] [kernel, point, based, gaussian, maximum, entropy, two, provided, mean, selected, given, selection] [process, information, search, value, robot, next] [better, performance, using, presented, work, dataset, perform]
The Power of Optimization from Samples
Eric Balkanski, Aviad Rubinstein, Yaron Singer
Eric Balkanski, Aviad Rubinstein, Yaron Singer
[number, many, obtained, influence, constraint, subset] [submodular, set, algorithm, function, optimal, obtain, monotone, contribution, lemma, learning, show, loss, since, element, best, bound, let, assume, greedy, expected, consider, cardinality, property, case, learnable, defined, polynomially, exists, concentration, impossibility, known, return, theorem, least, inequality, proof, tight, drawn, feasible, bounded] [log, approximation, optimization, marginal, size] [good, first, can, solution, analysis, second, main, also, high, small] [curvature, random, bad, distribution, given, two, result, construct, data] [value, model, poor, information, goal, optimize, decision] [learned, figure, using, work, better, consists, learn, used]
Single Pass PCA of Matrix Products
Shanshan Wu, Srinadh Bhojanapalli, Sujay Sanghavi, Alexandros G. Dimakis
Shanshan Wu, Srinadh Bhojanapalli, Sujay Sanghavi, Alexandros G. Dimakis
[number, spark] [algorithm, optimal, show, set, let, appendix, theorem, now, smaller, case, streaming, may, complexity, ratio, observe, online, annual] [approximation, compute, size, step, pass, sampling, large, log, distributed, datasets, machine, standard] [matrix, spectral, can, norm, rank, low, error, product, sketch, lela, column, one, sketching, also, sampled, dot, rescaled, first, qij, synthetic, cone, computes, pca, kai, note, kbj, entry, vector, svd, symposium, analysis] [two, given, embedding, sample, random, data, important] [information, angle, desired] [figure, better, using, compared, perform, single, three, use, dataset, performance, preprint, arxiv, outperforms, compare]
Efficient Nonparametric Smoothness Estimation
Shashank Singh, Simon S. Du, Barnabas Poczos
Shashank Singh, Simon S. Du, Barnabas Poczos
[probability, number] [since, theorem, let, general, may, function, bound, case, class, minimax, smoothing, learning] [inner, parameter, processing, efficient, convergence, variance, large, divergence] [can, error, norm, computational, denote, related, one, also, order, suppose] [sobolev, distance, estimator, density, test, estimation, nonparametric, estimated, asymptotic, estimating, section, statistical, squared, true, kernel, functionals, estimate, smoothness, functional, mean, theoretical, data, fourier, computationally, multivariate, bias, testing, journal, denotes, based, distribution, space, sample, gaussians, statistically, suggest, uniform] [information, time] [work, figure, neural, use, used, different, using]
Sequential Neural Models with Stochastic Layers
Marco Fraccaro, Søren Kaae Sønderby, Ulrich Paquet, Ole Winther
Marco Fraccaro, Søren Kaae Sønderby, Ulrich Paquet, Ole Winther
[average, obtained, structure, report] [learning, dependence, set, will, depends, show] [inference, stochastic, variational, srnn, posterior, approximation, deterministic, latent, log, blizzard, music, timit, parameterization, sequential, ssm, approximate, polyphonic, depend, parameterized, term, elbo, monte, compute, likelihood, resq, vrnn, storn, intractable] [can, also, one, see] [distribution, data, test, mean, nonlinear, gaussian, space, section, true, two, done, given] [model, state, time, prior, modeling, information, therefore, predictive, raw, temporal, uncertainty, future] [network, neural, using, recurrent, figure, rnn, sequence, generative, used, speech, hidden, use, improve, table, rnns, shown, residual, deep, output]
Stochastic Structured Prediction under Bandit Feedback
Artem Sokolov, Julia Kreutzer, Stefan Riezler, Christopher Lo
Artem Sokolov, Julia Kreutzer, Stefan Riezler, Christopher Lo
[pairwise, criterion, structure, present, partial, number, obtained] [learning, algorithm, loss, bandit, constant, function, expected, complexity, convex, optimal, minimization, set, best, smallest, lipschitz, smaller, since, sfo, will] [stochastic, convergence, gradient, machine, optimization, objective, standard, development, evaluation, respect, variance, chunking, iteration, full, update, large, analyzed] [preference, can, norm, analysis, following, interactive, sampled, note] [data, squared, given, asymptotic, numerical, practical, based, distribution, result, exponential, estimate] [feedback, information, model, speed, form, motivated] [task, structured, output, prediction, translation, feature, use, performance, training, approach, input, predicted, table, using, presented, sequence, gold, seen, three]
Bayesian Intermittent Demand Forecasting for Large Inventories
Matthias W. Seeger, David Salinas, Valentin Flunkert
Matthias W. Seeger, David Salinas, Valentin Flunkert
[forecasting, negative, larger] [risk, function, learning, smoothing, defined, algorithm, day, even] [likelihood, latent, large, intermittent, log, stock, inference, negbin, approximation, bayesian, term, approximate, forecast, ets, stage, optimization, inner, bursty, quadratic, expectation, machine, method, run, logistic, compute, posterior, implementation] [demand, linear, generalized, item, related, short, can, running, one] [gaussian, section, data, exponential, well, automated, given, laplace, two, space] [time, state, kalman, series, range, medium, poisson, model, mode, prior] [use, work, transfer, figure, used, prediction, novel, dataset, several, outperforms, training, approach]
Automatic Neuron Detection in Calcium Imaging Data Using Convolutional Networks
Noah Apthorpe, Alexander Riordan, Robert Aguilar, Jan Homann, Yi Gu, David Tank, H. Sebastian Seung
Noah Apthorpe, Alexander Riordan, Robert Aguilar, Jan Homann, Yi Gu, David Tank, H. Sebastian Seung
[false, david, detection, many, detect, number] [may, learning, active, precision, set, algorithm] [datasets, processing] [calcium, imaging, also, noise, one, analysis, require] [basis, test, automated, true, selection, two] [human, roi, time, series, temporal, cell, neuron, cortex, took, activity, brain] [network, image, convnet, output, ground, dataset, convolutional, pixel, truth, mec, accuracy, figure, conv, neural, training, input, using, use, trained, score, used, original, per, recall, transiently, supervised, annotation, convnets, video, stack, shown, visual, predicted, single, inactive, computer, col, traditional]
Minimax Optimal Alternating Minimization for Kernel Nonparametric Tensor Learning
Taiji Suzuki, Heishiro Kanagawa, Hayato Kobayashi, Nobuyuki Shimizu, Yukihiro Tagami
Taiji Suzuki, Heishiro Kanagawa, Hayato Kobayashi, Nobuyuki Shimizu, Yukihiro Tagami
[number, normalized] [function, algorithm, minimization, learning, optimal, minimax, problem, convex, conference, achieves, set, complexity, theorem, bounded, let, rate, bound] [method, convergence, optimization, processing, bayesian, initial, parameter, term, factor] [tensor, amp, alternating, linear, analysis, assumption, can, regression, solution, multitask, one, error, also, regularization, component, following, consumer, completion, computational, restaurant, rank, good, suppose, matrix, decomposition, sufficiently, optimality] [true, estimator, kernel, nonparametric, data, statistical, gaussian, procedure, given, theoretical, estimation, nonlinear, based, bayes, space, rkhs, generalization, sample] [model, information, process] [different, shown, task, several, updated, neural, performance, using, input, international, similarity, used]
“Congruent” and “Opposite” Neurons: Sisters for Multisensory Integration and Segregation
Wen-Hao Zhang, He Wang, K. Y. Michael Wong, Si Wu
Wen-Hao Zhang, He Wang, K. Y. Michael Wong, Si Wu
[probabilistic, connected, whose, present, number, becomes, strength] [concentration, optimal, since, rate] [processing, posterior, bayesian, decentralized, large] [can, vector, via, first, computational, one, also] [two, distribution, given, mean, gaussian, journal, exp, length] [opposite, congruent, cue, multisensory, integration, information, module, von, brain, neuron, disparity, model, segregation, vestibular, heading, neuroscience, reciprocal, mstd, direction, preferred, neuronal, sensory, stimulus, measured, geometrical, tuning, direct, experimental, firing, nature, indirect, prior, whether, interpretation, vip, whereas, reciprocally, form, implement, offset, circular, integrates] [visual, neural, connection, different, network, feedforward, similar, input, predicted, single, multiple, achieved]
Direct Feedback Alignment Provides Learning in Deep Neural Networks
Arild Nøkland
Arild Nøkland
[path, connected, disconnected] [learning, will, let, provide, minimize, loss, even, show, theorem] [update, gradient, method, machine, initial, derivative, steepest] [error, can, signal, zero, first, also, symmetric, matrix, see, order, one, initialization, asymmetric] [random, fixed, test, principle, data] [feedback, forward, direction, direct, biologically, target, neuron, indicate, descending, experimental, directly, learns] [hidden, layer, training, dfa, tanh, network, neural, deep, used, output, weight, able, trained, different, alignment, mnist, input, figure, calculated, plausible, table, activation, performed, convolutional, dropout, work, using]
End-to-End Kernel Learning with Supervised Convolutional Kernel Networks
Julien Mairal
Julien Mairal
[present, obtained, introduced, number] [learning, set, function, may, loss, consider, algorithm, obtain, scheme, defined, idea, since, appendix, rate, make, every] [gradient, respect, optimization, large, size, stochastic, method, parameter] [linear, also, matrix, one, can, local, subspace, error, projection, indeed, onto] [kernel, data, given, two, rkhs, classical, space, test, positive, definite, gaussian] [model, optimizing, multilayer, hierarchical] [image, convolutional, network, neural, deep, training, map, using, use, pooling, layer, approach, without, classification, supervised, svhn, perform, previous, unsupervised, representation, natural, pixel, achieved, prediction, spatial, single, several, consists]
On Multiplicative Integration with Recurrent Neural Networks
Yuhuai Wu, Saizheng Zhang, Ying Zhang, Yoshua Bengio, Ruslan R. Salakhutdinov
Yuhuai Wu, Saizheng Zhang, Ying Zhang, Yoshua Bengio, Ruslan R. Salakhutdinov
[number, block, many, level] [learning, general, achieves, show, almost, best] [gradient, term, size, large, extra, optimization] [can, order, also, second, one, first, formulation, following, product, vanilla, computational, matrix, note] [additive, test, two] [integration, model, information, time, simple, baseline, evaluated, state, effect, design] [multiplicative, rnn, building, neural, recurrent, compared, different, preprint, arxiv, using, hidden, performance, validation, bottom, table, use, training, bpc, character, speech, language, wxk, better, lstm, referred, rnns, including, lstms, initialized, final, task, top, outperforms, reported, gating, yoshua, figure, several, perform, without]
Selective inference for group-sparse linear models
Fan Yang, Rina Foygel Barber, Prateek Jain, John Lafferty
Fan Yang, Rina Foygel Barber, Prateek Jain, John Lafferty
[present] [set, interval, confidence, rate, hypothesis, case, lemma, theorem, let, give, since, coverage, setting, might, show] [inference, method, compute, size, develop, quadratic] [group, can, regression, truncated, linear, lasso, hard, also, projection, thresholding, iterative, sparse, subspace, see, main, lee, signal, first, vector, via] [selection, dirl, selected, given, fixed, testing, null, test, stepwise, distribution, selective, true, loftus, taylor, data, conditioning, result, procedure, section, normal, injury, density, percentile, two, death, physically, income, jonathan, interested, statistical, grouped] [model, forward, event, whether, response, information, write, design] [used, like, work, perform, using, region, without]
Quantum Perceptron Models
Ashish Kapoor, Nathan Wiebe, Krysta Svore
Ashish Kapoor, Nathan Wiebe, Krysta Svore
[probability, number, find, thus, misclassified, possible, either] [version, algorithm, theorem, learning, set, complexity, provide, margin, query, assume, bound, online, will, define, let, since, problem, now, mistake, may, finding, access, consider, case, make, follows, known, lemma, element, oracle, exists, proof] [machine, unitary] [vector, can, also, one, need, amplitude, following, first, computational, computation] [quantum, perceptron, space, classical, data, given, two, distribution, sample, marked, corresponding, statistical, point, separated, amplification, address, needed] [search, state, model, therefore, hyperplane, hyperplanes, new, scaling] [training, using, unit, feature, use, used, computer, traditional, figure, instead, preprint, arxiv]
Riemannian SVRG: Fast Stochastic Optimization on Riemannian Manifolds
Hongyi Zhang, Sashank J. Reddi, Suvrit Sra
Hongyi Zhang, Sashank J. Reddi, Suvrit Sra
[number] [theorem, convex, algorithm, complexity, problem, dependence, function, learning, corollary, analyze, fast, set, show, bounded, rate, defined, proof] [optimization, gradient, stochastic, svrg, convergence, geodesically, variance, nonconvex, computing, dominated, strongly, reduction, ifo, sectional, method, parallel, expx, rsvrg, siam, batch, key, inner, machine] [analysis, matrix, linear, global, vector, also, can, reduced, note, following, first, leading, solving, one, eigenvector, psd, symmetric, solution] [riemannian, manifold, exponential, curvature, geodesic, space, definite, nonlinear, transport, positive, two, euclidean, geometric, mean, tangent, journal, metric] [required, new, option] [map, using, accuracy, neural, use, shown, work, different, similar]
Fairness in Learning: Classic and Contextual Bandits
Matthew Joseph, Michael Kearns, Jamie H. Morgenstern, Aaron Roth
Matthew Joseph, Michael Kearns, Jamie H. Morgenstern, Aaron Roth
[probability, definition, number, many, attainable] [algorithm, fair, arm, regret, learning, bandit, kwik, contextual, fairness, bound, confidence, round, problem, set, active, will, classic, class, lemma, show, kwikt, optimal, prove, theorem, unknown, expected, least, lower, learner, setting, case, choose, context, function, interval, known, dependence, unfair, every, upper, study, now, let, play, implies, rjt, uti, general, constant, sti, give, may] [stochastic, machine, michael, initialize] [can, linear, technical, also, one, high, arg] [section, construct, distribution, polynomial, random, important] [reward, must, history, payoff, time, feedback] [without, use, prediction, sequence]
Swapout: Learning an ensemble of deep architectures
Saurabh Singh, Derek Hoiem, David Forsyth
Saurabh Singh, Derek Hoiem, David Forsyth
[number, block, average, represent, ensemble] [set, schedule, may, show, bernoulli, learning, general] [stochastic, inference, method, deterministic, batch, gradient, standard, performs, normalization, replace] [can, width, error, one, note, regularization] [random, corresponding, comparable, increasing, well, mean, equation, two] [model, form, experimental, early, others] [swapout, network, dropout, resnet, layer, residual, training, deep, use, trained, performance, depth, unit, table, using, convolutional, different, randomly, neural, wider, used, work, skip, better, output, outperforms, several, improves, similar, pooling, dropping, input, improve, outperform, relatively, recent]
A primal-dual method for conic constrained distributed optimization problems
Necdet Serhat Aybat, Erfan Yazdandoost Hamedani
Necdet Serhat Aybat, Erfan Yazdandoost Hamedani
[topology, graph, node, among, definition] [let, convex, algorithm, set, defined, rate, consider, theorem, dom, function, problem, define, optimal, conference, will, private, assume, svm, lemma] [distributed, consensus, optimization, convergence, xki, argmin, decentralized, pda, primal, constrained, iteration, computing, converges, saddle, computed, university, central, rmi, step, rki, angelia, iterate, machine] [can, following, local, ieee, suppose, denote, vector, also, connectivity, linear, min, assumption, solving, solution, note, one, signal, conic, global, matrix, computational] [given, point, denotes, section, journal, corresponding] [communication, dynamic, decision, information] [using, network, proposed, static, generated, sequence]
Minimax Estimation of Maximum Mean Discrepancy with Radial Kernels
Ilya O. Tolstikhin, Bharath K. Sriperumbudur, Prof. Bernhard Schölkopf
Ilya O. Tolstikhin, Bharath K. Sriperumbudur, Prof. Bernhard Schölkopf
[probability, present, theory, independence, fact] [minimax, theorem, lower, rate, bound, proof, set, get, optimal, dependence, let, learning, will, sup, conference, bounded, known, inequality, constant, inf, case, defined, class, exist, problem] [method, machine, distributed] [can, following, condition, note, optimality, need, main, second, first, order] [kernel, mmdk, estimation, based, gaussian, mmd, two, empirical, mean, result, radial, space, section, estimator, independent, universal, distribution, borel, shi, embedding, random, finite, measure, test, reproducing, dimensional, hilbert, sample, positive, journal, nonparametric, distance] [] [using, work, neural, used]
What Makes Objects Similar: A Unified Multi-Metric Learning Approach
Han-Jia Ye, De-Chuan Zhan, Xue-Min Si, Yuan Jiang, Zhi-Hua Zhou
Han-Jia Ye, De-Chuan Zhan, Xue-Min Si, Yuan Jiang, Zhi-Hua Zhou
[base, linkage, social, number, lml] [learning, set, instance, loss, function, convex, get, clearly, general, property, will] [latent, gradient, operator, datasets, supplementary, usually, proximal] [can, one, local, regularizer, global, unified, locality, related, ambiguous, overall, also, lovs] [metric, distance, based, data, specific, type, word, common, two] [framework, others, form] [different, semantic, similarity, multiple, mnn, similar, used, feature, single, learned, pair, semantics, lrgs, spatial, triplet, classification, discovering, table, dissimilar, performance, physical, various, pattern, hidden, figure, reflects, using, weak, approach, visualization]
A Bayesian method for reducing bias in neural representational similarity analysis
Mingbo Cai, Nicolas W. Schuck, Jonathan W. Pillow, Yael Niv
Mingbo Cai, Nicolas W. Schuck, Jonathan W. Pillow, Yael Niv
[structure, average, obtained] [assume, ratio, exists] [standard, bayesian, method, experiment, likelihood, variance, averaging] [matrix, can, noise, one, recovered, condition, also, analysis, imaging, low, diagonal, noisy, much] [covariance, bias, data, estimated, true, estimation, estimate, two, gaussian, corresponding, point, based, space] [rsa, fmri, representational, activity, fig, voxels, simulated, correlation, snr, design, directly, model, time, encoding, response, across, cognitive, temporal, princeton, cortex, brain, human, individual, measured, experimental, neuroscience, sensory, roi, autocorrelation, monkey, process] [similarity, neural, different, pattern, similar, visual, structured, task, using, used, map, spatial, higher, figure, learned, representation, applied, shared]
Learning in Games: Robustness of Fast Convergence
Dylan J. Foster, zhiyuan li, Thodoris Lykouris, Karthik Sridharan, Eva Tardos
Dylan J. Foster, zhiyuan li, Thodoris Lykouris, Karthik Sridharan, Eva Tardos
[number, probability, strong, social, theory, many, include] [regret, property, learning, cost, algorithm, hedge, price, fast, anarchy, bound, optimal, consider, satisfy, show, bandit, rate, realized, achieve, expected, even, shifting, setting, costi, satisfies, appendix, loss, utility, round, class, welfare, conference, online, cti, optimistic, will, repeated, turnover, function, best, close, minimization, rvu, implies, annual] [approximate, convergence, smooth, approximation, factor, expectation, full, requires, efficient, log] [low, can, small, proposition, satisfying, high, analysis, also] [population, section, converge, efficiency, result, well, smoothness] [game, player, action, information, feedback, dynamic, time, simple, new] [using, use, sequence, improve]
Optimal spectral transportation with application to music transcription
Rémi Flamary, Cédric Févotte, Nicolas Courty, Valentin Emiya
Rémi Flamary, Cédric Févotte, Nicolas Courty, Valentin Emiya
[fundamental, sum, number] [problem, cost, will, set, conference, optimal, every, may, defined, learning] [music, divergence, optimisation, method, fit, respect, energy] [spectral, matrix, frequency, harmonic, note, can, dictionary, transportation, ost, unmixing, oste, dirac, plca, transcription, cij, one, vector, small, musical, min, entry, transported, also, real, ostg, piano, local, first, explained, order, amplitude, pitch, ieee, decomposition] [section, data, measure, sample, given, particular, tij, based, distribution, metric, estimate, two] [template, time, new, form, placed, target, simple, model] [using, proposed, performance, consists, approach, international, frame, three, compared, used, without, audio, recognition]
Efficient Neural Codes under Metabolic Constraints
Zhuo Wang, Xue-Xin Wei, Alan A. Stocker, Daniel D. Lee
Zhuo Wang, Xue-Xin Wei, Alan A. Stocker, Daniel D. Lee
[constraint, monotonic, theory, david, department] [optimal, cost, function, case, problem, max, general, scheme, provide, active, rate, constant] [efficient, energy, university, due, detailed, processing, large] [noise, can, solution, magnitude, low, one, see, observed, linear] [mutual, curve, gaussian, distribution, two, fisher, uniform, population, mean, difference, limit, redundancy] [metabolic, response, coding, information, tuning, model, stimulus, neuron, code, ktotal, poisson, rmax, prior, range, dash, analytical, time, sensory, framework, limited, alan, simple, take, nature, assuming, advantage, substantial, early, focused, current, kthre] [neural, input, visual, single, different, pair, previous, natural, similar, using, red, available]
Multi-view Anomaly Detection via Robust Probabilistic Latent Variable Models
Tomoharu Iwata, Makoto Yamada
Tomoharu Iwata, Makoto Yamada
[detection, number, probabilistic, find, indicates, acm, represents, variable, obtained, probability, assignment] [instance, rate, since, precision, conference, private, set, every, assume] [latent, method, parameter, inference, likelihood, select, draw] [anomaly, auc, vector, can, view, projection, inconsistent, low, xnd, pcca, analysis, nth, anomalous, inconsistency, movie, ocsvm, matrix, noise, one, ieee, snd] [data, dimensionality, given, based, gaussian, two, mixture, estimated, joint, dirichlet, robust, space, canonical] [model, across, information, observation, process, assumes, infer, integrating, value, correlation] [proposed, different, using, used, generated, multiple, single, use, figure, performance, calculated, international, score, shared]
Eliciting Categorical Data for Optimal Aggregation
Chien-Ju Ho, Rafael Frongillo, Yiling Chen
Chien-Ju Ho, Rafael Frongillo, Yiling Chen
[interface, aggregation, partition, report, probability, number, scoring, elicit, eliciting, categorical, threshold, often, truthful, ppd, aggregating, elicitation, correctly, aggregate, knowledge, voting, payment, strictly] [optimal, conference, binary, proper, consider, problem, obtain, choose, setting, focus, learning, lemma, will, set, get, assume, drawn] [posterior, bayesian, method, full] [can, principal, also, one, observed, noisy, global, error] [data, sample, distribution, given, two, space, robust, theoretical] [information, agent, belief, model, design, maximize, prior, whether, answer, simple, goal, heuristic, cell, framework, count, bonus] [ground, prediction, figure, truth, question, work, single, task, predicting, using, use]
The non-convex Burer-Monteiro approach works on smooth semidefinite programs
Nicolas Boumal, Vlad Voroninski, Afonso Bandeira
Nicolas Boumal, Vlad Voroninski, Afonso Bandeira
[theory, constraint, number] [theorem, almost, cost, optimal, set, convex, problem, satisfy, bound, function, may, hold, general, show, bounded, feasible] [optimization, smooth, approximate, extreme, convergence, solve, university, size, method, quadratic, siam, step] [local, global, optimality, critical, rank, matrix, can, min, one, necessary, also, optimum, via, synchronization, globally, linear, burer, spurious, main, phase, sdp, hessf, note, rtr, computational, see, hessian] [point, riemannian, space, positive, journal, mathematical, two, result, tangent, manifold, robust, nonlinear, mathematics, important] [search, form] [compact, arxiv, preprint, using]
Approximate maximum entropy principles via Goemans-Williamson with applications to provable variational methods
Andrej Risteski, Yuanzhi Li
Andrej Risteski, Yuanzhi Li
[ising, partition, pairwise, ferromagnetic, graph, degree, many, fact, programming, science] [will, theorem, relaxation, function, case, constant, prove, convex, get, since, consider, provide, general, known, bound, lemma, implies, define, show, hardness, problem, upper, wish, satisfies, annual, max, learning] [variational, factor, approximation, log, optimization, due, approximate, calculating, machine, efficiently, term, standard] [can, rounding, one, following, provable, matrix, note, first, paper, main, symposium, computational] [entropy, distribution, maximum, exponential, principle, mean, given, based, section, comparable, exp, provides, family, journal, covariance, functionals] [information, value, model, time, optimizing] [multiplicative, work, produce, using, previous]
Deep Neural Networks with Inexact Matching for Person Re-Identification
Arulkumar Subramaniam, Moitreya Chatterjee, Anurag Mittal
Arulkumar Subramaniam, Moitreya Chatterjee, Anurag Mittal
[normalized, connected, partial, false, number] [learning, conference, set, every] [inexact, large, challenging] [ieee, also, first, one, small, existing, second] [two, metric, fused, test, given, space, corresponding] [model, search, correlation, grid, across] [person, matching, image, training, deep, feature, ahmed, relu, computer, layer, dataset, wider, conv, different, pattern, neural, performance, vision, input, convolution, maxpool, similarity, multiple, international, qmul, representation, architecture, output, network, fully, using, use, table, propose, viewpoint, proposed, work, learn, gallery, used, similar, labeled, pair, illumination]
A Powerful Generative Model Using Random Weights for the Deep Image Representation
Kun He, Yan Wang, John Hopcroft
Kun He, Yan Wang, John Hopcroft
[chinese, create, quality, purely, science, weighting] [loss, may, even, lower] [gradient, select, method, argmin, university] [high, first, row, following, success, noise] [random, reference, based, well] [new, prior, model, white] [image, deep, using, vgg, style, network, ranvgg, texture, trained, neural, representation, visualization, convolutional, content, weight, untrained, feature, layer, figure, gatys, pretrained, natural, use, work, original, artistic, reconstruction, inversion, several, different, training, input, architecture, higher, pooling, slightly, compare, perceptual, cnn, similar, better, transfer, understanding, synthesize, inverting, reconstruct, synthesized, impact, understand, generative, activation, alexnet, three, compared, synthesis, mahendran, starry]
Optimistic Bandit Convex Optimization
Scott Yang, Mehryar Mohri
Scott Yang, Mehryar Mohri
[average, structural] [loss, regret, convex, let, algorithm, bound, fbt, will, bandit, assume, lipschitz, lemma, smoothed, gbt, ftrl, barrier, function, ptimistic, set, regt, bounded, play, since, bco, always, learner, theorem, online, dekel, admits, kxt, cost, optimistic, round, surrogate, saha, best, improved, arbitrary, idea, rft, flaxman, feasible, get, admit, known, learning, subgradient] [gradient, optimization, key, stochastic, variance, around, showed, factor] [can, following, denote, one, also, analysis, dimension] [point, estimate, computationally, result, polynomial, given] [time, action, averaged, information, predictable, upon, new, closed] [use, sequence, work, predicting, original, using, prediction]
A Probabilistic Programming Approach To Probabilistic Data Analysis
Feras Saad, Vikash K. Mansinghka
Feras Saad, Vikash K. Mansinghka
[probabilistic, cgpms, cgpm, programming, logpdf, simulate, nan, member, bayesdb, create, zrk, satellite, composable, directed, cluster, orbital, metamodeling, variable, purely, override, crosscat, program, categorical, launch, law, panel, platform, subset, acyclic, ample] [return, query, learning, assume, composite, may, algorithm, define, function, set, general] [latent, importance, bayesian, log, machine, inference, likelihood, sampling, latents] [analysis, can, computational, paper, missing] [data, statistical, population, joint, section, sample, random, density, distribution, numerical, type, well, given, multivariate] [model, trace, baseline, infer, new, evidence, code] [using, generative, input, output, figure, table, discriminative, language, network, weight, used, generate]
Learning under uncertainty: a comparison between R-W and Bayesian approach
He Huang, Martin Paulus
He Huang, Martin Paulus
[stable, probability, thus, likely, contingency, influence] [learning, rate, optimal, will, function, lower, chosen, expected, always] [bayesian, parameter] [low, optimality, high, can, condition, also, frequency, significantly, linear] [estimation, based, estimated, positive, consistent, two, data, maximum] [model, volatility, belief, decision, reward, dbm, expert, stationarity, environmental, choice, inverted, behavioral, lose, poor, change, wsls, option, simulation, fig, shift, simulated, behavior, value, correlation, search, examine, explaining, differ, rewarded, across, target, environment, significant, information, positively, current] [three, different, shown, using, figure, shape, use, accuracy, visual, approach, higher, relationship, task, used]
How Deep is the Feature Analysis underlying Rapid Visual Categorization?
Sven Eberhardt, Jonah G. Cader, Thomas Serre
Sven Eberhardt, Jonah G. Cader, Thomas Serre
[number, correctly, possible, total] [complexity, may, confidence, set, best, fast, considered, function] [processing, experiment, computed] [also, analysis, computational] [data, underlying, well, maximum, corresponding, boundary] [human, model, correlation, decision, animal, individual, time, response, behavioral, issn, longer, found, experimental, answer, presentation, correct, hierarchical, timing, stimulus, psychophysics, whether, past] [visual, accuracy, deep, categorization, rapid, depth, shown, figure, image, intermediate, used, recognition, object, using, layer, higher, neural, convolutional, deeper, different, recent, vision, feature, natural, imagenet, fixation, performance, computer, classification, increased, cross, top, work, paradigm]
Optimal Architectures in a Solvable Model of Deep Networks
Jonathan Kadmon, Haim Sompolinsky
Jonathan Kadmon, Haim Sompolinsky
[number, rule, recursion, wide, level, cluster, total, average] [optimal, learning, will, function, assume, show, may, contribution, binary, set, consider, simplicity] [processing, initial, size, variance, due, large, central, convergence] [error, noise, noisy, noiseless, one, sparsity, load, can, lopt, linear, also, matrix, signal] [fixed, mean, two, limit, finite, random, infinitely, infinite] [model, neuron, sensory, information, simple, activity, system, behavior, state, time] [layer, readout, deep, input, network, neural, single, overlap, different, performance, intermediate, perform, architecture, figure, field, classification, using, feedforward, use, understanding, depth, andrew, training, pattern, seen]
Coresets for Scalable Bayesian Logistic Regression
Jonathan Huggins, Trevor Campbell, Tamara Broderick
Jonathan Huggins, Trevor Campbell, Tamara Broderick
[subset, number, clustering, thus, obtained, expect, probability, many] [algorithm, bound, let, upper, lemma, streaming, setting, theorem, may, show, conference, expected, since, constant, lower, provide, smaller, idea] [bayesian, inference, logistic, posterior, mcmc, variational, parameter, size, large, ebspam, scalable, approximation, distributed, likelihood, datasets, full, standard, sampling, monte, machine, processing, langevin, develop] [small, regression, can, synthetic, also, proposition, computational, one, via] [data, coreset, sensitivity, coresets, random, subsampling, mean, construction, given, inary, maximum, bayes, indep, section, important] [time, prior, choice, artificial, model, demonstrate, information] [used, dataset, using, use, performance, approach]
Sampling for Bayesian Program Learning
Kevin Ellis, Armando Solar-Lezama, Josh Tenenbaum
Kevin Ellis, Armando Solar-Lezama, Josh Tenenbaum
[program, ample, rogram, edit, list, sat, probability, number, solver, constraint, manipulation, counting, string, tilt, programming, often, many, structure, either, recursive, editing, parity, assignment, thus, probabilistic, ashish, held] [learning, set, problem, algorithm, uniformly, learner, upper, satisfy, will, least, lower] [sampling, posterior, approximate, log, processing, bayesian] [can, sketch, also, small, one, see, much, arbitrarily] [sample, random, distribution, space, length, consistent, data] [model, new, correct, search, time, decision, take, framework, prior, goal, write] [figure, text, approach, work, synthesis, like, using, use, neural, learn, description, used, different, synthesizing, training]
Bayesian optimization under mixed constraints with a slack-variable augmented Lagrangian
Victor Picheny, Robert B. Gramacy, Stefan Wild, Sebastien Le Digabel
Victor Picheny, Robert B. Gramacy, Stefan Wild, Sebastien Le Digabel
[constraint, valid, wmin, panel, many, involving, variable] [inequality, problem, optimal, function, show, improvement, may, best, set, expected, let, surrogate, observe, unknown, known, version] [optimization, slack, constrained, mixed, objective, blackbox, bayesian, method, albo, supplementary, progress, augmented, involves, lagrangian, implementation, step, efi, fmin, initial, comparators, ycj, nlopt, monte, standard] [can, global, via, one, local, plot, formulation, first, solution, see, min, solving] [random, distribution, two, statistical, gaussian, mean, section, provided, proportion, mathematical, numerical, empirical] [equality, search, new, predictive, closed, literature, value, unconstrained, design, acquisition, modeling] [original, figure, several, work, input, shown]
Composing graphical models with neural networks for structured representations and fast inference
Matthew Johnson, David K. Duvenaud, Alex Wiltschko, Ryan P. Adams, Sandeep R. Datta
Matthew Johnson, David K. Duvenaud, Alex Wiltschko, Ryan P. Adams, Sandeep R. Datta
[graphical, structure, variable, probabilistic, represent] [learning, general, consider, algorithm, bound, provide, max, function] [variational, latent, inference, svae, conjugate, gradient, objective, log, iid, compute, lsvae, respect, efficient, fit, factor, stochastic, conjugacy, optimization, flexible, approximate] [can, linear, also, see, local, mouse] [family, exponential, data, discrete, gaussian, mean, nonlinear, section, gmm, two, manifold, density, estimate] [model, observation, modeling, dynamical, lds, state, prior, rich, framework, system, continuous, behavior] [natural, neural, structured, recognition, field, image, video, using, used, use, network, figure, generative, speech, depth, autoencoders, approach, deep, learn, preprint, arxiv]
Safe Policy Improvement by Minimizing Robust Baseline Regret
Mohammad Ghavamzadeh, Marek Petrik, Yinlam Chow
Mohammad Ghavamzadeh, Marek Petrik, Yinlam Chow
[probability, number, represents, represent] [problem, algorithm, improvement, optimal, regret, return, may, set, theorem, function, show, appendix, example, loss, uncertain, analyze, max, randomized, even, bound, will, guarantee, conference, improved, consider, minimizing] [optimization, approximate, deterministic, standard, energy, due, batch, computing] [error, can, solution, solving, min, following, arg, formulation, main, guaranteed, significantly, note, one, also, good, high] [robust, section, given, true, data, based, space] [policy, baseline, model, uncertainty, transition, state, mdp, safe, markov, decision, reward, action, arbitrage, simple, conservative, directly] [performance, approach, use, using, better, figure, learned, propose, used]
Examples are not enough, learn to criticize! Criticism for Interpretability
Been Kim, Oluwasanmi O. Koyejo, Rajiv Khanna
Been Kim, Oluwasanmi O. Koyejo, Rajiv Khanna
[number, subset, represent, present] [function, submodular, set, let, greedy, consider, known, may, algorithm, study, theorem, learning, corollary, case, submodularity, cost, max, selecting, problem, class, show, example, selects] [bayesian, optimization, respect, large] [condition, can, also, diagonal, matrix, local, linear, following, denote] [criticism, kernel, prototype, data, given, selection, interpretability, nearest, proto, two, measure, selected, well, witness, important, positive, sample, useful, mmd, maximum, discrepancy, space, together] [model, human, complex, raw, framework, subject, interpretable] [using, performance, dataset, image, compared, approach, improve, use, work, similarity, applied, used, shown]
A Locally Adaptive Normal Distribution
Georgios Arvanitidis, Lars K. Hansen, Søren Hauberg
Georgios Arvanitidis, Lars K. Hansen, Søren Hauberg
[cluster, structure, number, probabilistic, find, clustering, denoted, definition, negative] [consider, learning, adaptive, algorithm, least, defined, function, define, problem, near, constant, changing, conference, known, since, scheme] [likelihood, standard, compute, gradient, normalization, monte, carlo, machine] [can, linear, local, matrix, synthetic, first, principal, low, analysis] [land, riemannian, distribution, data, manifold, mean, normal, metric, mixture, distance, covariance, maximum, sleep, locally, gaussian, space, nonlinear, density, point, tangent, geodesic, measure, logarithm, gmm, eeg, corresponding, exp, euclidean, entropy, given, logx, estimate, kernel, estimation] [model, intrinsic, subject] [using, use, learned, figure, smoothly, generative, work, better, neural]
One-vs-Each Approximation to Softmax for Scalable Estimation of Probabilities
Michalis Titsias RC AUEB
Michalis Titsias RC AUEB
[probability, obtained, categorical, sum, notice, number, pairwise, probabilistic] [bound, lower, remaining, since, function, set, learning, consider, multiclass, class, cost, conference, maximizing, label, wish, will] [log, stochastic, likelihood, ove, approximate, large, efk, variational, method, parameter, sgd, gradient, optimization, bouchard, approximation, maximized, scalable, efm, size, softmaxk, subsample, bibtex, full] [exact, can, also, one, following, small, error, product] [data, estimation, based, maximum, estimated, associated, subsampling, section, test, denotes] [value, form, model, evolution, maximize, simple] [softmax, training, classification, figure, neural, score, scale, single, language, using, several, input, shown, dataset]
Sub-sampled Newton Methods with Non-uniform Sampling
Peng Xu, Jiyan Yang, Farbod Roosta-Khorasani, Christopher Ré, Michael W. Mahoney
Peng Xu, Jiyan Yang, Farbod Roosta-Khorasani, Christopher Ré, Michael W. Mahoney
[block, partial, number, solver, among] [algorithm, complexity, problem, learning, convex, defined, lemma, randomized, achieve, dependence, scheme, assume, consider, show, least] [sampling, newton, convergence, approximation, machine, size, iteration, method, standard, logistic, approximate, optimization, log, augmented, michael] [leverage, matrix, can, ssn, condition, hessian, local, norm, linear, solution, see, following, error, order, one, tsolve, running, plevss, rnormss, exact, tconst, second] [uniform, section, given, well, two, ridge, distribution, construct, independent, based, locally] [time, form, information] [used, table, using, different, score, work, figure, preprint, arxiv, better, use, shown, per, performance]
Guided Policy Search via Approximate Mirror Descent
William H. Montgomery, Sergey Levine
William H. Montgomery, Sergey Levine
[constraint, manipulation, obtained, either] [learning, cost, algorithm, bound, since, show, conference, case, appendix, surrogate, unknown, convex, provide] [step, method, size, mirror, descent, dkl, optimization, gradient, convergence, initial, requires, dual, sampling, approximate, variant, iteration, end] [local, can, global, also, linear, small, projection, analysis, arg, following] [nonlinear, section, based] [policy, search, guided, prior, mdgps, peg, trajectory, state, epi, badmm, new, typically, simple, directly, form, adjustment, corresponds, reinforcement, optimize, lqr, complex, robotic, control, hole] [using, supervised, use, neural, work, deep, performance, task, train, used, shown, different, international, better, per]
Diffusion-Convolutional Neural Networks
James Atwood, Don Towsley
James Atwood, Don Towsley
[graph, node, dcnns, diffusion, cora, pubmed, relational, probabilistic, present, structure, breadth, mutag, graphical, definition, wjk, represented, ked, citation, either] [learning, set, label, provide, defined, function] [standard] [can, matrix, also, one, tensor, via, power, paper] [kernel, data, exponential, section, well, conditional, proportion, test, mean, curve, basis] [model, search, information, baseline, offer] [classification, accuracy, dcnn, performance, representation, neural, several, training, input, work, figure, prediction, validation, network, learned, used, convolutional, table, effective, including, consists, using, applied, deep, labeled, dataset, implemented, investigate, predict]
Solving Marginal MAP Problems with NP Oracles and Parity Constraints
Yexiang Xue, zhiyuan li, Stefano Ermon, Carla P. Gomes, Bart Selman
Yexiang Xue, zhiyuan li, Stefano Ermon, Carla P. Gomes, Bart Selman
[number, probability, saa, lbp, sum, represent, counting, probabilistic, strength, parity, coupling, xor, weighted, carla, pairwise, cascade, replicates, solver, find, node] [algorithm, problem, set, max, function, will, conference, theorem, return, bound, let, case, show, learning, upper, binary, lemma, optimal, since, alexander] [marginal, mixed, optimization, log, inference, approximation, intractable, machine, approximate, solve] [can, one, suppose, solution, first, also, exact, solving] [two, independent, sample, fixed, provides, result, well, random, selected] [mmap, found, median, information, time, value, belief, artificial] [map, figure, different, network, randomly, approach, use, performance, neural, original, proposed, international]
Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings
Tolga Bolukbasi, Kai-Wei Chang, James Y. Zou, Venkatesh Saligrama, Adam T. Kalai
Tolga Bolukbasi, Kai-Wei Chang, James Y. Zou, Venkatesh Saligrama, Adam T. Kalai
[fact, many, crowd, number, seed, present, identify, find, quality, normalized] [set, learning, algorithm, will, appendix, show, since, consider, define, might, may] [machine, parameter, step, standard] [also, can, one, vector, subspace, onto, projection, first, much, significantly] [gender, word, embedding, bias, embeddings, two, neutral, analogy, specific, grandfather, closer, given, debiasing, grandmother, female, well, woman, male, news, useful, section, preserving, described] [direction, indirect, implicit, whether, direct, across, exhibit] [figure, pair, use, used, language, original, different, unit, similar, generated, evaluate, man, computer, using, trained, shown, ten, classifier, natural, reducing]
The Limits of Learning with Missing Data
Brian Bullins, Elad Hazan, Tomer Koren
Brian Bullins, Elad Hazan, Tomer Koren
[number, attribute, probability, enough, attainable] [let, loss, algorithm, lower, learning, theorem, observe, absolute, set, learner, regressor, proof, show, prove, general, setting, may, example, precision, bound, exists, since, function, subgradient, conference, hinge, lemma, hazan, expected, lao, convex, consider, arbitrary, follows, koren, achieve, choose, always, complexity] [machine, supplementary, efficient] [regression, can, missing, linear, one, main, first, see, certain, arbitrarily, error, low, whereby, min, hard, gap, vector, necessary] [distribution, squared, two, data, given, result, sample, ridge, limit] [limited, information, determine, observation, framework] [classification, training, work, output, shown, similar, international, used, per]
Leveraging Sparsity for Efficient Submodular Data Summarization
Erik Lindgren, Shanshan Wu, Alexandros G. Dimakis
Erik Lindgren, Shanshan Wu, Alexandros G. Dimakis
[graph, runtime, number, threshold, subset, probability, representative] [set, greedy, algorithm, problem, function, submodular, sparsified, optimal, element, benefit, facility, sparsification, since, lsh, consider, quickly, personalized, covering, get, will, sensitive, instance, constant, learning, summarization, show, chosen, guarantee, theorem, lower, fast, smaller, lazy] [approximation, size, log, stochastic, method, large, approximate, optimization, factor, datasets, parameter] [matrix, can, one, solution, see, sparsity, pagerank, much, exact, also, small, largest, dot, product, locality, analysis, following, movielens] [neighbor, nearest, data, random, sample] [value, location, time, take] [using, use, similarity, figure, entire, work, used, feature, top, better]
A Comprehensive Linear Speedup Analysis for Asynchronous Stochastic Parallel Optimization from Zeroth-Order to First-Order
Xiangru Lian, Huan Zhang, Cho-Jui Hsieh, Yijun Huang, Ji Liu
Xiangru Lian, Huan Zhang, Cho-Jui Hsieh, Yijun Huang, Ji Liu
[node, child, number] [algorithm, rate, set, function, convex, learning, bound, upper, bounded, constant, may, let, best, theorem, show, special, define, agarwal, since] [asynchronous, stochastic, speedup, gradient, parallel, convergence, sgd, central, scd, descent, asgd, parameter, nonconvex, optimization, zeroth, coordinate, aszd, ascd, method, variance, music, machine, objective, distributed, large, iteration, blending, comprehensive, requirement, black] [linear, analysis, can, following, first, order, generic, running, existing, proved, note, one, component, also, much, matrix, ensure] [data, test, result, two, consistent, provides, random, estimate] [model, value, read, time] [different, using, neural, network, table, single, novel, use, computer, deep, including]
On Robustness of Kernel Clustering
Bowei Yan, Purnamrita Sarkar
Bowei Yan, Purnamrita Sarkar
[clustering, number, cluster, consistency, present, misclassified] [algorithm, theorem, show, lemma, will, upper, relaxation, proof, assume, bound, let, constant, weakly, rate, max, problem, now, learning, appendix, get, bounded, arbitrary, function] [log, machine, strongly] [matrix, sdp, analysis, one, eigenvectors, also, spectral, can, main, semidefinite, error, first, zero, norm, high, outlier, misclustered, eigenvalue, largest, fraction, singular, following, svd, blockwise] [kernel, two, data, consistent, based, robust, robustness, gaussian, section, separation, distance, population, inliers, result, random, corresponding, inlier] [model, value] [used, top, different, accuracy, performance, shown, use, proposed, without]
Binarized Neural Networks
Itay Hubara, Matthieu Courbariaux, Daniel Soudry, Ran El-Yaniv, Yoshua Bengio
Itay Hubara, Matthieu Courbariaux, Daniel Soudry, Ran El-Yaniv, Yoshua Bengio
[number, arithmetic, many, possible, either, multiplication] [binary, learning, algorithm, precision, function] [full, gradient, size, faster, reduce, stochastic, energy, method, computing, parameter, processing, batch, consumption, bnn, run, lead, normalization, machine] [can, first, also, matrix, noise, power, running, high] [kernel, two, section, test, point, based, estimator, quantization, dedicated] [forward, research, time, specifies] [neural, using, training, deep, binarized, memory, gpu, bnns, binarization, accuracy, arxiv, use, weight, work, performance, preprint, network, previous, imagenet, achieved, layer, without, hardware, updated, mnist, impact, mlp, svhn, trained, classification, drastically, last]
A Probabilistic Model of Social Decision Making based on Reward Maximization
Koosha Khalvati, Seongmin A. Park, Jean-Claude Dreher, Rajesh P. Rao
Koosha Khalvati, Seongmin A. Park, Jean-Claude Dreher, Rajesh P. Rao
[social, number, probability, average, probabilistic, larger, among, total] [round, function, expected, public, optimal, set, playing, contribution, outcome, make, maximizing, best] [beta, initial, fit] [group, error, one, can, also, first, partially, much] [based, two, distribution, data, space, given, observable] [model, reward, player, belief, state, decision, pomdp, game, action, prior, descriptive, making, value, mdp, individual, pgg, behavior, cooperation, brain, subject, human, cooperativeness, trial, framework, dilemma, markov, found, others, modeled, cognitive, next, transition, current, behavioral, modeling, fmri, activity, left] [neural, using, previous, different, predicted, shown, figure, computer, use]
The Robustness of Estimator Composition
Pingfan Tang, Jeff M. Phillips
Pingfan Tang, Jeff M. Phillips
[definition, many, ith, either, greater, number] [set, composite, define, composition, theorem, make, since, may, will, even, defined, obtain, ranked, get, know, consider, implies, give, example] [gradient, machine] [can, analysis, suppose, need, first, order, remark, main, real, much, anomalous, following, condition, onto] [breakdown, estimator, point, data, pflat, robust, given, robustness, percentile, asymptotic, result, two, finite, common, formal, sample, wingspread, section, company, important, space, changed] [median, new, value, target, change, move, individual, another, process, raw] [use, person, without, single, several, table, different, understanding]
Higher-Order Factorization Machines
Mathieu Blondel, Akinori Fujino, Naonori Ueda, Masakazu Ishihata
Mathieu Blondel, Akinori Fujino, Naonori Ueda, Masakazu Ishihata
[degree, number, recursion, negative, evaluating, distinct] [algorithm, learning, cost, let, defined, since, function, minimizing, even, set, conference, define] [hofms, anova, computing, gradient, compute, hofm, machine, efficient, objective, adagrad, coordinate, although, descent, stochastic, reduce, end] [can, factorization, also, main, order, link, need, denote, similarly, regression, sparse] [kernel, polynomial, section, data, two, positive, well] [model, new, therefore, time, advantage, simply, directly] [training, using, feature, use, table, prediction, learn, approach, per, network, propose, instead, shared, proposed, compared, different, dataset, neural]
Wasserstein Training of Restricted Boltzmann Machines
Grégoire Montavon, Klaus-Robert Müller, Marco Cuturi
Grégoire Montavon, Klaus-Robert Müller, Marco Cuturi
[probability, possible, obtained] [set, binary, learning, optimal, expected, consider, defined, smoothing, conference, function, best, let] [gradient, parameter, standard, dual, machine, size, large, variance, respect, term, divergence, approximation] [can, completion, one, restricted, observed, also, small, low, high, first, vector, noise, error] [wasserstein, distance, data, rbm, boltzmann, distribution, sample, metric, two, hamming, empirical, given, practical, equation, kde, bias, true, pcd, euclidean, explanatory, kantorovich, point] [model, state, simple, effect] [figure, training, generated, denoising, learned, image, neural, using, shown, used, original, performance, work, different]
Stochastic Multiple Choice Learning for Training Diverse Deep Ensembles
Stefan Lee, Senthil Purushwalkam Shiva Prakash, Michael Cogswell, Viresh Ranjan, David Crandall, Dhruv Batra
Stefan Lee, Senthil Purushwalkam Shiva Prakash, Michael Cogswell, Viresh Ranjan, David Crandall, Dhruv Batra
[smcl, ensemble, diverse, mcl, member, often, standing, riding, perched, many, present, specialization, tree] [oracle, loss, learning, conference, example, set, class, algorithm, provide, show, expected] [stochastic, size, gradient, standard, method, descent] [can, error, existing, ieee, one, group, also] [based, classical, people] [model, choice, research, search, producing, information] [multiple, training, deep, iou, figure, trained, image, produce, network, vision, top, task, single, neural, bird, computer, segmentation, train, performance, approach, using, beam, sitting, work, dey, different, classification, net, wave, convolutional, man, predicted, structured, output, accuracy, semantic, international, cnn, shown, captioning, dataset]
Fast Active Set Methods for Online Spike Inference from Calcium Imaging
Johannes Friedrich, Liam Paninski
Johannes Friedrich, Liam Paninski
[constraint, obtained, gurobi, possible, homogeneous] [algorithm, set, active, online, fast, problem, convex, function, will] [method, inference, optimization, step, run, parameter, supplementary, updating, update, warm, constrained] [calcium, imaging, fluorescence, oasis, can, one, solution, noise, nonnegative, also, order, sparsity, need, interior, signal, nat, formulation, running, violation, simultaneous, magnitude, autocovariance, sparse, high, isotonic] [data, estimate, equation, true, point, well, based] [time, spike, pool, activity, value, process, forward, model, whereas, move, simulated, research, optimizing, trace, spiking, series, track, new] [neural, figure, deconvolution, updated, compared, train, using, back, approach, used]
General Tensor Spectral Co-clustering for Higher-Order Data
Tao Wu, Austin R. Benson, David F. Gleich
Tao Wu, Austin R. Benson, David F. Gleich
[clustering, graph, probability, walk, cluster, number, partition, partitioning, find, represents, cut] [algorithm, will, set, case, chosen, call, general, let, may, show, appendix] [method, stochastic, standard, large, compute, datasets] [tensor, spectral, conductance, can, matrix, group, vector, following, eigenvector, square, one, gtsc, fiedler, surfer, sweep, rectangular, tsc, also, first, planted, sparse, spacey, multilinear, nonnegative, flight, solution] [random, data, chain, two, distribution, biased, stationary, section, based, procedure] [markov, transition, state, process, new, stop, form, information, english, model, time] [use, using, generalize, work, figure, several]
Maximal Sparsity with Deep Networks?
Bo Xin, Yizhou Wang, Wen Gao, David Wipf, Baoyuan Wang
Bo Xin, Yizhou Wang, Wen Gao, David Wipf, Baoyuan Wang
[many, structure, number, likely, probability, allow, degree] [will, may, learning, problem, optimal, consider, constant, function, algorithm, drawn, even, least, loss, set, uniformly, general] [iid, operator] [can, sparse, iht, support, dictionary, recovery, existing, iterative, maximally, linear, matrix, via, ieee, following, photometric, rip, nonzero, one, much, also, computational, signal, success, quite, whereby, coherent] [estimation, given, practical, fixed, data, distribution, section, described, true] [correlation, model, information, across] [network, training, deep, using, use, surface, layer, original, different, dnn, classification, figure, performance, lighting, lstm, unit, viewed, arxiv, residual, neural, preprint]
Stochastic Three-Composite Convex Minimization
Alp Yurtsever, Bang Cong Vu, Volkan Cevher
Alp Yurtsever, Bang Cong Vu, Volkan Cevher
[strong, present, average, variable, basic, denoted, probability, indicator] [algorithm, learning, convex, rate, problem, function, minimization, lipschitz, set, monotone, theorem, composite, minimize, almost, hold, best, assume, consider, uniformly, constant, define, optimal] [stochastic, method, gradient, splitting, optimization, convergence, proximal, deterministic, operator, smooth, processing, parameter, compute, standard, portfolio, expectation, strongly, unbiased, end, machine, markowitz] [can, remark, restricted, convexity, vector, see, solving, note, suppose, projection, solution, denote, following, one] [random, given, section, point, numerical, positive, empirical, squared, two] [template, information, framework, continuous, evidence, subject] [sequence, use, using, figure, neural]
A Sparse Interactive Model for Matrix Completion with Side Information
Jin Lu, Guannan Liang, Jiangwen Sun, Jinbo Bi
Jin Lu, Guannan Liang, Jiangwen Sun, Jinbo Bi
[interaction, unique] [algorithm, rate, problem, complexity, adaptive, theorem, max, let, conference, may, convex] [method, full, log, sampling, convergence, latent, machine, optimization, parameter, standard, datasets, solve] [matrix, side, can, rank, recovery, completion, low, row, observed, missing, interactive, ladmm, sparse, rmse, condition, exact, linear, synthetic, analysis, imc, singular, corrupted, solution, norm, column, ieee, sufficient, nuclear, dirtyimc, high, require, first, min] [two, sample, data, theoretical, given, inductive, space, empirical, journal, true] [model, information, new] [feature, use, approach, randomly, figure, performance, proposed, three, compared, used, recent, without, different, generated, international]
MetaGrad: Multiple Learning Rates in Online Learning
Tim van Erven, Wouter M. Koolen
Tim van Erven, Wouter M. Koolen
[master, theory] [learning, regret, convex, online, metagrad, algorithm, loss, theorem, bound, version, rate, surrogate, adaptive, fast, may, slave, rft, appendix, vtu, logarithmic, consider, general, function, offline, annual, will, let, depends, implies, bounded, get, satisfy, easier, kgt, hinge, bernstein, guarantee, koolen] [full, machine, stochastic, gradient, method, step, parameter, optimization, strongly, kuk] [can, diagonal, also, condition, analysis, running, related, one, main, first, optimum, projection, matrix] [section, data, point, exponential, two, based, fixed, covariance, corresponding] [grid, van, time, therefore, prior, simultaneously, expert, direction] [prediction, using, like, domain, work]
Attend, Infer, Repeat: Fast Scene Understanding with Generative Models
S. M. Ali Eslami, Nicolas Heess, Theophane Weber, Yuval Tassa, David Szepesvari, koray kavukcuoglu, Geoffrey E. Hinton
S. M. Ali Eslami, Nicolas Heess, Theophane Weber, Yuval Tassa, David Szepesvari, koray kavukcuoglu, Geoffrey E. Hinton
[number, variable, probabilistic, structure] [learning, show, will, consider, every, appear] [inference, air, latent, gradient, draw, likelihood, variational, posterior, dair, however, amortized, log] [can, one, also, via, first, note] [data, two, specified, well, given, random, distribution] [model, framework, learns, form, interpretable, therefore, inferred, demonstrate, continuous, infer, inverse, policy, count, prior] [network, scene, generative, using, object, neural, image, training, zpres, deep, recurrent, use, learn, used, different, structured, figure, pose, trained, attention, multiple, produce, task, alex, single, learned, approach, accuracy, supervised, unsupervised, koray, perform, geoffrey, identity]
Dimension-Free Iteration Complexity of Finite Sum Optimization Problems
Yossi Arjevani, Ohad Shamir
Yossi Arjevani, Ohad Shamir
[number, degree, whose, definition, many, sum] [lower, bound, oracle, complexity, problem, convex, algorithm, function, theorem, rate, class, learning, bounded, derive, setting, set, tight, may, scheme, attained, proof, minimization] [optimization, iteration, stochastic, oblivious, convergence, iterates, gradient, machine, sdca, accelerated, dual, coordinate, proximal, cli, depend, deterministic, method, smooth, fsm, descent, arjevani, parameter, sampling, rlm, equipped, processing, approximation, ohad] [can, linear, following, one, min, computational, first, also] [given, section, finite, point, well, based, multivariate] [form, information, new, framework, current] [approach, using, neural, without, different, preprint, arxiv]
An Efficient Streaming Algorithm for the Submodular Cover Problem
Ashkan Norouzi-Fard, Abbas Bazzi, Ilija Bogunovic, Marwa El Halabi, Ya-Ping Hsieh, Volkan Cevher
Ashkan Norouzi-Fard, Abbas Bazzi, Ilija Bogunovic, Marwa El Halabi, Ya-Ping Hsieh, Volkan Cevher
[number, graph, vertex, acm, subset, find, possible, representative] [algorithm, set, streaming, cover, submodular, greedy, offline, problem, utility, oracle, theorem, active, dominating, smallest, optimal, consider, lazy, function, studied, cost, conference, massive, achieve, least, ratio, lower, bicriteria, tight, may, defined, element, best] [size, approximation, large, pass, requires, efficient, log, factor, distributed, machine, select, optimization, run, designed] [can, solution, ssc, fraction, small, one, first, phase, also, note, certain] [data, given, selection, random, section, corresponding, kernel, two, selected] [goal] [memory, single, performance, international, using, dataset, stream, figure]
Matching Networks for One Shot Learning
Oriol Vinyals, Charles Blundell, Tim Lillicrap, koray kavukcuoglu, Daan Wierstra
Oriol Vinyals, Charles Blundell, Tim Lillicrap, koray kavukcuoglu, Daan Wierstra
[many] [set, learning, class, function, defined, label, example, make] [batch, large, full, ets] [support, can, also, one, much, following, related, note] [data, test, given, section, metric, based, embedding, nearest, two, distribution, word, well] [model, new, baseline, form, modeling, simple] [matching, training, neural, task, cosine, imagenet, deep, trained, language, used, work, fine, attention, network, omniglot, inception, classifier, memory, classification, sequence, use, image, arxiv, preprint, convolutional, table, performance, using, lassifier, atching, sentence, recent, lstm, like, setup, single, google, softmax, siamese, accuracy, classify, vision]
Pairwise Choice Markov Chains
Stephen Ragain, Johan Ugander
Stephen Ragain, Johan Ugander
[probability, pairwise, thus, partition, theory, definition, number, transitivity] [set, let, rate, show, implies, property, regularity, case, special, known, choosing, chosen, least] [stochastic, logit, datasets, nested, supplementary, inference, likelihood] [matrix, one, can, qij, also, proposition, error, see] [uniform, axiom, discrete, chain, data, empirical, distribution, expansion, stationary, independent, selection, multinomial, random, two, well, maximum, strict, journal, given] [model, choice, pcmc, mnl, markov, contractible, exhibit, qji, mmnl, transition, sfwork, btl, time, elimination, defines, rum, ttest, closed, communicating, yellott, sfshop] [work, learned, figure, recent, using, train, different]
Safe Exploration in Finite Markov Decision Processes with Gaussian Processes
Matteo Turchetta, Felix Berkenkamp, Andreas Krause
Matteo Turchetta, Felix Berkenkamp, Andreas Krause
[probability, constraint, path, say] [lemma, set, algorithm, assume, know, function, least, consider, learning, unknown, may, since, confidence, return, every, theorem, prove, known, problem, implies, define, induction, exists, let, depends, proof, follows, achieve, conference, stuck, smallest] [step, optimization] [can, order, one, high, also, see] [gaussian, given, finite, distance, two] [safe, safety, ret, state, exploration, mdp, unsafe, safely, model, rreach, explore, agent, action, exploring, reachability, visiting, transition, reinforcement, reachable, therefore, goal, starting, rnret, considers, rsafe, rover, priori, within, information, system, control, rret, afe] [without, use, used, feature, using, able]
Combining Low-Density Separators with CNNs
Yu-Xiong Wang, Martial Hebert
Yu-Xiong Wang, Martial Hebert
[number, block, thus] [learning, set, problem] [large, standard, optimization, additional, sampling] [can, leading, also, first, linear, still, generic] [space, data, specific, two, sample] [lds, target, action, new, across, decision] [layer, activation, unsupervised, unlabeled, using, cnn, cnns, top, figure, training, use, recognition, novel, feature, learn, performance, scene, deep, labeled, different, image, separator, train, approach, improve, used, transferability, visual, neural, generate, category, convolutional, per, imagenet, supervised, discriminative, flickr, unit, dataset, classification, learned, shown, work, original]
Threshold Learning for Optimal Decision Making
Nathan F. Lepora
Nathan F. Lepora
[threshold, rule, thus, find, many, distinct, probability, average, greater, number] [learning, function, optimal, cost, problem, ratio, considered, algorithm, rate, consider] [optimization, bayesian, method, log, standard, stochastic, variance, gradient, sequential, converges, sampling, challenging] [can, one, error, sampled, linear, comparison] [two, mean, gaussian, type, equal, data, derived, bayes, journal] [decision, reward, reinforce, model, time, making, acquisition, reinforcement, policy, exhaustive, trial, process, evidence, choice, optimize, sprt, animal, basal, experimental, psychological, sensory] [neural, figure, single, performance, used, use, approach, accuracy, forced, learn, perceptual, unit, multiple]
Phased LSTM: Accelerating Recurrent Network Training for Long or Event-based Sequences
Daniel Neil, Michael Pfeiffer, Shih-Chii Liu
Daniel Neil, Michael Pfeiffer, Shih-Chii Liu
[regular, number] [learning, every, rate, even, drawn, conference, since] [sampling, standard, update, ron, asynchronous, processing, gradient, epoch, faster] [can, sampled, first, note, frequency, phase, one, high] [data, two, important] [time, state, cell, model, period, long, new, longer, event, sensor, neuron, timing, controlled] [lstm, phased, input, gate, recurrent, network, neural, video, output, accuracy, rnn, training, using, memory, audio, layer, different, three, task, sine, presented, use, plstm, open, figure, preprint, arxiv, rnns, trained, previous, rhythmic, used, speech, deep, mfccs, stream, vision, digit, shown, oscillation, work]
Maximizing Influence in an Ising Network: A Mean-Field Optimal Solution
Christopher Lynn, Daniel D. Lee
Christopher Lynn, Daniel D. Lee
[external, stable, ising, influence, interaction, opinion, social, focusing, unique, structure, magnetization, block, jij, number, find, rise, viral, strength, represents, theory, marketing, susceptibility, spread, node, among, present, resulting, ferromagnetic] [optimal, theorem, budget, algorithm, exists, maximization, consider, maximizing, problem, concave, set, existence, since, general, max, even] [smooth, gradient, ascent, efficiently, apply, calculate, machine] [one, can, high, low, phase, solution, exact, critical, plot, sufficient, also, arg] [given, positive, boltzmann, random, statistical, maximum, mean, common] [system, transition, model, individual] [field, network, figure, performance, shown]
Scalable Adaptive Stochastic Optimization Using Random Projections
Gabriel Krummenacher, Brian McWilliams, Yannic Kilcher, Joachim M. Buhmann, Nicolai Meinshausen
Gabriel Krummenacher, Brian McWilliams, Yannic Kilcher, Joachim M. Buhmann, Nicolai Meinshausen
[dag, combination, root] [loss, algorithm, learning, regret, randomized, opt, setting, adaptive, fast, consider, rate, convex, dependence] [rad, adag, full, gradient, ada, variance, optimization, proximal, stochastic, approximate, iteration, faster, approximation, update, term, compute, reduction, processing, standard, step, variant, efficient, method, machine] [matrix, can, order, projection, low, diagonal, computational, second, first, svd, rank, high, square, following, also] [random, data, test, computationally, section, particular, practical] [information, inverse, range] [neural, training, using, similar, convolutional, performance, used, use, deep, preprint, arxiv, propose, effective, figure, dense, several, work, different]
Probabilistic Linear Multistep Methods
Onur Teymur, Kostas Zygalakis, Ben Calderhead
Onur Teymur, Kostas Zygalakis, Ben Calderhead
[probabilistic, number, present] [will, function, now, appendix, give, problem, consider, make, constant, define, proof, case, since, deviation] [method, deterministic, convergence, bayesian, ode, additional, initial, standard, posterior, university] [error, order, can, truncation, solution, local, plot, linear, also, computational, proposition, first, following, formulation, analysis] [given, numerical, gaussian, integrator, basis, multistep, differential, mean, distribution, polynomial, conditional, chua, difference, family, derivation, particular, way, covariance, statistical, csab, mathematics, equal, section, corresponding, estimate] [process, value, required, model, prior, uncertainty, new, therefore] [used, using, use, generate, figure, approach]
Spectral Learning of Dynamic Systems from Nonequilibrium Data
Hao Wu, Frank Noe
Hao Wu, Frank Noe
[potential, number, diffusion, denoted, obtained] [learning, algorithm, equilibrium, appendix, function, problem, theorem, general, will] [stochastic, operator, distributed, parameter, compute] [spectral, can, assumption, error, comparison, analysis, computational, related, order, singular] [ooms, data, estimation, binless, nonequilibrium, oom, length, observable, based, two, empirical, statistical, molecular, space, alanine, hmm, density, identically, metastable, discrete, lim, estimated, gaussian, independent, distribution, given, limit, finite, estimate, procedure] [observation, continuous, process, markov, dynamic, state, modeling, simulation, trajectory, time, predictive, starting] [proposed, feature, hidden, neural, without, used, shown, generated, figure]
Learning Deep Parsimonious Representations
Renjie Liao, Alex Schwing, Richard Zemel, Raquel Urtasun
Renjie Liao, Alex Schwing, Richard Zemel, Raquel Urtasun
[clustering, cluster, connected, number, chw, report, assigned] [learning, set, show, since, loss, may, case, online, every] [update, compute, gradient, method, objective, standard, end, size] [regularization, matrix, one, can, note, first, via, also, tensor, error, vector] [sample, based, center, test, data, generalization] [model, within, current, framework, baseline, new] [neural, deep, layer, representation, network, convolutional, different, use, fully, learned, parsimonious, table, using, spatial, applied, dataset, preprint, unsupervised, arxiv, training, feature, shown, investigate, classification, approach, trained, work, receptive, image, compare, input, architecture, unfolding, recent]
Learning User Perceived Clusters with Feature-Level Supervision
Ting-Yu Cheng, Guiguan Lin, xinyang gong, Kang-Jun Liu, Shan-Hung Wu
Ting-Yu Cheng, Guiguan Lin, xinyang gong, Kang-Jun Liu, Shan-Hung Wu
[] [] [] [] [] [] []
Conditional Image Generation with PixelCNN Decoders
Aaron van den Oord, Nal Kalchbrenner, Lasse Espeholt, koray kavukcuoglu, Oriol Vinyals, Alex Graves
Aaron van den Oord, Nal Kalchbrenner, Lasse Espeholt, koray kavukcuoglu, Oriol Vinyals, Alex Graves
[den, vertical, quality] [every, learning, class, conference, blind] [machine, processing, masked, latent, autoregressive, aaron] [can, also, one, row] [conditional, given, embeddings, two, distribution, conditioning, embedding] [model, information, new, van, current, modelled, another] [image, pixelcnn, neural, gated, pixel, figure, convolutional, deep, single, arxiv, preprint, different, using, used, generative, network, generate, performance, training, use, trained, able, conditioned, layer, shown, receptive, decoder, imagenet, field, deepmind, lstm, pixelrnn, natural, original, similar, google, pixelcnns, person, ivo, karol, table, international, instead, encoder, input, modelling, stack, generation, daan, improve, generated, visual]
On Valid Optimal Assignment Kernels and Applications to Graph Classification
Nils M. Kriege, Pierre-Louis Giscard, Richard Wilson
Nils M. Kriege, Pierre-Louis Giscard, Richard Wilson
[assignment, strong, hierarchy, graph, vertex, valid, base, histogram, intersection, path, induces, induced, subtree, rooted, tree, edge, root, definition, obtained, unique, kmax, according, number, rise, called, colour, ultrametric] [optimal, set, every, let, function, defined, class, since, problem, theorem, consider, general, may, element, show, will, obtain, provide] [inner, computed, refinement] [can, following, note, one, see, computation, restricted, related, product, dirac, first] [kernel, data, two, derived, associated, positive, given, section] [time, new] [feature, convolution, classification, figure, used, similarity, approach, shown, weight, using, applied, proposed, similar, different, image, three]
Can Active Memory Replace Attention?
Łukasz Kaiser, Samy Bengio
Łukasz Kaiser, Samy Bengio
[basic, number, markovian] [active, learning, will, focus, set, problem, might, every, since, provide, dependence] [machine, size, standard, processing, compute, step, pure] [can, one, good, tensor, first, also, matrix] [length, test, embedding] [model, extended, state, baseline, information, another, new, teacher, long, change] [neural, memory, attention, output, gpu, use, used, translation, image, figure, recurrent, using, input, bleu, network, mechanism, better, convolutional, training, previous, sfin, task, shape, lstm, different, sentence, deep, gru, decoder, residual, single, generative, language, part, cgrud, symbol, alex, ivo, table, perplexity, rnn, score, similar, generated, natural, cgru, convolution, whole, recent, source]
Consistent Kernel Mean Estimation for Functions of Random Variables
Carl-Johann Simon-Gabriel, Adam Scibior, Ilya O. Tolstikhin, Prof. Bernhard Schölkopf
Carl-Johann Simon-Gabriel, Adam Scibior, Ilya O. Tolstikhin, Prof. Bernhard Schölkopf
[variable, consistency, probabilistic, probability, programming] [set, function, provide, theorem, let, general, may, show, bounded, bound, upper, defined, assume, rate, even, learning, proof, case] [convergence, monte, carlo, apply, approximate, size, converges, university, machine] [can, reduced, also, good, product, one, need, main, first] [random, kernel, estimator, section, mean, embedding, sample, joint, kme, two, distribution, embeddings, kxy, expansion, consistent, space, based, estimate, hilbert, finite, rkhs, data, journal, theoretical, positive, given, provides] [continuous, rather, information] [using, input, used, applying, representation, multiple, approach, different, use, work, figure]
Cooperative Graphical Models
Josip Djolonga, Stefanie Jegelka, Sebastian Tschiatschek, Andreas Krause
Josip Djolonga, Stefanie Jegelka, Sebastian Tschiatschek, Andreas Krause
[pairwise, partition, graphical, edge, strength, strong, bethe, variable, graph, many, cut, linearized, resulting, disagreement] [problem, submodular, function, upper, lower, convex, bound, obtain, will, set, algorithm, even, polytope, example, minimization, general, learning, defined, binary] [inference, log, variational, trwbp, approximate, marginals, optimization, approximation, pmap, marginal, computing, method, linearization, compute, energy, solved, tractable, strongly] [can, one, vector, sdp, error, via, also, first, linear, solving, still] [entropy, random, family, estimate, result, two, discrete, exp, journal] [model, cooperative, new, optimize, belief] [figure, using, use, map, different, used, image]
Optimal Tagging with Markov Chain Optimization
Nir Rosenfeld, Amir Globerson
Nir Rosenfeld, Amir Globerson
[probability, vertex, many, adding, subset, describe, walk, monotonicity, social] [set, problem, optimal, will, algorithm, greedy, may, maximizing, show, now, prove, every, general, choose, submodular, assume, conference, since, consider, function, choosing] [optimization, objective, efficient, method, computing] [can, qij, tagging, tag, item, one, absorbing, pagerank, first, user, browsing, linear, denote, interesting, solving, matrix, recommendation, also, link, via, collaborative, decomposition, hence, related] [chain, given, random, based, corresponding, true, two, distribution, transient] [transition, markov, reaching, new, state, information, reach, optimizing, system, search, model, maximize] [task, using, used, use, work, add, different]
Online Convex Optimization with Unconstrained Domains and Losses
Ashok Cutkosky, Kwabena A. Boahen
Ashok Cutkosky, Kwabena A. Boahen
[average, number, knowledge] [lmax, rescaledexp, regret, algorithm, loss, online, convex, learning, maxt, setting, bound, achieve, theorem, oco, adadelta, max, ftrl, pistol, klmax, learner, subgradient, set, problem, conference, prove, lower, optimal, let, lemma, mmax, francesco, adversary, rtj, every, unknown, strategy, annual, element, kxt, adaptive] [optimization, adagrad, machine, stochastic, log, processing, gradient, update] [can, invariant, min, require, linear, first, order, analysis, following, vector, one, much, computational] [exp, adam, exponential, space, two, generalization] [hyperparameter, unconstrained, information, prior, value, showing, time] [neural, scale, using, without, figure, classification, use, arxiv, preprint]
Stochastic Gradient Richardson-Romberg Markov Chain Monte Carlo
Alain Durmus, Umut Simsekli, Eric Moulines, Roland Badeau, Gaël RICHARD
Alain Durmus, Umut Simsekli, Eric Moulines, Roland Badeau, Gaël RICHARD
[obtained] [theorem, algorithm, rate, show, bound, set, function, observe, optimal, consider, proof, assume, defined, let] [sgld, sgrrld, stochastic, step, gradient, mse, langevin, extrapolation, convergence, size, variance, monte, posterior, run, bayesian, carlo, converges, supplementary, sgrrhmc, large, parallel, distributed, monitor] [order, can, matrix, first, factorization, also, observed, computation, following, one, synthetic] [bias, two, distribution, estimator, gaussian, chain, given, data, based, random, fixed, numerical, mean, equation, brownian, true] [choice, model, therefore, time, markov] [sequence, using, figure, use, proposed, different, performance, shown, accuracy, applied, similar, approach, novel, applying]
Launch and Iterate: Reducing Prediction Churn
Mahdi Milani Fard, Quentin Cormier, Kevin Canini, Maya Gupta
Mahdi Milani Fard, Quentin Cormier, Kevin Canini, Maya Gupta
[resulting, probability, report] [set, learning, will, bound, ratio, expected, svm, problem, may, let, case, loss] [operator, mcmc, machine, initial, datasets, large, reduction, run] [can, see, one, regression, stability, note, also, sampled, suppose, first, regularization] [churn, chain, rcp, wlr, test, stabilization, random, pwin, fixed, diffs, two, acc, distribution, empirical, unnecessary, nomao, statistical, ridge, needed, statistically, diplopia, measure, testing, generalization] [model, markov, change, new, baseline, effect, future, towards] [training, classifier, accuracy, different, trained, figure, using, train, proposed, used, use, previous, table, without, dataset, compared, classification, feature, neural, candidate, shown]
Dense Associative Memory for Pattern Recognition
Dmitry Krotov, John J. Hopfield
Dmitry Krotov, John J. Hopfield
[number, many, rule, configuration, thus] [function, case, learning, set, problem, will, contribution, even] [energy, update, large, standard, capacity, step] [one, can, small, computational, order, linear, error, power, sign, also] [test, two, equal, prototype, family, limit, polynomial, given, data, well, section] [model, state, corresponds, information, neuron, behavior, backpropagation, transition, simple] [memory, network, neural, associative, activation, rectified, hidden, pattern, used, classification, layer, deep, different, visible, higher, training, stored, recognition, dense, image, using, feature, use, similar, output, performance, feedforward, presented, compared, store, several, input, digit, better, various]
Distributed Flexible Nonlinear Tensor Factorization
Shandian Zhe, Kai Zhang, Pengyuan Wang, Kuang-chih Lee, Zenglin Xu, Yuan Qi, Zoubin Ghahramani
Shandian Zhe, Kai Zhang, Pengyuan Wang, Kuang-chih Lee, Zenglin Xu, Yuan Qi, Zoubin Ghahramani
[number, subset, contains] [binary, set, algorithm, tight, learning, function, will, bound, online, obtain, may, optimal, complexity, lower] [latent, distributed, variational, inference, large, educe, gradient, inftuckerex, log, elbo, inducing, size, inftucker, gigatensor, develop, efficient, full, shuffling, subtensors, dintucker, optimization, standard, parallel, flexible, tgp, scalability] [tensor, can, factorization, tucker, nonzero, zero, following, small, decomposition, also, first, sparse, computational, xij, sampled, matrix] [data, nonlinear, gaussian, covariance, kernel, infinite, based, given, space, introduce, mean] [model, process, continuous, prior, modeling, mode, exploit] [use, using, used, fully, training, proposed, performance, perform, performed, figure]
DECOrrelated feature space partitioning for distributed sparse regression
Xiangyu Wang, David B. Dunson, Chenlei Leng
Xiangyu Wang, David B. Dunson, Chenlei Leng
[number, subset, false, partitioning, runtime, partition, consistency, variable] [theorem, algorithm, will, rate, set, since, strategy, chosen] [deco, size, full, decorrelation, parallel, fitting, distributed, stage, log, partitioned, mse, datasets, large, due, convergence, machine, parameter, processing, depend, step] [lasso, can, error, via, one, regression, first, condition, sparse, matrix, dimension, also, linear, computational, electricity] [data, sample, space, selection, statistical, section, estimation, random, journal, two, ridge, independent] [model, framework, time, information, new, eye, form, correlation] [performance, feature, using, dataset, approach, different, table, neural, compare, three, used]
SURGE: Surface Regularized Geometry Estimation from a Single Image
Peng Wang, Xiaohui Shen, Bryan Russell, Scott Cohen, Brian Price, Alan L. Yuille
Peng Wang, Xiaohui Shen, Bryan Russell, Scott Cohen, Brian Price, Alan L. Yuille
[edge, pairwise, inside, consistency, indicates, present] [confidence, since, show, learning, defined, upper, set, fast, improvement, loss] [inference, evaluation, term, implementation] [can, regularization, also, first, compatibility, comparison] [normal, joint, geometry, estimation, two, random, given] [information, within, whether, inferred] [planar, depth, surface, dcrf, network, image, plane, using, prediction, map, pixel, affinity, output, semantic, single, training, better, neural, use, dense, cnn, predicted, convolutional, approach, nyu, used, ground, truth, deep, eigen, four, figure, regularize, cnns, bilateral, segmentation, input, depicted, adopt, propose, back, entire, feature, layout, different, improves, evaluate, explicitly]
Can Peripheral Representations Improve Clutter Metrics on Complex Scenes?
Arturo Deza, Miguel Eckstein
Arturo Deza, Miguel Eckstein
[number, coefficient, detection, hit, regular, total, create, present, edge, influence] [will, rate, function, max, since, loss, every] [computed, away, experiment, size] [can, global, also, one, high] [journal, metric, difference, entropy, mean, measure, distance, section, point] [model, target, human, search, value, information, time, across, roi, retinal, observer, location, correlation, orientation, response] [clutter, feature, congestion, foveated, map, image, peripheral, score, fixation, deg, visual, eccentricity, pifc, architecture, pooling, forced, pyramid, used, scale, different, fovea, figure, final, crowding, input, proposed, region, subband, periphery, using, previous, original, ffc, pff, vision, simoncelli, representation, color, texture, freeman, dense]
Learning Transferrable Representations for Unsupervised Domain Adaptation
Ozan Sener, Hyun Oh Song, Ashutosh Saxena, Silvio Savarese
Ozan Sener, Hyun Oh Song, Ashutosh Saxena, Silvio Savarese
[consistency, rule, enforce, among] [learning, label, algorithm, problem, since, show, function, class, loss] [method, transductive, large, batch] [can, one, following, first, order, also, arg] [nearest, data, metric, neighbor, two, well, based, point, transformation, given, joint] [target, model, shift, state, framework, cyclic] [domain, source, unsupervised, adaptation, using, mnist, transduction, deep, feature, dataset, reject, classification, digit, use, learn, figure, object, discriminative, office, similarity, jointly, different, svhn, learned, proposed, network, image, fully, structured, task, table, neural, ethod, labelled, supervised, training, art, evaluate, convolutional, recognition, accuracy, webcam, input, without, transfer]
Linear dynamical neural population models through nonlinear embeddings
Yuanjun Gao, Evan W. Archer, Liam Paninski, John P. Cunningham
Yuanjun Gao, Evan W. Archer, Liam Paninski, John P. Cunningham
[variable, find, structure, distinct] [rate, function, class, algorithm, may, learning] [latent, inference, variational, approximate, posterior, log, reduction, likelihood, large, fitting] [linear, can, noise, analysis, also, much, recover] [data, nonlinear, population, true, sample, gaussian, distribution, dimensionality, result, mean, testing, fitted] [model, pflds, plds, time, predictive, dynamical, neuron, observation, spike, stimulus, aevb, poisson, gcflds, activity, lds, macaque, across, orientation, zrt, state, flds, system, lapem, simulation, count, grating, upon, firing, reaching, primary, simulated, recorded, xrti, kalman] [neural, use, generative, training, performance, approach, capture, using, preprint, arxiv, figure, compare, compared]
Learning to Communicate with Deep Multi-Agent Reinforcement Learning
Jakob Foerster, Yannis M. Assael, Nando de Freitas, Shimon Whiteson
Jakob Foerster, Yannis M. Assael, Nando de Freitas, Shimon Whiteson
[message, among, partial, many] [learning, differentiable, day, optimal, consider, since, binary, set, setting, chooses] [parameter, select, gradient, evaluation] [can, one, also, noise, order, first] [two, discrete, independent, space, important, address] [communication, agent, dial, action, rial, centralised, protocol, dqn, state, decentralised, policy, environment, reinforcement, interrogation, across, nocomm, receives, mat, essential, room, uat, switch, communicate, reward, another, share, must, time, cooperative, multiagent, execution] [deep, network, figure, neural, sharing, mnist, preprint, using, arxiv, learn, performance, language, recurrent, training, without, learned, approach, input, different, task, used]
Finite-Sample Analysis of Fixed-k Nearest Neighbor Density Functional Estimators
Shashank Singh, Barnabas Poczos
Shashank Singh, Barnabas Poczos
[probability, fact, strictly] [bound, lemma, rate, learning, conference, function, since, may, show, let, known, supported, constant, bounded, theorem, define, lower, class] [variance, machine, divergence, processing, log, convergence, additional, due] [can, ieee, analysis, error, also, require, one, suppose, assumption, via, support] [bias, estimator, density, estimation, entropy, based, functional, neighbor, mutual, functionals, estimating, estimate, sample, consistent, positive, shannon, correction, boundary, nonparametric, multivariate, asymptotic, given, increasing, random, mean, finite, fixing, alfred, nearest] [information, continuous, form, statement] [approach, neural, international, using, table, used, work, several]
Statistical Inference for Pairwise Graphical Models Using Score Matching
Ming Yu, Mladen Kolar, Varun Gupta
Ming Yu, Mladen Kolar, Varun Gupta
[graphical, scoring, rule, graph, edge, structure, number, valid, probabilistic] [let, confidence, asymptotically, set, coverage, learning, theorem, rate, case, constant, will, lemma, since, consider] [inference, log, parameter, large, step] [can, first, assumption, min, regularized, one, following, vector, via, linear, see, sparse, matrix, arg, high, condition, suitable] [exponential, estimator, gaussian, estimation, density, conditional, procedure, selection, family, distribution, data, normal, dimensional, consistent, given, statistical, empirical, based, asymptotic, estimated, construct, well, robust, random, sample, corresponding, signaling] [model, literature, form, next, event] [score, matching, using, work, use, table, used, arxiv, figure]
New Liftable Classes for First-Order Probabilistic Inference
Seyed Mehran Kazemi, Angelika Kimmig, Guy Van den Broeck, David Poole
Seyed Mehran Kazemi, Angelika Kimmig, Guy Van den Broeck, David Poole
[theory, lifted, recursion, wfomc, probabilistic, rule, clause, prvs, den, predicate, grounding, unary, relational, contains, srl, lvs, guy, weighted, number, david, logic, assigned, prv, represent, transitivity, dan, liftable, formula, recursively, birthday, called, branch, logical, compiling, clausal] [set, let, case, example, consider, every, learning, problem, binary, show, finding, class, lemma, may, assume] [inference, size, compute, calculating] [can, one, symmetric, also, first, analysis, exactly, suppose, decomposition, following] [two, population, exponential, true, polynomial, given, statistical, done, random] [van, time, model, markov, new, individual] [domain, using, ground, applying, unit, different, without]
Structured Sparse Regression via Greedy Hard Thresholding
Prateek Jain, Nikhil Rao, Inderjit S. Dhillon
Prateek Jain, Nikhil Rao, Inderjit S. Dhillon
[overlapping, number] [algorithm, greedy, let, set, theorem, convex, problem, show, even, least, obtain, provide, learning, function, lemma, appendix, setting, general, arbitrary, proof, conference, case, consider] [log, approximate, method, standard, processing, operator, solve, optimization] [group, iht, can, sparse, sparsity, regression, projection, hard, existing, following, thresholding, error, arg, min, also, condition, via, signal, vector, iterative, linear, recovery, gomp, note, one, suppose, require, restricted, ieee, exact, significantly, onto, analysis, high, sog, synthetic] [result, data, based, selection, given, well] [information, model, time] [using, similar, neural, international, overlap, used, structured, figure]
Fundamental Limits of Budget-Fidelity Trade-off in Label Crowdsourcing
Farshad Lahouti, Babak Hassibi
Farshad Lahouti, Babak Hassibi
[crowdsourcing, worker, number, probability, kic, average, incidence, memoryless, labeling, crowdsourcer, fundamental, taskmaster, crowd, possible, clustering, valid, level, identify, indicates] [query, may, case, consider, theorem, rate, problem, set, function, follows, scheme, unknown, budget, algorithm, known, appendix, label, show] [large, inference, iid, optimization, processing] [error, can, item, one, following, min, suitable, overall, small, noisy, certain, analysis] [two, distortion, theoretic, given, discrete, described, section, joint, remain] [skill, coding, code, information, model, design, feedback, within, optimized, purpose] [channel, source, performance, decoder, figure, presented, output, different, per, neural, dataset, input, used, work, task, using]
Safe and Efficient Off-Policy Reinforcement Learning
Remi Munos, Tom Stepleton, Anna Harutyunyan, Marc Bellemare
Remi Munos, Tom Stepleton, Anna Harutyunyan, Marc Bellemare
[thus, cut, coefficient, sum, possible, definition, notice] [learning, greedy, algorithm, consider, theorem, will, defined, since, case, online, function, arbitrary, proof, assume, conference, depends, best, may, make, lemma, deduce, general, close, show] [operator, convergence, variance, evaluation, full, machine, importance, sampling, efficient, around, converges] [assumption, can, contraction, product, first, low, need, min] [fixed, point, sample, result, infinite, estimate, given, finite, provided] [policy, control, target, behaviour, trace, increasingly, value, reinforcement, txe, bellman, action, choice, write, glie, safe, time, temporal, replay] [sequence, using, score, use, mapping, international, single, deepmind]
Unsupervised Feature Extraction by Time-Contrastive Learning and Nonlinear ICA
Aapo Hyvarinen, Hiroshi Morioka
Aapo Hyvarinen, Hiroshi Morioka
[number, monotonic, structure, obtained] [learning, function, case, since, will, blind, general, corollary, constant, show, theorem, proof, assume, define] [method, machine, logistic] [linear, can, also, one, mlr, component, note, analysis, see, vector] [nonlinear, data, ica, tcl, extractor, independent, segment, given, based, nonstationary, nonstationarity, modulated, principle, point, identifiability, equal, well, mixing, mean, estimation, transformation, distribution, chance] [model, time, temporal, must, new, meg, series] [feature, used, source, using, neural, unsupervised, generative, different, deep, classification, figure, use, layer, hidden, spatial, training, performance, trained, learn, similar, network, higher]
Sparse Support Recovery with Non-smooth Loss Functions
Kévin Degraux, Gabriel Peyré, Jalal Fadili, Laurent Jacques
Kévin Degraux, Gabriel Peyré, Jalal Fadili, Laurent Jacques
[probability, identifiable, stable, lagrange, constraint, sharp, thus, consistency] [case, lemma, loss, proof, theorem, let, problem, general, function, will, supported, assume, since, show, hold, define, now, set, instance, observe, convex, satisfied, provide, special] [dual, smooth, solve, large] [support, can, noise, solution, condition, stability, sparse, recovery, note, vector, sensing, one, compressed, analysis, main, first, certificate, restricted, also, small, order, norm, matrix, following, injectivity, subspace, see, ieee, sign, observed, min, require, sparsity, signal] [section, result, corresponding, data, theoretical, particular, journal, important, noting, associated, specific, correspond, equal] [extended, information, model] [different, using, figure, able]
Learning values across many orders of magnitude
Hado P. van Hasselt, Arthur Guez, Matteo Hessel, Volodymyr Mnih, David Silver
Hado P. van Hasselt, Arthur Guez, Matteo Hessel, Volodymyr Mnih, David Silver
[normalized, many] [learning, function, algorithm, consider, may, conference, defined, adaptive, changing, loss, lower, now, depends] [normalization, update, clipping, sgd, step, unnormalized, gradient, size, large, appropriate, machine, straightforward, processing] [can, magnitude, much, proposition, first] [data, important, squared, true, mean] [dqn, reinforcement, target, clipped, change, reward, new, shift, atari, adapt, normalize, mnih, median, van, thereby, action, time, information, heuristic, unclipped, hasselt, policy] [neural, scale, double, deep, using, without, different, performance, output, input, network, international, single, tune, used, use, natural, figure, work, layer]
Scan Order in Gibbs Sampling: Models in Which it Matters and Bounds on How Much
Bryan D. He, Christopher M. De Sa, Ioannis Mitliagkas, Christopher Ré
Bryan D. He, Christopher M. De Sa, Ioannis Mitliagkas, Christopher Ré
[variable, probability, number, conjecture] [will, show, theorem, prove, bound, set, lazy, best, always] [sampling, factor, augmented, method, faster, distributed, due] [order, can, conductance, one, good, relative, onto, small, also, analysis, much, matrix] [scan, systematic, mixing, random, two, bridge, gibbs, island, chain, tmix, distribution, permutation, space, efficiency, mix, polynomial, true, mass, stationary, section, needed, sample, corresponding, statistical, asymptotic, discrete, slower, sampler, result, way, nsf] [model, state, time, markov, transition, effect, move, information, choice, must] [using, different, sequence, single, pyramid, used, figure, several, compared]
Stochastic Gradient Methods for Distributionally Robust Optimization with f-divergences
Hongseok Namkoong, John C. Duchi
Hongseok Namkoong, John C. Duchi
[] [algorithm, convex, regret, problem, loss, bound, give, show, appendix, let, function, inf, risk, provide, minimization, learning, confidence, optimal, set, theorem, define, case, obtain, lemma, may, bandit, proof, now, conference, calibrated, differentiable, sup, online] [gradient, descent, stochastic, optimization, log, machine, objective, mirror, divergence, method, standard, efficient, dual, university, variance, duchi, step, convergence, update, strongly, distributionally, term, compute] [can, first, following, vector, formulation, solving, plot, solution, high, computational] [robust, empirical, section, procedure, journal, provides, sample, point, data, test, url] [time, identical, uncertainty, choice] [use, approach, using, performance, classification, figure]
Truncated Variance Reduction: A Unified Approach to Bayesian Optimization and Level-Set Estimation
Ilija Bogunovic, Jonathan Scarlett, Andreas Krause, Volkan Cevher
Ilija Bogunovic, Jonathan Scarlett, Andreas Krause, Volkan Cevher
[according, potential, theory] [algorithm, set, confidence, function, cost, gchk, setting, provide, regret, since, truvar, best, maximizers, choosing, dependence, learning, lower, bound, choose, define, chosen, corollary, version, smaller, consider] [variance, optimization, posterior, bayesian, supplementary, log, reduction, performs, respect, heteroscedastic] [noise, can, truncated, also, one, via, unified, synthetic, significantly, high, following, lake, small] [point, theoretical, data, mean, estimation, result, gaussian, lse, given, two, based, kernel, selected] [time, within, search, target, found, value, process, choice, typically, directly, track] [use, figure, previous, domain, performance, instead, different, using, better, similar, three]
Scaled Least Squares Estimator for GLMs in Large-Scale Problems
Murat A. Erdogdu, Lee H. Dicker, Mohsen Bayati
Murat A. Erdogdu, Lee H. Dicker, Mohsen Bayati
[number, find] [algorithm, least, cost, let, theorem, may, assume, rate, problem, constant, bound, show, provide, general, defined] [glm, ols, convergence, glms, method, step, optimization, bfgs, lbfgs, gradient, agd, proportionality, term, logistic, university, batch, cubic, quadratic] [regression, linear, error, can, denote, lasso, proposition, computational, matrix, second, order, min, first, vector, generalized, via, approximately] [mle, estimator, section, random, test, estimating, normal, gaussian, distribution, scaled, mean, covariance, covariates, estimate, corresponding, two, based, data, maximum] [design, time, model, right] [performance, used, use, proposed, dataset, relationship, figure, using, several]
Mixed Linear Regression with Multiple Components
Kai Zhong, Prateek Jain, Inderjit S. Dhillon
Kai Zhong, Prateek Jain, Inderjit S. Dhillon
[clustering, neighborhood] [algorithm, complexity, will, theorem, convex, let, show, max, function, appendix, set, optimal, constant, define, problem, case, guarantee, learning, proof, assume, minimization, provide] [method, objective, convergence, compute, gradient, mixed, requires, converges, standard, descent, initial, strongly, step] [linear, tensor, subspace, can, initialization, computational, min, mlr, regression, global, denote, also, power, small, recovery, local, solving, sampled, still, solution, dimension, exact, zit, formulation, certain, one, analysis, via, following] [sample, data, random, two, point, resampling, independent, locally, empirical] [model, time, required] [use, different, using, multiple, ground, generated, table, propose, daniel, truth, proposed, dataset]
Following the Leader and Fast Rates in Linear Prediction: Curved Constraint Sets and Other Regularities
Ruitong Huang, Tor Lattimore, András György, Csaba Szepesvari
Ruitong Huang, Tor Lattimore, András György, Csaba Szepesvari
[constraint, thus, theory] [regret, ftl, learning, convex, online, let, bound, loss, will, algorithm, differentiable, theorem, assume, show, set, case, since, function, lemma, even, bounded, satisfies, achieve, lower, logarithmic, setting, constant, known, inequality, achieves, minimax, get, proof, leader, conference, expected, prove, corollary, adaptive, fast, consider, example] [stochastic, away, machine, university, strongly, processing, optimization] [linear, can, proposition, also, one, principal, small, note, paper, following, support, denote, norm, first, much, enjoys, vector] [curvature, data, result, boundary, curve] [information, angle, next, whether, long] [prediction, used, unit, neural, compact, work, previous]
Variational Inference in Mixed Probabilistic Submodular Models
Josip Djolonga, Sebastian Tschiatschek, Andreas Krause
Josip Djolonga, Sebastian Tschiatschek, Andreas Krause
[probabilistic, called, partition, quality, represent, describe] [submodular, modular, set, function, problem, algorithm, upper, bound, will, consider, binary, facility, diversity, conference, learning, best, considered, let, andreas, general, property, arbitrary] [inference, approximate, variational, mixed, fldc, flic, flid, optimization, large, attractive, latent, repulsive, approximation, djolonga, machine, solve, log, inner, psms, josip, efficient] [can, product, one, item, recommendation, following, via, first, note] [distribution, given, corresponding, based, section, point, procedure] [model, location, form, optimize, optimizing] [use, used, performance, different, proposed, similar, task, using, accuracy, ground, better, dataset, approach, international]
A scalable end-to-end Gaussian process adapter for irregularly sampled time series classification
Steven Cheng-Xian Li, Benjamin M. Marlin
Steven Cheng-Xian Li, Benjamin M. Marlin
[number, block] [algorithm, show, expected, learning, loss, since, set, defined, label] [lanczos, approximation, adapter, posterior, gradient, method, compute, interpolation, inducing, uac, sampling, marginal, approximate, irregularly, computing, irregular, ski, log, efficiently, conjugate, approximating, due, computed, wilson, cubic, expectation] [can, sparse, matrix, error, also, sampled, computation, regression, exact, subspace, linear, krylov, one, symmetric, square] [kernel, gaussian, data, random, covariance, section, described, given, mean, space, sample, based] [time, series, framework, meg, process, uncertainty, model] [using, classification, use, classifier, training, train, different, feature, applied, shown, neural, including, representation, proposed, work, output, approach, structured]
Neurons Equipped with Intrinsic Plasticity Learn Stimulus Intensity Statistics
Travis Monk, Cristina Savin, Jörg Lücke
Travis Monk, Cristina Savin, Jörg Lücke
[circuit, variable, average, negative, number, represents, rule] [learning, class, function, set, optimal, let, algorithm, may, show, appendix] [parameter, posterior, standard, bayesian] [can, also, one, preprocessing, link] [data, wcd, given, distribution, robust, mixture, mean, comp] [model, plasticity, intensity, excitability, intrinsic, neuron, synaptic, stimulus, gain, evolution, spiking, artificial, learns, maximize, ppg, role, process, dull, change, wcgen, gsm, observation, cortical, activity, hebbian] [neural, network, input, hidden, figure, generative, used, classification, using, learn, different, unsupervised, dataset, mnist, training, digit, use, work, performance, generate, learned, without, novel, generated]
Domain Separation Networks
Konstantinos Bousmalis, George Trigeorgis, Nathan Silberman, Dilip Krishnan, Dumitru Erhan
Konstantinos Bousmalis, George Trigeorgis, Nathan Silberman, Dilip Krishnan, Dumitru Erhan
[htc, number] [loss, private, learning, set, class, function, let, label] [datasets, gradient, respect] [synthetic, error, also, can, one, real, orthogonality, subspace] [mmd, mean, data, two, squared, kernel, separation] [target, model, traffic, soft] [domain, shared, source, adaptation, unsupervised, representation, use, dataset, classification, used, object, training, classifier, trained, using, pose, mnist, image, reconstruction, dann, svhn, similarity, work, similar, background, neural, task, encoder, applied, like, dsn, deep, ground, network, truth, input, hsc, layer, hidden, different, labeled, recognition, validation, predict, without]
Online and Differentially-Private Tensor Decomposition
Yining Wang, Anima Anandkumar
Yining Wang, Anima Anandkumar
[definition, probability, variable, number, level] [algorithm, privacy, theorem, private, online, differentially, learning, improved, show, complexity, streaming, bound, exists, guarantee, let, consider, bounded, sup, even, appendix, arbitrary, least, set] [method, step, end, efficient, latent, linearly, requires, objective] [tensor, power, noise, decomposition, analysis, can, matrix, one, assumption, perturbation, dimension, symmetric, also, first, order, noisy, spectral, condition, recovery, popular, orthogonal, via, component, linear, error, suppose, high, initialization, remark, norm, recovers, guaranteed, whitening] [data, robust, random, sample, important, practical, gaussian, moment, procedure, differential, journal, population] [model, simple] [memory, input, using]
Architectural Complexity Measures of Recurrent Neural Networks
Saizheng Zhang, Yuhuai Wu, Tong Che, Zhouhan Lin, Roland Memisevic, Ruslan R. Salakhutdinov, Yoshua Bengio
Saizheng Zhang, Yuhuai Wu, Tong Che, Zhouhan Lin, Roland Memisevic, Ruslan R. Salakhutdinov, Yoshua Bengio
[graph, number, node, directed, coefficient, definition, path, adding, edge, dependency, larger] [learning, set, consider, complexity, general, let, since, even, might] [term, size, sequential, step, large, extra, increase] [can, also, one, first, following, note, denote] [two, given, length, increasing, test, specific, nonlinear, measure, result] [time, model, connecting, long, cyclic, information, new, transition, baseline] [recurrent, skip, rnn, depth, neural, gun, feedforward, tanh, rnns, different, mnist, performance, architecture, figure, sequence, lstm, preprint, using, input, arxiv, hidden, table, shown, architectural, stacked, yoshua, previous, similar, multiple, unfolded, output, investigate]
An Online Sequence-to-Sequence Model Using Partial Conditioning
Navdeep Jaitly, Quoc V. Le, Oriol Vinyals, Ilya Sutskever, David Sussillo, Samy Bengio
Navdeep Jaitly, Quoc V. Le, Oriol Vinyals, Ilya Sutskever, David Sussillo, Samy Bengio
[block, number, probability, thus, possible, partial] [best, learning, show, every, function, conference, context, let, since] [step, computed, log, end, size, compute, processing, timit, large] [can, one, vector, first, note] [two, equation, data, section] [model, time, state, next, target, current, information, across, emitted, brain] [transducer, neural, output, input, sequence, attention, using, encoder, speech, recurrent, used, lstm, last, layer, figure, alignment, network, recognition, produce, rnn, entire, produced, symbol, conditioned, three, work, deep, prediction, mechanism, impact, arxiv, preprint, trained, oriol, per, google, generated, softmax, seen, achieved, accuracy, task, table, dzmitry, yoshua]
Community Detection on Evolving Graphs
Stefano Leonardi, Aris Anagnostopoulos, Jakub Łącki, Silvio Lattanzi, Mohammad Mahdian
Stefano Leonardi, Aris Anagnostopoulos, Jakub Łącki, Silvio Lattanzi, Mohammad Mahdian
[cluster, node, clustering, evolving, probability, structure, block, graph, number, community, many, social, detection, cit] [algorithm, query, case, lemma, show, assume, may, every, theorem, constant, expected, even, now, bound, let, problem, give, strategy, study, exists, conference, optimal] [step, stochastic, log, size, large] [can, first, assumption, high, also, error, one, note, analysis, main, detecting, denote, second] [data, given, random, two, distribution, section, theoretical] [model, time, dynamic, evolution, new, information] [per, single, use, work, different, entire, using, input, used, network, probing, international, able]
Tight Complexity Bounds for Optimizing Composite Objectives
Blake E. Woodworth, Nati Srebro
Blake E. Woodworth, Nati Srebro
[find, number, average, thus] [lower, algorithm, randomized, convex, oracle, prox, bound, function, access, complexity, consider, make, kxk, upper, appendix, defined, round, mlb, show, even, optimal, obtain, exists, theorem, since, problem, will, dependence, least, composite, provide, minimizing, query, define] [deterministic, log, gradient, optimization, accelerated, smooth, stochastic, strongly, distributed, term, svrg, large, iterates, machine, atyusha] [can, component, order, dimension, also, following, first, orthogonal, exact, much, sufficiently, need] [point, finite, construction, based] [must, optimizing, information, required] [using, previous, arxiv, preprint, similar, table, several, without, different, neural, presented, use]
Inference by Reparameterization in Neural Population Codes
Rajkumar Vasudeva Raju, Xaq Pitkow
Rajkumar Vasudeva Raju, Xaq Pitkow
[probabilistic, graphical, node, many, lbp, propagation, represent, pairwise, probability, loopy, distinct, tree, rule] [algorithm, general, since, set, will, even] [inference, marginal, distributed, quadratic, normalization, reparameterization, posterior, approximate, bayesian, processing, performs, update] [can, linear, one, local, sufficient, computation, noise] [nonlinear, population, distribution, two, joint, gaussian, true, important, multivariate, section, exp] [activity, information, belief, trp, model, ppcs, time, tst, neuroscience, brain, pseudomarginals, evidence, marginalization, nature, slow, beck, divisive, neuronal, sensory] [neural, network, different, natural, figure, using, performance, perform, use, encode, multiple, representation, work]
Data Poisoning Attacks on Factorization-Based Collaborative Filtering
Bo Li, Yining Wang, Aarti Singh, Yevgeniy Vorobeychik
Bo Li, Yining Wang, Aarti Singh, Yevgeniy Vorobeychik
[knowledge, complete, resulting] [attacker, minimization, algorithm, optimal, set, utility, function, problem, learning, uniformly, defined, since, availability, conference, consider] [gradient, optimization, compute, parameter, sgld, machine, posterior, computed, step, university] [malicious, matrix, attack, poisoning, can, collaborative, norm, nuclear, filtering, alternating, user, mij, one, solution, also, first, recommendation, formulation, avoid, rmse, completion, rating, kkt, rated, integrity, item, singular, min, approximately, popular, pga, via, projected, regularization] [data, normal, two, based, specific, random, section, robust] [system, model, research, prior, information, value] [using, use, different, figure, adversarial, original, similar, feature, effective]
k*-Nearest Neighbors: From Global to Local
Oren Anava, Kfir Levy
Oren Anava, Kfir Levy
[number, weighted, thus, find, present, consistency] [optimal, algorithm, implies, bound, choosing, adaptive, best, theorem, assume, consider, may, latter, since, problem, cost, confidence] [standard, term, objective, yielding, datasets, log, method, requires, efficiently] [solution, first, note, following, order, running, condition, error, analysis, second, vector, regression, local, also, nonzero, noise] [data, given, kernel, nearest, neighbor, equation, distance, section, two, point, locally, metric, theoretical, well, based, estimation, regime, bandwidth] [value, decision, time, form, therefore, information, whereas, towards, new] [prediction, approach, weight, using, several, classification, different, per, used, dataset, three, text, table]
Learning Multiagent Communication with Backpropagation
Sainbayar Sukhbaatar, arthur szlam, Rob Fergus
Sainbayar Sukhbaatar, arthur szlam, Rob Fergus
[number, thus, block, graph, structure] [set, will, learning, get, now, call, function, show, expected, conference, let] [gradient, processing] [can, vector, also, one, linear, matrix, note, order, view] [two, discrete, given, well, independent, address, specific] [model, communication, agent, controller, time, form, state, take, reward, dynamically, baseline, reinforcement, multilayer, followed, control, nonlinearity, concatenation, communicate, another, simple, feed, rather, cooperating, extend, policy, abstract, information, within, scalar, simplify, commnet, sized, email, broadcast, module, range] [neural, input, network, used, output, single, architecture, layer, hidden, applied, consists, use, feedforward, viewed, using, final, different]
Combinatorial Energy Learning for Image Segmentation
Jeremy B. Maitin-Shepard, Viren Jain, Michal Januszewski, Peter Li, Pieter Abbeel
Jeremy B. Maitin-Shepard, Viren Jain, Michal Januszewski, Peter Li, Pieter Abbeel
[kevin, merge, connected, pairwise, larger, structure] [learning, obtain, binary, combinatorial, may, defined, define, general] [energy, method, computed, large, step, size, initial, efficient, evaluation] [local, connectivity, can, error, also, existing, vector, sebastian] [based, two, corresponding, given, boundary, test, automated, true, well, data, type] [model, information, within, prior, box, neuronal, required, voxels, modeling, potentially, action] [segmentation, shape, image, descriptor, neural, celis, agglomeration, training, bounding, network, using, deep, used, position, use, microscopy, convolutional, reconstruction, voxel, accuracy, winfried, electron, shown, volume, single, score, moritz, classification, dataset, rand, several, approach, feature, scale, viren, object, computer, watershed, architecture]
Perspective Transformer Nets: Learning Single-View 3D Object Reconstruction without 3D Supervision
Xinchen Yan, Jimei Yang, Ersin Yumer, Yijie Guo, Honglak Lee
Xinchen Yan, Jimei Yang, Ersin Yumer, Yijie Guo, Honglak Lee
[introduced] [learning, loss, problem, set] [silhouette, latent] [projection, can, view, matrix, good, one, following, also] [test, transformation, two, corresponding, point, generalization, based, given, project, data] [model, agent, azimuth, demonstrate, experimental] [volume, shape, network, object, image, perspective, volumetric, training, representation, using, trained, figure, single, convolutional, input, shown, transformer, ground, category, reconstruction, used, without, neural, proposed, generated, table, truth, unseen, three, learn, camera, deep, viewpoint, encoder, chair, decoder, voxel, combined, performance, learned, conv, propose, better, multiple, train, arxiv, prediction, recurrent, preprint, computer, able, understanding, output]
PAC Reinforcement Learning with Rich Observations
Akshay Krishnamurthy, Alekh Agarwal, John Langford
Akshay Krishnamurthy, Alekh Agarwal, John Langford
[number, path, many, probability] [learning, algorithm, function, complexity, optimal, dependence, contextual, set, bound, since, may, lower, assume, class, least, problem, even, general, show, exponentially, pac, focus, bandit] [step, consensus, large, size, requires, deterministic, efficient] [can, assumption, one, also, require, regression, global, denote, first] [sample, polynomial, empirical, estimate, given, space, finite, true, theoretical, result, two] [state, reinforcement, policy, model, value, observation, elimination, exploration, reactive, dfs, pomdps, rich, new, search, decision, action, future, surviving, transition, lsvee, must, time, reward, current, required, planning, realizability, agent, goal] [use, using]
Multi-armed Bandits: Competing with Optimal Sequences
Zohar S. Karnin, Oren Anava
Zohar S. Karnin, Oren Anava
[total, block, bin, average, thus, identify, number, say, inside, denoted, knowledge, notice] [regret, algorithm, loss, arm, bound, optimal, setting, deviation, absolute, online, exploitation, problem, will, est, round, bandit, set, concentrated, adversary, consider, defined, observe, whenever, obtaining, bounded, best, example, assume, might, provide] [stochastic, strongly, standard, optimization] [can, phase, analysis, one, linear, first, observed, sublinear, hard, also, still, def] [mean, test, random, two, statistical, given, fixed, stationary] [action, exploration, environment, value, dynamic, feedback, whether, change, player, time, within, benchmark, determine, long] [sequence, variation, use, static, work, without, using, adversarial]
Incremental Boosting Convolutional Neural Network for Facial Action Unit Recognition
Shizhong Han, Zibo Meng, AHMED-SHEHAB KHAN, Yan Tong
Shizhong Han, Zibo Meng, AHMED-SHEHAB KHAN, Yan Tong
[strong, number, expression, average] [learning, loss, function, active, set, improvement, will] [incremental, boosting, iteration, boosted, standard, due] [can, ieee, analysis] [selected, based, well, data, positive] [action, decision, limited, information, current, blue, individual, benchmark, target, new] [facial, cnn, proposed, layer, classifier, recognition, training, score, performance, weak, figure, database, calculated, feature, convolutional, cnns, employed, face, used, semaine, neural, spontaneous, input, traditional, activation, shown, red, different, unit, recognizing, novel, learned, disfa, classification, using, deep, trained, updated, illustrated, table, discriminative, previous, four, network]
SoundNet: Learning Sound Representations from Unlabeled Video
Yusuf Aytar, Carl Vondrick, Antonio Torralba
Yusuf Aytar, Carl Vondrick, Antonio Torralba
[knowledge] [learning, show, may, loss, since, even] [large, standard, five] [also, can, comparison, order, leverage, one, synchronization, existing] [data, suggest, mean, length] [teacher, model, rich, human] [sound, network, unlabeled, visual, convolutional, video, deep, vision, use, soundnet, scene, using, classification, natural, trained, acoustic, layer, training, performance, representation, transfer, dataset, figure, table, train, used, audio, accuracy, recognition, learn, object, without, labeled, hidden, better, discriminative, neural, approach, deeper, supervision, imagenet, vgg, transferring, andrew, feature, antonio, architecture, recognize, propose, depth, learned, preprint, perform, visualize, dcase, carl]
Improved Error Bounds for Tree Representations of Metric Spaces
Samir Chowdhury, Facundo Mémoli, Zane T. Smith
Samir Chowdhury, Facundo Mémoli, Zane T. Smith
[tree, ultrametric, number, linkage, clustering, find, notice, claim, present] [bound, theorem, let, max, optimal, defined, cardinality, define, set, upper, covering, proof, depending, prove, obtain, dependence, bounded, will, appendix, function, provide, duality, consider, notion, problem, observe, since, show] [method, university] [can, dimension, stability, one, min, remark, also, denote, note, low, first] [metric, space, embedding, distortion, doubling, data, slhc, additive, finite, given, dxn, dln, two, gromov, hyperbolicity, uxn, ultrametricity, voronoi, result, journal, preserving, point, construction, numerical, dendrogram] [write, hierarchical, typically] [using, question, able, different, map, multiplicative]
Satisfying Real-world Goals with Dataset Constraints
Gabriel Goh, Andrew Cotter, Maya Gupta, Michael P. Friedlander
Gabriel Goh, Andrew Cotter, Maya Gupta, Michael P. Friedlander
[constraint, negative, number, often] [problem, convex, will, algorithm, fairness, function, upper, ramp, rate, deployed, set, tight, svm, may, randomized, appendix, loss, every, zafar, hinge, thresholded, learning, let, minimize, svmoptimizer, egregious, feasible, bound, expected] [optimization, large, datasets, method, objective, dual, parameter] [can, one, linear, vector, error, also, support, small, certain] [churn, positive, section, testing, two, given, test, expressed] [model, optimizing, new, desired, optimize, unconstrained, found] [training, classifier, dataset, using, proposed, classification, recall, approach, unlabeled, propose, used, use, labeled, figure, table, different, prediction, candidate, accuracy]
High-Rank Matrix Completion and Clustering under Self-Expressive Models
Ehsan Elhamifar
Ehsan Elhamifar
[clustering, notice, number, incomplete, complete, subset, whose, graph] [algorithm, since, function, convex, problem, conference, set, learning, consider, show, finding, selecting] [machine, performs, optimization, due, method, objective, large, solve] [missing, completion, matrix, fraction, cij, subspace, can, error, lrmc, recover, corrupted, solution, lie, sparse, union, ssc, ieee, rank, synthetic, real, min, via, nonzero, cin, lifting, first, ambient] [data, point, given, journal, important] [framework, deal, information] [using, mfa, computer, figure, different, motion, international, propose, performance, similarity, better, use, representation, dataset, pattern]
Sorting out typicality with the inverse moment matrix SOS polynomial
Edouard Pauwels, Jean B. Lasserre
Edouard Pauwels, Jean B. Lasserre
[degree, detection, number, sum, level, panel, potential, unique, average, collection] [set, theorem, let, will, problem, learning, satisfies, defined, function, lemma, appendix, optimal, provide] [suggests, machine, evaluation, optimization, method] [matrix, can, orthogonal, assumption, following, real, solution, computation, outlier, global, denote, order, also, one] [polynomial, moment, section, empirical, measure, data, given, mathematical, monomials, distinguished, christoffel, cloud, positive, outlyingness, intrusion, well, npd, estimation, based, sublevel, borel, rationale, density, result, family, equal] [inverse, value, simple, left] [shape, network, figure, use, using, similar, score, higher, different, connection, international]
Optimal Binary Classifier Aggregation for General Losses
Akshay Balsubramani, Yoav S. Freund
Akshay Balsubramani, Yoav S. Freund
[ensemble, partial, potential, aggregation, structure, constraint, lagrange, false] [loss, function, binary, minimax, convex, general, learning, optimal, problem, label, theorem, set, example, max, defined, lemma, minimize, will, bound, algorithm, minimizer, class, earlier, may, randomized, appendix] [optimization, slack, dual, efficient, convergence, sigmoid, discussed, applies] [can, vector, also, min, assumption, linear, formulation, solution, convexity, error, paper] [test, data, true, given, predictor, well, result, uniform, increasing, particularly, corresponding, equation] [game, information, decision, convenient, typically, simply] [prediction, unlabeled, using, classifier, classification, used, work, different, including, predict, like, without, use, figure]
Completely random measures for modelling block-structured sparse networks
Tue Herlau, Mikkel N. Schmidt, Morten Mørup
Tue Herlau, Mikkel N. Schmidt, Morten Mørup
[number, caron, fox, degree, obtained, block, edge, exchangeability, probability, exchangeable, vertex, aij, according, variable, expression, completely, many, assignment, crmsbm, notice, kallenberg, total, structure] [will, define, obtain, function, assume, consider, since, selecting] [sampling, latent, standard, supplementary, method, parameter, update, posterior, bayesian, iid, inference] [can, first, suppose, link, sparse, following, order] [random, measure, distribution, important, infinite, procedure, corresponding, crm, mass, statistical, associated, construction, based] [model, process, poisson, must, form, found, simple, prior] [network, using, figure, representation, use, generated, modelling, different, prediction, invariance, work, generate]
Globally Optimal Training of Generalized Polynomial Neural Networks with Nonlinear Spectral Methods
Antoine Gautier, Quynh N. Nguyen, Matthias Hein
Antoine Gautier, Quynh N. Nguyen, Matthias Hein
[thus, unique, complete, number] [theorem, let, defined, function, every, lemma, problem, show, optimal, satisfies, constant, achieve, algorithm, proof, guarantee, loss, implies, max, learning, exists, rate, class, prove, consider, lipschitz] [method, objective, sgd, convergence, gradient, optimization, parameter, large, stochastic] [global, one, spectral, can, matrix, note, order, globally, main, optimality, small, see, optimum, certain, also, first, nonnegative, critical, paper, following, fuab, linear, tensor] [fixed, point, metric, nonlinear, radius, two, positive, data, section, space, test, given, uci] [model, reason, new] [neural, hidden, use, training, network, layer, deep, figure, performance, architecture, work]
Fast learning rates with heavy-tailed losses
Vu C. Dinh, Lam S. Ho, Binh Nguyen, Duy Nguyen
Vu C. Dinh, Lam S. Ho, Binh Nguyen, Duy Nguyen
[probability, clustering, obtained, number] [learning, fast, lemma, function, hypothesis, rate, loss, class, risk, sup, bounded, satisfies, mendelson, exists, define, theorem, heavy, let, consider, exist, defined, least, prove, study, bound, optimal, obtain, derive, mehta, concentration, brownlees, depending, regularity, constant, implies] [convergence, standard, verify, log, machine, separable] [condition, can, assumption, denote, order, following, also, vector, arbitrarily, min] [empirical, envelope, distribution, finite, result, given, journal, random, positive, definite, robust, annals] [unbounded, information, failure, extend, new, van, assuming] [source, previous, using, enable, work, recall, neural]
Adaptive Concentration Inequalities for Sequential Decision Problems
Shengjia Zhao, Enze Zhou, Ashish Sabharwal, Stefano Ermon
Shengjia Zhao, Enze Zhou, Ashish Sabharwal, Stefano Ermon
[probability, number, walk, many, variable, often, threshold, definition, identification, law] [bound, algorithm, theorem, stopping, let, hoeffding, arm, adaptive, problem, set, best, asymptotically, concentration, inequality, optimal, bounded, will, achieve, show, hold, terminate, confidence, make, lil, exponentially, chosen, learning, positiveness, adaptively, constant, function, risk, crossed, unlikely, robert] [log, sequential, large, university, parameter] [can, also, zero, sparse, plot, one, analysis, following] [random, mean, boundary, empirical, based, sample, distribution, well, asymptotic, testing, lim, journal] [time, new, agent, decision, behavior, value, design] [figure, performance, similar, use, various, including, using]
Blazing the trails before beating the path: Sample-efficient Monte-Carlo planning
Jean-Bastien Grill, Michal Valko, Remi Munos
Jean-Bastien Grill, Michal Valko, Remi Munos
[trailblazer, node, number, avg, tree, root, called, definition, possible, either, subset, child, uct, probability, notice, structure, identify, branching] [complexity, max, case, bound, algorithm, set, want, define, problem, call, consider, provide, optimistic, optimal, munos, may, defined, oracle, theorem, difficulty, conference, expected, function] [sampling, end, factor] [can, order, following, first, high, min, note, one] [sample, finite, measure, infinite, two, maximum, estimate, exponential, specific, quantity, polynomial, random, result] [planning, value, state, model, next, action, transition, information, mdp, control, potentially, policy, reward, decision, stop] [generative, depth, figure, previous, use, using, neural, like]
Maximization of Approximately Submodular Functions
Thibaut Horel, Yaron Singer
Thibaut Horel, Yaron Singer
[definition, many, called, among, theory] [submodular, function, algorithm, case, set, constant, learning, ratio, lower, greedy, problem, monotone, show, cardinality, max, let, maximizing, obtain, exists, coverage, bound, submodularity, lemma, theorem, implies, class, even, returned, close, achieves, define, access, drawn, uniformly, inequality, active, general, proof, matroid, version, optimal, consider, since, convex, otherwise, bounded] [approximation, approximate, optimization, size, objective] [can, approximately, one, error, high, assumption, noise, via, solution, noisy, main, second, note, also, first] [random, given, result, curvature, two, application, additive, estimated, independent, exponential] [value, information, range] [using, used, input]
Adversarial Multiclass Classification: A Risk Minimization Perspective
Rizal Fathony, Anqi Liu, Kaiser Asif, Brian Ziebart
Rizal Fathony, Anqi Liu, Kaiser Asif, Brian Ziebart
[consistency, constraint, number, potential, possible] [loss, multiclass, minimization, function, set, learning, risk, hinge, theorem, algorithm, label, svm, margin, llw, binary, universally, argmaxj, surrogate, max, since, best, peter, minimax, equilibrium, conference, convex, case] [dual, machine, method, optimization, parameter, datasets, large, processing] [vector, support, linear, can, formulation, also, relative, min, computational, low] [kernel, fisher, empirical, data, provides, journal, space, consistent, gaussian, true, universal, theoretical, two, given, multivariate, efficiency] [game, model, information, value] [adversarial, feature, training, classification, using, classifier, prediction, three, table, performance, accuracy, neural, input, use, generation, dataset, figure]
Learning Sensor Multiplexing Design through Back-propagation
Ayan Chakrabarti
Ayan Chakrabarti
[number, find, level] [learning, optimal, since, even, set] [inference, full, method, standard] [noise, measurement, one, can, also, see, hard, significantly, first, sensing, sparse, note, computational, ieee] [corresponding, joint, measure, based, two, entropy, given] [sensor, design, measured, across, light, learns, choice, intensity] [color, pattern, reconstruction, network, training, image, bayer, multiplexing, neural, using, learned, learn, approach, layer, use, jointly, different, used, channel, input, demosaicking, rgb, traditional, camera, output, train, deep, trained, psnr, higher, cfa, performance, single, able, better, similar, table, aliasing, encode, coded, convolutional, proposed, final, figure]
Multi-step learning and underlying structure in statistical models
Maia Fraser
Maia Fraser
[possible, probability, definition, structure, subset] [learning, class, bound, lower, case, let, shattering, function, algorithm, assume, concept, will, upper, complexity, expected, problem, example, learner, prove, define, consider, supremum, theorem, worst, loss, access, version, depends, defined, sup, hypothesis, since, implies, pac, best, now] [size, marginal, marginals] [dimension, can, also, error, one, compatibility, assumption, noisy, group, first, invariant, sufficient] [sample, distribution, joint, given, data, true, measure, manifold, uniform, two, generalization, determined, underlying, section, statistical, conditionals, finite, reference] [framework, model, action, stick, next] [using, labeled, unlabeled, use, used, figure, training, three]
Learning Bayesian networks with ancestral constraints
Eunice Yuh-Jie Chen, Yujia Shen, Arthur Choi, Adnan Darwiche
Eunice Yuh-Jie Chen, Yujia Shen, Arthur Choi, Adnan Darwiche
[ancestral, dag, constraint, edge, ordering, structure, path, decomposable, variable, graph, directed, iff, tree, contains, negative, find, cpdag, topological, number, ancestor, entailed, programming, adding, knowledge, enforce, compatible, shortest, denoted, represent] [set, learning, optimal, consider, satisfies, conference, algorithm, problem, finding, since, function, oracle, let, satisfy, show, exists, will, case, may] [bayesian, machine] [can, one, denote, also, linear, note, following, first, percentage] [given, based, section, two, space, positive, equation, maximum] [search, infer, artificial, time, heuristic] [approach, network, using, table, used, proposed, different, dataset, impact, international, available]
Coevolutionary Latent Feature Processes for Continuous-Time User-Item Interactions
Yichen Wang, Nan Du, Rakshit Trivedi, Le Song
Yichen Wang, Nan Du, Rakshit Trivedi, Le Song
[interaction, base, number, mae, influence, contains] [function, will, convex, may, online, service, algorithm, since] [latent, averaging, method, gradient, objective, compute, efficient, epoch, term, optimization, due] [user, can, item, rank, low, poissontensor, lowrankhawkes, coevolving, one, matrix, also, observed, stic, factorization, recommendation, collaborative, hence, group, iptv, movie, product, interact, fip, ank] [point, data, two, conditional, based, space, given, testing, kernel, exponential] [time, model, temporal, intensity, process, event, hawkes, history, drift, nature, reddit, new, coevolutionary, yelp, poisson, framework] [feature, prediction, different, using, figure, capture, better]
Universal Correspondence Network
Christopher B. Choy, JunYoung Gwak, Silvio Savarese, Manmohan Chandraker
Christopher B. Choy, JunYoung Gwak, Silvio Savarese, Manmohan Chandraker
[negative, many, mining, number] [learning, loss, since, fast, surrogate, margin] [faster, efficient, large, normalization, optimization] [also, accurate, note, global, hard, second, sparse, invariant, can] [metric, geometric, neighbor, nearest, space, test, universal, estimation] [raw, prior, across, directly] [correspondence, convolutional, spatial, network, use, image, dense, feature, transformer, semantic, patch, fully, contrastive, sift, visual, figure, similarity, table, training, ucn, dataset, pck, performance, used, ground, flow, truth, using, pair, kitti, different, propose, train, shape, deep, layer, keypoints, neural, keypoint, learn, daisy, accuracy, without, trained, siamese, object, pascal, cnn, novel, matching, various, cub, generate, appearance, several]
Active Learning with Oracle Epiphany
Tzu-Kuo Huang, Lihong Li, Ara Vartanian, Saleema Amershi, Xiaojin Zhu
Tzu-Kuo Huang, Lihong Li, Ara Vartanian, Saleema Amershi, Xiaojin Zhu
[number, disagreement, probability, induced, threshold, total] [oracle, epiphany, active, learning, query, algorithm, epical, complexity, version, will, agnostic, hypothesis, label, may, passive, cal, bound, case, unknown, set, erm, learner, define, let, mcal, since, online, problem, theorem, induce, every, webpage, risk, appendix, least, assume, realizable, analyze, abstain, regret, show, huang, consider, formalize] [standard, end, term, processing, machine, step, factor] [analysis, can, error, one, note, also, following, small, order] [space, two, empirical, section, distribution, given, additive, theoretical, test] [information, model, human, basketball, current] [region, neural, input, without, unlabeled, classification, labeled, figure]
Exploiting Tradeoffs for Exact Recovery in Heterogeneous Stochastic Block Models
Amin Jalali, Qiyang Han, Ioana Dumitriu, Maryam Fazel
Amin Jalali, Qiyang Han, Ioana Dumitriu, Maryam Fazel
[community, sbm, number, block, heterogenous, nmin, pmin, detection, program, many, probability, recoverability, graph, recoverable, configuration, yij, summary, notice, possible, edge, total, clustering] [theorem, convex, consider, example, case, provide, bound, appendix, equivalent, algorithm, proof, assume, concentration, considered, general, upper, lower] [log, stochastic, size, likelihood, parameter, large, efficiently] [can, recovery, small, exact, matrix, semidefinite, one, observed, condition, via, recover, see, high, still, spectral, connectivity, following, first, relative, sparse, analysis, computational] [random, given, well, maximum, section, estimator, two, based, statistical] [model, information] [similar, different, dense, understanding, use]
Measuring the reliability of MCMC inference with bidirectional Monte Carlo
Roger B. Grosse, Siddharth Ancha, Daniel M. Roy
Roger B. Grosse, Siddharth Ancha, Daniel M. Roy
[probabilistic, reverse, programming, number, many, partition, quality] [bound, upper, lower, appendix, may, function, show, ratio, obtain, consider, set, since] [log, posterior, inference, approximate, divergence, bread, mcmc, stochastic, jeffreys, monte, stan, carlo, webppl, dkl, method, hyperparameters, bdmc, unbiased, sampling, validate, convergence, expectation, qrev, importance, bidirectional, initial, automatic, rev] [can, one, real, exact, analysis, order, matrix, also, first, nonnegative] [data, distribution, true, section, estimate, sample, chain, given, estimator, two, based, described, statistical, fitted] [simulated, forward, model, state, behavior, target, markov] [using, used, use, accuracy, bounding, evaluate, produced, sequence, different]
Adaptive Maximization of Pointwise Submodular Functions With Budget Constraint
Nguyen Cuong, Huan Xu
Nguyen Cuong, Huan Xu
[possible, partial, constraint, half] [cost, utility, adaptive, set, submodular, function, greedy, submodularity, budget, pointwise, problem, theorem, active, learning, modular, fworst, satisfies, setting, best, optimal, consider, budgeted, general, let, alc, blc, constant, monotone, assume, maximizes, case, guarantee, put, coverage, class, observe, proof, may, show, andreas, will, considered, depends, define, worst] [optimization, factor, select, supplementary] [can, also, item, one, minimal, still, note] [two, data, given, selected, section, uniform, theoretical, generalization] [policy, realization, state, information, value, next] [better, previous, using, table]
NESTT: A Nonconvex Primal-Dual Splitting Method for Distributed and Stochastic Optimization
Davood Hajinezhad, Mingyi Hong, Tuo Zhao, Zhaoran Wang
Davood Hajinezhad, Mingyi Hong, Tuo Zhao, Zhaoran Wang
[variable] [problem, algorithm, convex, function, define, case, rate, consider, will, set, theorem, let, setting, constant, show, learning, achieves, worst] [nonconvex, convergence, gradient, stochastic, distributed, nonsmooth, optimization, saga, nestt, smooth, method, sampling, dual, proximal, iteration, update, rgi, quadratic, incremental, picked, descent, processing, splitting, primal, sgd, converges, siam] [can, following, note, component, linear, suppose, local, optimality, min, assumption, solution, one, gap, also, analysis, matrix, signal, sublinear, see, comparison, related] [based, uniform, given, stationary, section, result, data, journal] [information, agent] [different, randomly, proposed, using, used, generated, neural, recent, table]
Generalized Correspondence-LDA Models (GC-LDA) for Identifying Functional Regions in the Brain
Timothy Rubin, Oluwasanmi O. Koyejo, Michael N. Jones, Tal Yarkoni
Timothy Rubin, Oluwasanmi O. Koyejo, Michael N. Jones, Tal Yarkoni
[number, probability, indicator, variable, assigned, total, david] [function, set, version, equivalent] [constrained, processing, inference] [can, one, via, note, generalized, also, paper, onto, sampled] [word, topic, distribution, functional, subregions, gaussian, peak, token, document, data, subregion, neurosynth, neuroimaging, sample, symmetry, multinomial, two, based, section, associated, mixture, lda, linguistic, correspond, multivariate, refer, given, corresponding, provided, described] [model, brain, cognitive, fmri, unconstrained, human, ability, significant, across, modeling] [spatial, activation, figure, neural, single, three, using, generative, correspondence, capture, database, shown, predicting, language, extracted, novel, used, different, text, region]
A Minimax Approach to Supervised Learning
Farzan Farnia, David Tse
Farzan Farnia, David Tse
[rule, probability, find, variable] [loss, minimax, problem, learning, theorem, function, set, svm, consider, max, minimizing, will, expected, duality, version, logarithmic, convex, optimal, call, erm, defined, risk, show, hinge, define, general, conference, let] [likelihood, quadratic, machine, step, supplementary, marginal, minimizes, divergence, material, logistic, standard] [linear, can, generalized, regression, following, also, solving, error, via, formulation, min] [maximum, distribution, entropy, bayes, conditional, given, principle, empirical, robust, based, specific, journal, statistical, discrete, underlying, denotes, address] [decision, information, model, uncertainty] [approach, figure, prediction, training, classification, using, supervised, propose, applying, task, predict, proposed, different, seen]
Residual Networks Behave Like Ensembles of Relatively Shallow Networks
Andreas Veit, Michael J. Wilber, Serge Belongie
Andreas Veit, Michael J. Wilber, Serge Belongie
[path, number, many, collection, removing, block, valid, ensemble] [show, remaining, even, learning, follows] [gradient, stochastic, depend, strongly, batch, processing] [can, error, one, short, view, first, behave, though, fraction] [length, distribution, test, data, result, sample] [individual, module, long, simple, whether, early, exhibit] [residual, network, neural, figure, deleting, training, layer, deep, depth, effective, single, input, vgg, performance, output, highway, dropping, convolutional, skip, unraveled, building, preprint, impact, lesion, investigate, computer, contribute, like, arxiv, shown, trained, downsampling, use, seen, smoothly, flow, vision, agnitude, perform, visual, imagenet, vanishing, deeper, traditional, work, previous]
Variational Bayes on Monte Carlo Steroids
Aditya Grover, Stefano Ermon
Aditya Grover, Stefano Ermon
[variable, directed, probability, number, parity, probabilistic] [learning, lower, algorithm, bound, will, defined, theorem, set, problem, tight, class, binary] [variational, inference, latent, marginal, log, sampling, posterior, sigmoid, approximate, mod, monte, approximation, carlo, approximating, gradient, amortized, iid, importance, stochastic, optimization, intractable, caltech, key] [can, projected, projection, high] [random, discrete, based, distribution, data, given, family, estimator, true, mean, empirical, test, sample, theoretical] [belief, model, new, median, choice, demonstrate] [using, network, generative, approach, performance, neural, used, use, dataset, single, deep, layer, trained]
Gradient-based Sampling: An Adaptive Importance Sampling for Least-squares
Rong Zhu
Rong Zhu
[according] [algorithm, since, get, bound, cost, fast, risk, make, theorem, problem, improved, show, excess, argue, considered] [sampling, grad, subsample, pilot, lev, size, gradient, large, method, full, importance, efficient, approximating, supplementary, approximate, xti, replacement, mse, unif, initial, optimization, alev, datasets, computing, log, establishes] [can, computational, solution, matrix, error, also, analysis, real, one, much, small, synthetic, running, leverage] [data, estimate, random, sample, section, uniform, mixture, logarithm, given, statistical, empirical, two, statistically, way, journal] [time, model, poisson, simple, information, choice] [performance, figure, input, use, better, table, various, proposed, similar, different]
Stochastic Variational Deep Kernel Learning
Andrew G. Wilson, Zhiting Hu, Ruslan R. Salakhutdinov, Eric P. Xing
Andrew G. Wilson, Zhiting Hu, Ruslan R. Salakhutdinov, Eric P. Xing
[structure, base, number, many, runtime, probabilistic] [learning, show, conference] [variational, stochastic, inducing, inference, large, scalable, marginal, likelihood, sampling, extra, latent, exploiting, interpolation, scalability, gradient, method, full, machine, intelligence, expressive, wilson, processing, hensman, airline, standard, efficient] [can, correlated, matrix, local, see, regression, sparse] [gaussian, kernel, additive, data, covariance, mixing, section, procedure] [model, process, time, information, artificial] [deep, classification, dnn, training, accuracy, neural, layer, output, figure, input, hidden, network, used, approach, proposed, use, performance, learn, trained, multiple, table, using, applied, architecture, preprint, arxiv, international, learned]
Threshold Bandits, With and Without Censored Feedback
Jacob D. Abernethy, Kareem Amin, Ruihao Zhu
Jacob D. Abernethy, Kareem Amin, Ruihao Zhu
[threshold, potential, number, probability, present] [arm, regret, bandit, bound, ucb, algorithm, censored, function, learner, setting, learning, xit, problem, max, round, now, define, let, nit, kmucb, uncensored, assume, expected, lemma, optimistic, mab, will, theorem, set, inequality, proof, consider, playing, show, peter, chosen, confidence, dkwucb, receive, bounded, deviation, drawn] [stochastic, standard, log, end] [can, one, order, observed, note, error, analysis, following, arg, significantly, relies, min] [given, estimator, sample, distribution, estimate, uniform, two, fixed, mean, based, classical, section, empirical, survival, exp, procedure, dark] [value, reward, time, feedback, information, policy, rti, new, pool, payoff, simply] [adversarial, use]
Error Analysis of Generalized Nyström Kernel Regression
Hong Chen, Haifeng Xia, Heng Huang, Weidong Cai
Hong Chen, Haifeng Xia, Heng Huang, Weidong Cai
[subset, coefficient, obtained, theory] [learning, function, hypothesis, fast, theorem, general, rate, bound, dependent, complexity, studied, let, least, optimal, sin, drawn, consider, just, constant, since] [approximation, sampling, method, convergence, select, key, year, parameter, operator, university] [can, analysis, regression, error, regularization, matrix, following, computation, min, symmetric, norm, condition, also, related, computational, paper, column, linear, generalized, square, leverage, rmse] [kernel, gnkr, data, generalization, theoretical, space, empirical, subsampling, positive, knn, sample, associated, given, indefinite, random, satisfactory, two, well, gaussian, fixed, introduce] [continuous, design] [used, previous, performance, table, prediction, different]
Without-Replacement Sampling for Stochastic Gradient Methods
Ohad Shamir
Ohad Shamir
[number, average, split] [algorithm, learning, convex, loss, will, least, bound, function, online, expected, consider, get, regret, smaller, uniformly, rademacher, let, show, complexity, theorem, assume, since, lipschitz, proof, instance, upper, chosen, setting, notation, set] [stochastic, gradient, sampling, machine, svrg, distributed, convergence, optimization, descent, transductive, parameter, suboptimality, size, respect, usually, expectation] [can, also, one, condition, need, note, solution, regularized, assumption, linear, suppose, order, good, analysis, much, require] [data, random, permutation, well, application, result] [individual, communication, required, long, assuming] [using, without, similar, use, randomly, single, several, used, arxiv, preprint, applied]