title
stringlengths
12
138
paper_url
stringlengths
44
58
authors
listlengths
1
22
type
stringclasses
0 values
primary_area
stringclasses
0 values
abstract
large_stringlengths
233
1.97k
keywords
listlengths
0
0
TL;DR
large_stringclasses
0 values
submission_number
int64
1
621
arxiv_id
stringlengths
10
10
arxiv_id_source
stringclasses
2 values
embedding
listlengths
768
768
Improved Regret Bounds for Thompson Sampling in Linear Quadratic Control Problems
https://proceedings.mlr.press/v80/abeille18a.html
[ "Marc Abeille", "Alessandro Lazaric" ]
null
null
Thompson sampling (TS) is an effective approach to trade off exploration and exploration in reinforcement learning. Despite its empirical success and recent advances, its theoretical analysis is often limited to the Bayesian setting, finite state-action spaces, or finite-horizon problems. In this paper, we study an ins...
[]
null
1
null
null
[ -0.037046875804662704, -0.027332479134202003, -0.021953262388706207, 0.045912109315395355, 0.04018595814704895, 0.021726613864302635, 0.02524791657924652, -0.013387707062065601, -0.014934814535081387, -0.057546067982912064, -0.012172832153737545, 0.0010285088792443275, -0.08914127200841904, ...
State Abstractions for Lifelong Reinforcement Learning
https://proceedings.mlr.press/v80/abel18a.html
[ "David Abel", "Dilip Arumugam", "Lucas Lehnert", "Michael Littman" ]
null
null
In lifelong reinforcement learning, agents must effectively transfer knowledge across tasks while simultaneously addressing exploration, credit assignment, and generalization. State abstraction can help overcome these hurdles by compressing the representation used by an agent, thereby reducing the computational and sta...
[]
null
2
null
null
[ -0.0415889248251915, -0.026000428944826126, -0.008969602175056934, 0.0549367293715477, 0.03795163333415985, 0.003121028421446681, -0.009422515518963337, -0.004527810495346785, -0.029295995831489563, -0.01298263855278492, -0.021590005606412888, 0.010960687883198261, -0.060168053954839706, -...
Policy and Value Transfer in Lifelong Reinforcement Learning
https://proceedings.mlr.press/v80/abel18b.html
[ "David Abel", "Yuu Jinnai", "Sophie Yue Guo", "George Konidaris", "Michael Littman" ]
null
null
We consider the problem of how best to use prior experience to bootstrap lifelong learning, where an agent faces a series of task instances drawn from some task distribution. First, we identify the initial policy that optimizes expected performance over the distribution of tasks for increasingly complex classes of poli...
[]
null
3
null
null
[ -0.029713099822402, -0.044526807963848114, -0.005401694215834141, 0.04498424008488655, 0.04659326747059822, 0.028207430616021156, 0.0026968077290803194, -0.004748146515339613, -0.007202092558145523, -0.03036460466682911, -0.023116154596209526, 0.029676059260964394, -0.08016718178987503, -0...
INSPECTRE: Privately Estimating the Unseen
https://proceedings.mlr.press/v80/acharya18a.html
[ "Jayadev Acharya", "Gautam Kamath", "Ziteng Sun", "Huanyu Zhang" ]
null
null
We develop differentially private methods for estimating various distributional properties. Given a sample from a discrete distribution p, some functional f, and accuracy and privacy parameters alpha and epsilon, the goal is to estimate f(p) up to accuracy alpha, while maintaining epsilon-differential privacy of the sa...
[]
null
4
1803.00008
title_snapshot
[ -0.000553494377527386, 0.014844299294054508, -0.013598605059087276, 0.05520664528012276, 0.059472329914569855, 0.046720415353775024, 0.04460032284259796, -0.043695949018001556, -0.016907949000597, -0.012631026096642017, 0.026551708579063416, -0.009230747818946838, -0.062026139348745346, -0...
Learning Representations and Generative Models for 3D Point Clouds
https://proceedings.mlr.press/v80/achlioptas18a.html
[ "Panos Achlioptas", "Olga Diamanti", "Ioannis Mitliagkas", "Leonidas Guibas" ]
null
null
Three-dimensional geometric data offer an excellent domain for studying representation learning and generative modeling. In this paper, we look at geometric data represented as point clouds. We introduce a deep AutoEncoder (AE) network with state-of-the-art reconstruction quality and generalization ability. The learned...
[]
null
5
1707.02392
title_snapshot
[ 0.009283892810344696, -0.004690170753747225, -0.0154506741091609, 0.0705036148428917, 0.034187201410532, 0.0638325959444046, 0.012491954490542412, 0.011114300228655338, -0.018846016377210617, -0.06245461478829384, -0.03805616497993469, -0.03442270681262016, -0.06347919255495071, 0.02417570...
Discovering Interpretable Representations for Both Deep Generative and Discriminative Models
https://proceedings.mlr.press/v80/adel18a.html
[ "Tameem Adel", "Zoubin Ghahramani", "Adrian Weller" ]
null
null
Interpretability of representations in both deep generative and discriminative models is highly desirable. Current methods jointly optimize an objective combining accuracy and interpretability. However, this may reduce accuracy, and is not applicable to already trained models. We propose two interpretability frameworks...
[]
null
6
null
null
[ -0.00920221209526062, -0.01456037349998951, -0.029177002608776093, 0.034687455743551254, 0.02887328527867794, 0.049572959542274475, -0.0071335649117827415, -0.02956179529428482, 0.00791207980364561, -0.04598674550652504, -0.021940946578979492, 0.006745956838130951, -0.06275691092014313, 0....
A Reductions Approach to Fair Classification
https://proceedings.mlr.press/v80/agarwal18a.html
[ "Alekh Agarwal", "Alina Beygelzimer", "Miroslav Dudik", "John Langford", "Hanna Wallach" ]
null
null
We present a systematic approach for achieving fairness in a binary classification setting. While we focus on two well-known quantitative definitions of fairness, our approach encompasses many other previously studied definitions as special cases. The key idea is to reduce fair classification to a sequence of cost-sens...
[]
null
7
1803.02453
title_snapshot
[ -0.03815162181854248, -0.0263633131980896, -0.03480581194162369, 0.030368180945515633, 0.03810488432645798, 0.04605740308761597, -0.008112994022667408, -0.015499421395361423, -0.034813299775123596, -0.021396862342953682, -0.0044806236401200294, -0.003895088564604521, -0.08312619477510452, ...
Accelerated Spectral Ranking
https://proceedings.mlr.press/v80/agarwal18b.html
[ "Arpit Agarwal", "Prathamesh Patil", "Shivani Agarwal" ]
null
null
The problem of rank aggregation from pairwise and multiway comparisons has a wide range of implications, ranging from recommendation systems to sports rankings to social choice. Some of the most popular algorithms for this problem come from the class of spectral ranking algorithms; these include the rank centrality (RC...
[]
null
8
null
null
[ -0.013188640587031841, -0.04076861962676048, 0.00843887496739626, 0.014694426208734512, 0.008405452594161034, -0.015482449904084206, 0.05686153843998909, 0.006540672853589058, -0.029922496527433395, -0.055971816182136536, -0.013866158202290535, 0.008433392271399498, -0.08043842017650604, -...
MISSION: Ultra Large-Scale Feature Selection using Count-Sketches
https://proceedings.mlr.press/v80/aghazadeh18a.html
[ "Amirali Aghazadeh", "Ryan Spring", "Daniel Lejeune", "Gautam Dasarathy", "Anshumali Shrivastava", "baraniuk" ]
null
null
Feature selection is an important challenge in machine learning. It plays a crucial role in the explainability of machine-driven decisions that are rapidly permeating throughout modern society. Unfortunately, the explosion in the size and dimensionality of real-world datasets poses a severe challenge to standard featur...
[]
null
9
1806.04310
title_snapshot
[ -0.019404949620366096, -0.029376117512583733, -0.016490258276462555, 0.024614715948700905, 0.050486478954553604, 0.03904447332024574, 0.04239441454410553, -0.004208107013255358, -0.016035713255405426, -0.04920663684606552, -0.03385891392827034, -0.005276056472212076, -0.07023126631975174, ...
Minimal I-MAP MCMC for Scalable Structure Discovery in Causal DAG Models
https://proceedings.mlr.press/v80/agrawal18a.html
[ "Raj Agrawal", "Caroline Uhler", "Tamara Broderick" ]
null
null
Learning a Bayesian network (BN) from data can be useful for decision-making or discovering causal relationships. However, traditional methods often fail in modern applications, which exhibit a larger number of observed variables than data points. The resulting uncertainty about the underlying network as well as the de...
[]
null
10
1803.05554
title_snapshot
[ -0.015858998522162437, -0.015522850677371025, -0.00526233809068799, 0.029412129893898964, 0.02901385724544525, 0.02524210885167122, 0.049537356942892075, -0.006572067271918058, -0.011439110152423382, -0.0383111834526062, 0.025719579309225082, 0.012146523222327232, -0.044576603919267654, -0...
Proportional Allocation: Simple, Distributed, and Diverse Matching with High Entropy
https://proceedings.mlr.press/v80/agrawal18b.html
[ "Shipra Agrawal", "Morteza Zadimoghaddam", "Vahab Mirrokni" ]
null
null
Inspired by many applications of bipartite matching in online advertising and machine learning, we study a simple and natural iterative proportional allocation algorithm: Maintain a priority score $\priority_a$ for each node $a\in \mathds{A}$ on one side of the bipartition, initialized as $\priority_a=1$. Iteratively a...
[]
null
11
null
null
[ -0.012732725590467453, -0.01795339770615101, -0.01695398986339569, 0.024469314143061638, 0.03377685323357582, 0.058205995708703995, 0.005893825553357601, -0.0031979214400053024, -0.015090994536876678, -0.03904321417212486, -0.026664383709430695, -0.03256939351558685, -0.06448014825582504, ...
Bucket Renormalization for Approximate Inference
https://proceedings.mlr.press/v80/ahn18a.html
[ "Sungsoo Ahn", "Michael Chertkov", "Adrian Weller", "Jinwoo Shin" ]
null
null
Probabilistic graphical models are a key tool in machine learning applications. Computing the partition function, i.e., normalizing constant, is a fundamental task of statistical inference but is generally computationally intractable, leading to extensive study of approximation methods. Iterative variational methods ar...
[]
null
12
1803.05104
title_snapshot
[ 0.022296113893389702, 0.0016725915484130383, -0.01998053304851055, 0.02650570124387741, 0.051443468779325485, 0.028677983209490776, 0.029211336746811867, -0.012541665695607662, -0.036805279552936554, -0.040869951248168945, 0.024741144850850105, -0.0037485151551663876, -0.07497365772724152, ...
oi-VAE: Output Interpretable VAEs for Nonlinear Group Factor Analysis
https://proceedings.mlr.press/v80/ainsworth18a.html
[ "Samuel K. Ainsworth", "Nicholas J. Foti", "Adrian K. C. Lee", "Emily B. Fox" ]
null
null
Deep generative models have recently yielded encouraging results in producing subjectively realistic samples of complex data. Far less attention has been paid to making these generative models interpretable. In many scenarios, ranging from scientific applications to finance, the observed variables have a natural groupi...
[]
null
13
null
null
[ 0.006526377517729998, 0.004512324929237366, 0.022671736776828766, -0.0016305806348100305, 0.02393638715147972, 0.057792503386735916, 0.04865698516368866, 0.013369076885282993, -0.023267731070518494, -0.029309194535017014, -0.0015281159430742264, -0.002781923860311508, -0.05338325724005699, ...
Limits of Estimating Heterogeneous Treatment Effects: Guidelines for Practical Algorithm Design
https://proceedings.mlr.press/v80/alaa18a.html
[ "Ahmed Alaa", "Mihaela Schaar" ]
null
null
Estimating heterogeneous treatment effects from observational data is a central problem in many domains. Because counterfactual data is inaccessible, the problem differs fundamentally from supervised learning, and entails a more complex set of modeling choices. Despite a variety of recently proposed algorithmic solutio...
[]
null
14
null
null
[ -0.018528321757912636, -0.03547737002372742, -0.03206358477473259, 0.008146215230226517, 0.03314825892448425, 0.007161763962358236, 0.05392402783036232, -0.0037068284582346678, -0.010857305489480495, -0.0240924134850502, 0.00723282853141427, -0.00456798542290926, -0.07875029742717743, -0.0...
AutoPrognosis: Automated Clinical Prognostic Modeling via Bayesian Optimization with Structured Kernel Learning
https://proceedings.mlr.press/v80/alaa18b.html
[ "Ahmed Alaa", "Mihaela Schaar" ]
null
null
Clinical prognostic models derived from largescale healthcare data can inform critical diagnostic and therapeutic decisions. To enable off-theshelf usage of machine learning (ML) in prognostic research, we developed AUTOPROGNOSIS: a system for automating the design of predictive modeling pipelines tailored for clinical...
[]
null
15
1802.07207
title_snapshot
[ 0.0013080901699140668, -0.012600107118487358, 0.021960748359560966, -0.009398453868925571, 0.05958611145615578, 0.031788963824510574, 0.029700929298996925, -0.044518716633319855, -0.002755229128524661, -0.033745553344488144, 0.00980242621153593, 0.011025168001651764, -0.022343818098306656, ...
Information Theoretic Guarantees for Empirical Risk Minimization with Applications to Model Selection and Large-Scale Optimization
https://proceedings.mlr.press/v80/alabdulmohsin18a.html
[ "Ibrahim Alabdulmohsin" ]
null
null
In this paper, we derive bounds on the mutual information of the empirical risk minimization (ERM) procedure for both 0-1 and strongly-convex loss classes. We prove that under the Axiom of Choice, the existence of an ERM learning rule with a vanishing mutual information is equivalent to the assertion that the loss clas...
[]
null
16
null
null
[ -0.042978014796972275, 0.002807975746691227, 0.0052178106270730495, 0.041528888046741486, 0.023254400119185448, 0.04959633946418762, 0.037063971161842346, -0.010282807052135468, -0.029209354892373085, -0.026345275342464447, -0.025583239272236824, 0.006002532783895731, -0.048759061843156815, ...
Fixing a Broken ELBO
https://proceedings.mlr.press/v80/alemi18a.html
[ "Alexander Alemi", "Ben Poole", "Ian Fischer", "Joshua Dillon", "Rif A. Saurous", "Kevin Murphy" ]
null
null
Recent work in unsupervised representation learning has focused on learning deep directed latentvariable models. Fitting these models by maximizing the marginal likelihood or evidence is typically intractable, thus a common approximation is to maximize the evidence lower bound (ELBO) instead. However, maximum likelihoo...
[]
null
17
1711.00464
title_snapshot
[ -0.0051942672580480576, -0.028589988127350807, -0.0413200706243515, 0.047597941011190414, 0.03800520673394203, 0.007167898118495941, 0.039099834859371185, -0.004306674003601074, -0.011934674344956875, -0.03294552117586136, -0.020003538578748703, 0.004633171483874321, -0.066587895154953, -0...
Differentially Private Identity and Equivalence Testing of Discrete Distributions
https://proceedings.mlr.press/v80/aliakbarpour18a.html
[ "Maryam Aliakbarpour", "Ilias Diakonikolas", "Ronitt Rubinfeld" ]
null
null
We study the fundamental problems of identity and equivalence testing over a discrete population from random samples. Our goal is to develop efficient testers while guaranteeing differential privacy to the individuals of the population. We provide sample-efficient differentially private testers for these problems. Our ...
[]
null
18
1707.05497
title_judge
[ -0.009046469815075397, 0.020127123221755028, -0.012782709673047066, 0.07760787010192871, 0.06144406646490097, 0.013524886220693588, 0.04186785966157913, -0.02606191858649254, -0.004412070382386446, -0.024166664108633995, 0.03683354705572128, -0.01730348914861679, -0.06494425237178802, -0.0...
Katyusha X: Simple Momentum Method for Stochastic Sum-of-Nonconvex Optimization
https://proceedings.mlr.press/v80/allen-zhu18a.html
[ "Zeyuan Allen-Zhu" ]
null
null
The problem of minimizing sum-of-nonconvex functions (i.e., convex functions that are average of non-convex ones) is becoming increasing important in machine learning, and is the core machinery for PCA, SVD, regularized Newton’s method, accelerated non-convex optimization, and more. We show how to provably obtain an ac...
[]
null
19
1802.03866
title_judge
[ -0.034801483154296875, -0.032213255763053894, 0.0052044992335140705, 0.019510401412844658, 0.01232917606830597, 0.07020476460456848, 0.04000982269644737, 0.01633811742067337, -0.03674948960542679, -0.04342220723628998, -0.02842123433947563, -0.03464154154062271, -0.05219276249408722, -0.00...
Make the Minority Great Again: First-Order Regret Bound for Contextual Bandits
https://proceedings.mlr.press/v80/allen-zhu18b.html
[ "Zeyuan Allen-Zhu", "Sebastien Bubeck", "Yuanzhi Li" ]
null
null
Regret bounds in online learning compare the player’s performance to $L*$, the optimal performance in hindsight with a fixed strategy. Typically such bounds scale with the square root of the time horizon $T$. The more refined concept of first-order regret bound replaces this with a scaling $\sqrt{L*}$, which may be muc...
[]
null
20
1802.03386
title_snapshot
[ -0.044555194675922394, 0.004776113200932741, -0.014116712845861912, 0.03032642975449562, 0.04235508292913437, 0.016160454601049423, 0.03715331479907036, 0.010046965442597866, -0.0390264131128788, -0.03869536519050598, -0.027975376695394516, 0.03513848036527634, -0.06254614889621735, -0.044...
Augmented CycleGAN: Learning Many-to-Many Mappings from Unpaired Data
https://proceedings.mlr.press/v80/almahairi18a.html
[ "Amjad Almahairi", "Sai Rajeshwar", "Alessandro Sordoni", "Philip Bachman", "Aaron Courville" ]
null
null
Learning inter-domain mappings from unpaired data can improve performance in structured prediction tasks, such as image segmentation, by reducing the need for paired data. CycleGAN was recently proposed for this problem, but critically assumes the underlying inter-domain mapping is approximately deterministic and one-t...
[]
null
21
1802.10151
title_snapshot
[ 0.010598232969641685, -0.030760793015360832, -0.03175293654203415, 0.0374661386013031, 0.03188109025359154, 0.009745798073709011, 0.006025101523846388, -0.00022361039009410888, -0.022480901330709457, -0.03568211942911148, -0.007688738871365786, 0.00014074564387556165, -0.06769750267267227, ...
Meta-Learning by Adjusting Priors Based on Extended PAC-Bayes Theory
https://proceedings.mlr.press/v80/amit18a.html
[ "Ron Amit", "Ron Meir" ]
null
null
In meta-learning an agent extracts knowledge from observed tasks, aiming to facilitate learning of novel future tasks. Under the assumption that future tasks are ‘related’ to previous tasks, accumulated knowledge should be learned in such a way that they capture the common structure across learned tasks, while allowing...
[]
null
22
1711.01244
title_snapshot
[ 0.001790389302186668, 0.017364785075187683, -0.00882725603878498, 0.044937875121831894, 0.04234137013554573, 0.03531727194786072, 0.030547982081770897, -0.008794872090220451, -0.05047938972711563, -0.03097151778638363, -0.02520272508263588, 0.03089432790875435, -0.06149879842996597, -0.019...
MAGAN: Aligning Biological Manifolds
https://proceedings.mlr.press/v80/amodio18a.html
[ "Matthew Amodio", "Smita Krishnaswamy" ]
null
null
It is increasingly common in many types of natural and physical systems (especially biological systems) to have different types of measurements performed on the same underlying system. In such settings, it is important to align the manifolds arising from each measurement in order to integrate such data and gain an impr...
[]
null
23
1803.00385
title_snapshot
[ -0.019699621945619583, 0.00034122320357710123, -0.02862829901278019, 0.03416512906551361, 0.016734447330236435, 0.027542024850845337, 0.038081444799900055, 0.009154005907475948, -0.028149083256721497, -0.041507065296173096, 0.0035459198988974094, 0.0026620111893862486, -0.09828247129917145, ...
Subspace Embedding and Linear Regression with Orlicz Norm
https://proceedings.mlr.press/v80/andoni18a.html
[ "Alexandr Andoni", "Chengyu Lin", "Ying Sheng", "Peilin Zhong", "Ruiqi Zhong" ]
null
null
We consider a generalization of the classic linear regression problem to the case when the loss is an Orlicz norm. An Orlicz norm is parameterized by a non-negative convex function G: R_+ - > R_+ with G(0) = 0: the Orlicz norm of a n-dimensional vector x is defined as |x|_G = inf{ alpha > 0 | sum_{i = 1}^n G( |x_i| / a...
[]
null
24
1806.06430
title_snapshot
[ -0.024943538010120392, -0.019949501380324364, 0.03493610396981239, 0.009102225303649902, 0.06869218498468399, 0.02485791966319084, 0.03454681858420372, -0.02092861197888851, -0.0015398155665025115, -0.0464920774102211, -0.048356425017118454, -0.009219365194439888, -0.07095425575971603, -0....
Efficient Gradient-Free Variational Inference using Policy Search
https://proceedings.mlr.press/v80/arenz18a.html
[ "Oleg Arenz", "Gerhard Neumann", "Mingjun Zhong" ]
null
null
Inference from complex distributions is a common problem in machine learning needed for many Bayesian methods. We propose an efficient, gradient-free method for learning general GMM approximations of multimodal distributions based on recent insights from stochastic search methods. Our method establishes information-geo...
[]
null
25
null
null
[ -0.014509905129671097, 0.011603904888033867, -0.0025225188583135605, 0.06419575959444046, 0.028570497408509254, 0.04997671768069267, 0.029948370531201363, 0.018930615857243538, -0.033381424844264984, -0.033057164400815964, -0.004062812775373459, 0.03674857318401337, -0.08827372640371323, 0...
On the Optimization of Deep Networks: Implicit Acceleration by Overparameterization
https://proceedings.mlr.press/v80/arora18a.html
[ "Sanjeev Arora", "Nadav Cohen", "Elad Hazan" ]
null
null
Conventional wisdom in deep learning states that increasing depth improves expressiveness but complicates optimization. This paper suggests that, sometimes, increasing depth can speed up optimization. The effect of depth on optimization is decoupled from expressiveness by focusing on settings where additional layers am...
[]
null
26
1802.06509
title_snapshot
[ -0.03697850927710533, -0.0198042131960392, 0.00786442868411541, 0.034023236483335495, 0.025880420580506325, 0.049175698310136795, 0.03631269931793213, 0.02077462710440159, -0.003850033739581704, -0.05346712842583656, -0.001452375785447657, -0.016571475192904472, -0.033815957605838776, 0.01...
Stronger Generalization Bounds for Deep Nets via a Compression Approach
https://proceedings.mlr.press/v80/arora18b.html
[ "Sanjeev Arora", "Rong Ge", "Behnam Neyshabur", "Yi Zhang" ]
null
null
Deep nets generalize well despite having more parameters than the number of training samples. Recent works try to give an explanation using PAC-Bayes and Margin-based analyses, but do not as yet result in sample complexity bounds better than naive parameter counting. The current paper shows generalization bounds that a...
[]
null
27
1802.05296
title_snapshot
[ 0.0016040803166106343, -0.035448867827653885, -0.02720440924167633, 0.041195161640644073, 0.053559161722660065, 0.047552790492773056, 0.025068752467632294, -0.027117609977722168, -0.01643538847565651, -0.04070413485169411, -0.0028764503076672554, -0.002298913663253188, -0.08407944440841675, ...
Lipschitz Continuity in Model-based Reinforcement Learning
https://proceedings.mlr.press/v80/asadi18a.html
[ "Kavosh Asadi", "Dipendra Misra", "Michael Littman" ]
null
null
We examine the impact of learning Lipschitz continuous models in the context of model-based reinforcement learning. We provide a novel bound on multi-step prediction error of Lipschitz models where we quantify the error using the Wasserstein metric. We go on to prove an error bound for the value-function estimate arisi...
[]
null
28
1804.07193
title_snapshot
[ -0.05574733018875122, -0.010300531052052975, -0.006893948186188936, 0.05110618472099304, 0.05285454913973808, 0.02144668996334076, 0.03604409471154213, 0.021216241642832756, -0.00913135427981615, -0.03822532668709755, -0.02357414737343788, 0.030258547514677048, -0.06821812689304352, -0.017...
Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examples
https://proceedings.mlr.press/v80/athalye18a.html
[ "Anish Athalye", "Nicholas Carlini", "David Wagner" ]
null
null
We identify obfuscated gradients, a kind of gradient masking, as a phenomenon that leads to a false sense of security in defenses against adversarial examples. While defenses that cause obfuscated gradients appear to defeat iterative optimization-based attacks, we find defenses relying on this effect can be circumvente...
[]
null
29
1802.00420
title_snapshot
[ -0.007261656224727631, -0.014345014467835426, -0.005702652502804995, 0.05671807751059532, 0.045603856444358826, 0.005033824127167463, 0.053927794098854065, -0.02058001235127449, -0.015571271069347858, -0.01990250125527382, -0.018230486661195755, -0.020703822374343872, -0.045903440564870834, ...
Synthesizing Robust Adversarial Examples
https://proceedings.mlr.press/v80/athalye18b.html
[ "Anish Athalye", "Logan Engstrom", "Andrew Ilyas", "Kevin Kwok" ]
null
null
Standard methods for generating adversarial examples for neural networks do not consistently fool neural network classifiers in the physical world due to a combination of viewpoint shifts, camera noise, and other natural transformations, limiting their relevance to real-world systems. We demonstrate the existence of ro...
[]
null
30
1707.07397
title_snapshot
[ -0.007577648852020502, 0.012846321798861027, -0.0026439714711159468, 0.049485694617033005, 0.03136782348155975, 0.03096691519021988, 0.01566050946712494, -0.013683883473277092, -0.02646627277135849, -0.04014042764902115, -0.029629355296492577, -0.00881635770201683, -0.059388164430856705, 0...
Contextual Graph Markov Model: A Deep and Generative Approach to Graph Processing
https://proceedings.mlr.press/v80/bacciu18a.html
[ "Davide Bacciu", "Federico Errica", "Alessio Micheli" ]
null
null
We introduce the Contextual Graph Markov Model, an approach combining ideas from generative models and neural networks for the processing of graph data. It founds on a constructive methodology to build a deep architecture comprising layers of probabilistic models that learn to encode the structured information in an in...
[]
null
31
1805.10636
title_snapshot
[ 0.00038487763958983123, 0.011730496771633625, -0.026750095188617706, 0.05643872171640396, 0.034085169434547424, 0.031435322016477585, 0.01898336410522461, 0.03233722224831581, -0.018713656812906265, -0.05420077219605446, -0.00644027441740036, -0.018028436228632927, -0.06479975581169128, -0...
Greed is Still Good: Maximizing Monotone Submodular+Supermodular (BP) Functions
https://proceedings.mlr.press/v80/bai18a.html
[ "Wenruo Bai", "Jeff Bilmes" ]
null
null
We analyze the performance of the greedy algorithm, and also a discrete semi-gradient based algorithm, for maximizing the sum of a suBmodular and suPermodular (BP) function (both of which are non-negative monotone non-decreasing) under two types of constraints, either a cardinality constraint or $p\geq 1$ matroid indep...
[]
null
32
1801.07413
title_judge
[ -0.039921388030052185, -0.007166051771491766, -0.0013974625617265701, 0.06375307589769363, 0.05318065360188484, 0.03655783832073212, -0.003470886265859008, -0.020449277013540268, -0.02193969301879406, -0.037668969482183456, -0.017830654978752136, 0.0170417632907629, -0.07063212245702744, -...
Comparing Dynamics: Deep Neural Networks versus Glassy Systems
https://proceedings.mlr.press/v80/baity-jesi18a.html
[ "Marco Baity-Jesi", "Levent Sagun", "Mario Geiger", "Stefano Spigler", "Gerard Ben Arous", "Chiara Cammarota", "Yann LeCun", "Matthieu Wyart", "Giulio Biroli" ]
null
null
We analyze numerically the training dynamics of deep neural networks (DNN) by using methods developed in statistical physics of glassy systems. The two main issues we address are the complexity of the loss-landscape and of the dynamics within it, and to what extent DNNs share similarities with glassy systems. Our findi...
[]
null
33
1803.06969
title_snapshot
[ -0.0448303185403347, -0.01977563090622425, -0.008348328992724419, 0.0679706558585167, 0.04349041357636452, 0.01502339169383049, 0.01655282825231552, 0.019595107063651085, -0.05136241391301155, -0.0367579385638237, 0.00795330386608839, -0.01844611018896103, -0.03357182815670967, 0.011644710...
SMAC: Simultaneous Mapping and Clustering Using Spectral Decompositions
https://proceedings.mlr.press/v80/bajaj18a.html
[ "Chandrajit Bajaj", "Tingran Gao", "Zihang He", "Qixing Huang", "Zhenxiao Liang" ]
null
null
We introduce a principled approach forsimultaneous mapping and clustering(SMAC) for establishing consistent maps across heterogeneous object collections (e.g., 2D images or 3D shapes). Our approach takes as input a heterogeneous object collection and a set of maps computed between some pairs of objects, and outputs a h...
[]
null
34
null
null
[ -0.011748427525162697, -0.0005847790162079036, -0.007179221138358116, 0.018853245303034782, 0.03452065587043762, 0.07440195232629776, 0.0029423879459500313, -0.008204098790884018, -0.055086031556129456, -0.050655703991651535, -0.04702713340520859, -0.013276435434818268, -0.08123123645782471,...
A Boo(n) for Evaluating Architecture Performance
https://proceedings.mlr.press/v80/bajgar18a.html
[ "Ondrej Bajgar", "Rudolf Kadlec", "Jan Kleindienst" ]
null
null
We point out important problems with the common practice of using the best single model performance for comparing deep learning architectures, and we propose a method that corrects these flaws. Each time a model is trained, one gets a different result due to random factors in the training process, which include random ...
[]
null
35
1807.01961
title_snapshot
[ 0.0042650760151445866, -0.009065280668437481, -0.019378894940018654, 0.024995895102620125, 0.031820666044950485, 0.01370846014469862, 0.04987207055091858, -0.0029038258362561464, 0.002421472454443574, -0.038039132952690125, 0.0032775546424090862, -0.013765518553555012, -0.050760552287101746,...
Learning to Branch
https://proceedings.mlr.press/v80/balcan18a.html
[ "Maria-Florina Balcan", "Travis Dick", "Tuomas Sandholm", "Ellen Vitercik" ]
null
null
Tree search algorithms, such as branch-and-bound, are the most widely used tools for solving combinatorial problems. These algorithms recursively partition the search space to find an optimal solution. To keep the tree small, it is crucial to carefully decide, when expanding a tree node, which variable to branch on at ...
[]
null
36
1803.10150
title_snapshot
[ -0.010151772759854794, -0.021405963227152824, -0.018967390060424805, 0.02803749591112137, 0.06463011354207993, 0.04554907977581024, 0.0213254913687706, -0.0136465635150671, -0.0345931239426136, -0.024311287328600883, -0.023008735850453377, -0.008449388667941093, -0.07674454152584076, -0.00...
The Mechanics of n-Player Differentiable Games
https://proceedings.mlr.press/v80/balduzzi18a.html
[ "David Balduzzi", "Sebastien Racaniere", "James Martens", "Jakob Foerster", "Karl Tuyls", "Thore Graepel" ]
null
null
The cornerstone underpinning deep learning is the guarantee that gradient descent on an objective converges to local minima. Unfortunately, this guarantee fails in settings, such as generative adversarial nets, where there are multiple interacting losses. The behavior of gradient-based methods in games is not well unde...
[]
null
37
1802.05642
title_snapshot
[ -0.06472783535718918, -0.012978524900972843, 0.005871997680515051, 0.032594386488199234, 0.0109671950340271, 0.03132094442844391, 0.010323803871870041, 0.02178865671157837, -0.03699563816189766, -0.054919298738241196, -0.00004420576078700833, 0.003123684786260128, -0.06877032667398453, 0.0...
Spline Filters For End-to-End Deep Learning
https://proceedings.mlr.press/v80/balestriero18a.html
[ "Randall Balestriero", "Romain Cosentino", "Herve Glotin", "Richard Baraniuk" ]
null
null
We propose to tackle the problem of end-to-end learning for raw waveform signals by introducing learnable continuous time-frequency atoms. The derivation of these filters is achieved by defining a functional space with a given smoothness order and boundary conditions. From this space, we derive the parametric analytica...
[]
null
38
null
null
[ -0.019571395590901375, -0.01452683750540018, -0.006005581934005022, 0.01630425825715065, 0.041644081473350525, 0.04740482196211815, 0.016816888004541397, -0.020055271685123444, -0.021082982420921326, -0.0737130343914032, -0.0015691797016188502, -0.010791007429361343, -0.06337360292673111, ...
A Spline Theory of Deep Learning
https://proceedings.mlr.press/v80/balestriero18b.html
[ "Randall Balestriero", "baraniuk" ]
null
null
We build a rigorous bridge between deep networks (DNs) and approximation theory via spline functions and operators. Our key result is that a large class of DNs can be written as a composition ofmax-affine spline operators(MASOs), which provide a powerful portal through which to view and analyze their inner workings. Fo...
[]
null
39
null
null
[ -0.047697730362415314, 0.003741330234333873, 0.009607655927538872, 0.033149171620607376, 0.03687258064746857, 0.05140941962599754, 0.018739603459835052, 0.0010229514446109533, -0.025511592626571655, -0.057197462767362595, 0.007252337411046028, 0.005537538323551416, -0.04665173590183258, 0....
Approximation Guarantees for Adaptive Sampling
https://proceedings.mlr.press/v80/balkanski18a.html
[ "Eric Balkanski", "Yaron Singer" ]
null
null
In this paper we analyze an adaptive sampling approach for submodular maximization. Adaptive sampling is a technique that has recently been shown to achieve a constant factor approximation guarantee for submodular maximization under a cardinality constraint with exponentially fewer adaptive rounds than any previously s...
[]
null
40
null
null
[ -0.024361800402402878, -0.032161541283130646, 0.012701543979346752, 0.045667070895433426, 0.060194142162799835, 0.05383303016424179, 0.018531514331698418, -0.02468183822929859, -0.040920715779066086, -0.05720667913556099, -0.005413097329437733, -0.0018008769256994128, -0.06228823959827423, ...
Improving the Gaussian Mechanism for Differential Privacy: Analytical Calibration and Optimal Denoising
https://proceedings.mlr.press/v80/balle18a.html
[ "Borja Balle", "Yu-Xiang Wang" ]
null
null
The Gaussian mechanism is an essential building block used in multitude of differentially private data analysis algorithms. In this paper we revisit the Gaussian mechanism and show that the original analysis has several important limitations. Our analysis reveals that the variance formula for the original mechanism is ...
[]
null
41
1805.06530
title_snapshot
[ -0.01437692902982235, 0.01952105574309826, -0.0015697265043854713, 0.04687163233757019, 0.04892469570040703, 0.017427237704396248, 0.04188109561800957, -0.042354878038167953, -0.02536228485405445, -0.032264918088912964, 0.0008643802721053362, -0.005659752059727907, -0.05015553906559944, -0...
Dissecting Adam: The Sign, Magnitude and Variance of Stochastic Gradients
https://proceedings.mlr.press/v80/balles18a.html
[ "Lukas Balles", "Philipp Hennig" ]
null
null
The ADAM optimizer is exceedingly popular in the deep learning community. Often it works very well, sometimes it doesn’t. Why? We interpret ADAM as a combination of two aspects: for each weight, the update direction is determined by the sign of stochastic gradients, whereas the update magnitude is determined by an esti...
[]
null
42
1705.07774
title_snapshot
[ -0.015197835862636566, -0.02070561610162258, -0.016544891521334648, 0.013090760447084904, 0.01004408672451973, 0.035594623535871506, 0.042238999158144, 0.02776338905096054, -0.006840039510279894, -0.03572949767112732, -0.020455801859498024, 0.008613510988652706, -0.05683627352118492, -0.00...
Differentially Private Database Release via Kernel Mean Embeddings
https://proceedings.mlr.press/v80/balog18a.html
[ "Matej Balog", "Ilya Tolstikhin", "Bernhard Schölkopf" ]
null
null
We lay theoretical foundations for new database release mechanisms that allow third-parties to construct consistent estimators of population statistics, while ensuring that the privacy of each individual contributing to the database is protected. The proposed framework rests on two main ideas. First, releasing (an esti...
[]
null
43
1710.01641
title_snapshot
[ -0.025034334510564804, -0.0008661719621159136, 0.013988671824336052, 0.07159236818552017, 0.07309094816446304, 0.03746003285050392, 0.036148201674222946, -0.01906808279454708, -0.026812635362148285, -0.019722342491149902, -0.02414010465145111, -0.016216067597270012, -0.060096465051174164, ...
Improving Optimization for Models With Continuous Symmetry Breaking
https://proceedings.mlr.press/v80/bamler18a.html
[ "Robert Bamler", "Stephan Mandt" ]
null
null
Many loss functions in representation learning are invariant under a continuous symmetry transformation. For example, the loss function of word embeddings (Mikolov et al., 2013) remains unchanged if we simultaneously rotate all word and context embedding vectors. We show that representation learning models for time ser...
[]
null
44
1803.03234
title_snapshot
[ -0.014144678600132465, -0.016620172187685966, 0.0152594530954957, 0.03325158357620239, 0.014601599425077438, 0.023899592459201813, 0.03874517232179642, 0.009896029718220234, -0.030130185186862946, -0.022503579035401344, -0.0013969060964882374, -0.010295268148183823, -0.06121661514043808, 0...
Improved Training of Generative Adversarial Networks Using Representative Features
https://proceedings.mlr.press/v80/bang18a.html
[ "Duhyeon Bang", "Hyunjung Shim" ]
null
null
Despite the success of generative adversarial networks (GANs) for image generation, the trade-off between visual quality and image diversity remains a significant issue. This paper achieves both aims simultaneously by improving the stability of training GANs. The key idea of the proposed approach is to implicitly regul...
[]
null
45
1801.09195
title_snapshot
[ -0.005055008921772242, -0.032125674188137054, -0.009961863048374653, 0.030400069430470467, 0.02771805226802826, 0.016921313479542732, 0.029328221455216408, -0.019134560599923134, -0.027093200013041496, -0.0460326112806797, -0.03314804285764694, -0.018736550584435463, -0.06317037343978882, ...
Using Inherent Structures to design Lean 2-layer RBMs
https://proceedings.mlr.press/v80/bansal18a.html
[ "Abhishek Bansal", "Abhinav Anand", "Chiranjib Bhattacharyya" ]
null
null
Understanding the representational power of Restricted Boltzmann Machines (RBMs) with multiple layers is an ill-understood problem and is an area of active research. Motivated from the approach ofInherent Structure formalism(Stillinger & Weber, 1982), extensively used in analysing Spin Glasses, we propose a novel measu...
[]
null
46
1806.04577
title_snapshot
[ -0.03410184010863304, -0.02405422553420067, -0.0059320819564163685, 0.03650764375925064, 0.04618977755308151, 0.00537537457421422, 0.03320367634296417, -0.017604447901248932, -0.04033829644322395, -0.022489236667752266, 0.01305884588509798, -0.02444901503622532, -0.03829813748598099, -0.01...
Classification from Pairwise Similarity and Unlabeled Data
https://proceedings.mlr.press/v80/bao18a.html
[ "Han Bao", "Gang Niu", "Masashi Sugiyama" ]
null
null
Supervised learning needs a huge amount of labeled data, which can be a big bottleneck under the situation where there is a privacy concern or labeling cost is high. To overcome this problem, we propose a new weakly-supervised learning setting where only similar (S) data pairs (two examples belong to the same class) an...
[]
null
47
1802.04381
title_snapshot
[ 0.008918918669223785, -0.030908871442079544, -0.0077938511967659, 0.03784675896167755, 0.037792596966028214, 0.015507313422858715, 0.016634251922369003, -0.03410416096448898, -0.005861666519194841, -0.012276225723326206, -0.01644796133041382, 0.012314780615270138, -0.10244232416152954, 0.0...
Bayesian Optimization of Combinatorial Structures
https://proceedings.mlr.press/v80/baptista18a.html
[ "Ricardo Baptista", "Matthias Poloczek" ]
null
null
The optimization of expensive-to-evaluate black-box functions over combinatorial structures is an ubiquitous task in machine learning, engineering and the natural sciences. The combinatorial explosion of the search space and costly evaluations pose challenges for current techniques in discrete optimization and machine ...
[]
null
48
1806.08838
title_snapshot
[ -0.011035703122615814, 0.000007477259259758284, -0.029897285625338554, 0.04306218773126602, 0.046368956565856934, 0.03664404898881912, 0.010014339350163937, -0.03023812174797058, -0.01476961374282837, -0.036897845566272736, -0.014277145266532898, 0.012690741568803787, -0.057596903294324875, ...
Geodesic Convolutional Shape Optimization
https://proceedings.mlr.press/v80/baque18a.html
[ "Pierre Baque", "Edoardo Remelli", "Francois Fleuret", "Pascal Fua" ]
null
null
Aerodynamic shape optimization has many industrial applications. Existing methods, however, are so computationally demanding that typical engineering practices are to either simply try a limited number of hand-designed shapes or restrict oneself to shapes that can be parameterized using only few degrees of freedom. In ...
[]
null
49
1802.04016
title_snapshot
[ -0.0320465974509716, 0.004066310357302427, 0.03379422426223755, 0.0445457324385643, 0.008293846622109413, 0.07215545326471329, -0.012213768437504768, 0.0191268902271986, -0.018928639590740204, -0.0648985207080841, -0.013440923765301704, -0.039678603410720825, -0.04293699190020561, 0.029227...
Learning to Coordinate with Coordination Graphs in Repeated Single-Stage Multi-Agent Decision Problems
https://proceedings.mlr.press/v80/bargiacchi18a.html
[ "Eugenio Bargiacchi", "Timothy Verstraeten", "Diederik Roijers", "Ann Nowé", "Hado Hasselt" ]
null
null
Learning to coordinate between multiple agents is an important problem in many reinforcement learning problems. Key to learning to coordinate is exploiting loose couplings, i.e., conditional independences between agents. In this paper we study learning in repeated fully cooperative games, multi-agent multi-armed bandit...
[]
null
50
null
null
[ -0.03492383286356926, -0.0012159089092165232, -0.006650214083492756, 0.03391563147306442, 0.028679613023996353, 0.00996523816138506, 0.013616367243230343, -0.006822520401328802, -0.03474004939198494, -0.031376443803310394, 0.0033310779836028814, 0.005614842753857374, -0.07257569581270218, ...
Testing Sparsity over Known and Unknown Bases
https://proceedings.mlr.press/v80/barman18a.html
[ "Siddharth Barman", "Arnab Bhattacharyya", "Suprovat Ghoshal" ]
null
null
Sparsity is a basic property of real vectors that is exploited in a wide variety of machine learning applications. In this work, we describe property testing algorithms for sparsity that observe a low-dimensional projec- tion of the input. We consider two settings. In the first setting, we test sparsity with respect to...
[]
null
51
1608.01275
title_snapshot
[ 0.0015707524726167321, 0.00041857437463477254, 0.022864816710352898, 0.030154243111610413, 0.0477808341383934, 0.03161940723657608, 0.008486411534249783, -0.006369014736264944, -0.034972500056028366, -0.06155182793736458, 0.02514178678393364, -0.005859438329935074, -0.04476986825466156, 0....
Transfer in Deep Reinforcement Learning Using Successor Features and Generalised Policy Improvement
https://proceedings.mlr.press/v80/barreto18a.html
[ "Andre Barreto", "Diana Borsa", "John Quan", "Tom Schaul", "David Silver", "Matteo Hessel", "Daniel Mankowitz", "Augustin Zidek", "Remi Munos" ]
null
null
The ability to transfer skills across tasks has the potential to scale up reinforcement learning (RL) agents to environments currently out of reach. Recently, a framework based on two ideas, successor features (SFs) and generalised policy improvement (GPI), has been introduced as a principled way of transferring skills...
[]
null
52
1901.10964
title_snapshot
[ -0.035448238253593445, -0.04769095405936241, 0.020188234746456146, 0.026454022154211998, 0.046040426939725876, 0.037816740572452545, 0.009505750611424446, -0.022355560213327408, -0.0161171555519104, -0.04598774015903473, -0.02061127871274948, 0.013920344412326813, -0.0682230219244957, -0.0...
Measuring abstract reasoning in neural networks
https://proceedings.mlr.press/v80/barrett18a.html
[ "David Barrett", "Felix Hill", "Adam Santoro", "Ari Morcos", "Timothy Lillicrap" ]
null
null
Whether neural networks can learn abstract reasoning or whether they merely rely on superficial statistics is a topic of recent debate. Here, we propose a dataset and challenge designed to probe abstract reasoning, inspired by a well-known human IQ test. To succeed at this challenge, models must cope with various gener...
[]
null
53
1807.04225
title_snapshot
[ -0.02736896462738514, -0.028980709612369537, 0.015168371610343456, 0.04868423566222191, 0.03472478315234184, 0.014095582999289036, 0.014994087629020214, 0.007581108249723911, -0.032483577728271484, -0.009126272052526474, -0.016953401267528534, 0.03738287463784218, -0.054832860827445984, -0...
Gradient descent with identity initialization efficiently learns positive definite linear transformations by deep residual networks
https://proceedings.mlr.press/v80/bartlett18a.html
[ "Peter Bartlett", "Dave Helmbold", "Philip Long" ]
null
null
We analyze algorithms for approximating a function $f(x) = \Phi x$ mapping $\Re^d$ to $\Re^d$ using deep linear neural networks, i.e. that learn a function $h$ parameterized by matrices $\Theta_1,...,\Theta_L$ and defined by $h(x) = \Theta_L \Theta_{L-1} ... \Theta_1 x$. We focus on algorithms that learn through gradie...
[]
null
54
1802.06093
title_snapshot
[ -0.011379812844097614, -0.006562963128089905, 0.021775878965854645, 0.02379787154495716, 0.048259008675813675, 0.04734694957733154, 0.01661848835647106, -0.029538804665207863, -0.013647310435771942, -0.02512645721435547, -0.016932522878050804, 0.0003574452712200582, -0.05862567946314812, 0...
Mutual Information Neural Estimation
https://proceedings.mlr.press/v80/belghazi18a.html
[ "Mohamed Ishmael Belghazi", "Aristide Baratin", "Sai Rajeshwar", "Sherjil Ozair", "Yoshua Bengio", "Aaron Courville", "Devon Hjelm" ]
null
null
We argue that the estimation of mutual information between high dimensional continuous random variables can be achieved by gradient descent over neural networks. We present a Mutual Information Neural Estimator (MINE) that is linearly scalable in dimensionality as well as in sample size, trainable through back-prop, an...
[]
null
55
null
null
[ -0.004040202125906944, -0.003760498482733965, -0.024537093937397003, 0.04746200144290924, 0.022188153117895126, 0.04153675213456154, 0.036021918058395386, -0.001801751321181655, -0.009571975097060204, -0.024303067475557327, 0.006602917332202196, 0.020974203944206238, -0.062143225222826004, ...
To Understand Deep Learning We Need to Understand Kernel Learning
https://proceedings.mlr.press/v80/belkin18a.html
[ "Mikhail Belkin", "Siyuan Ma", "Soumik Mandal" ]
null
null
Generalization performance of classifiers in deep learning has recently become a subject of intense study. Deep models, which are typically heavily over-parametrized, tend to fit the training data exactly. Despite this “overfitting", they perform well on test data, a phenomenon not yet fully understood. The first point...
[]
null
56
1802.01396
title_snapshot
[ -0.026079485192894936, -0.03944740444421768, 0.009799077175557613, 0.06317108124494553, 0.04729840159416199, 0.024582011625170708, 0.016852041706442833, 0.003841131692752242, -0.005430371966212988, -0.031484246253967285, -0.013574719429016113, 0.015499407425522804, -0.06115247681736946, 0....
Understanding and Simplifying One-Shot Architecture Search
https://proceedings.mlr.press/v80/bender18a.html
[ "Gabriel Bender", "Pieter-Jan Kindermans", "Barret Zoph", "Vijay Vasudevan", "Quoc Le" ]
null
null
There is growing interest in automating neural network architecture design. Existing architecture search methods can be computationally expensive, requiring thousands of different architectures to be trained from scratch. Recent work has exploredweight sharingacross models to amortize the cost of training. Although pre...
[]
null
57
null
null
[ -0.004026991780847311, -0.0165964774787426, -0.007590155582875013, 0.051055967807769775, 0.059201594442129135, 0.023525405675172806, 0.02142167277634144, -0.01403494831174612, -0.008330618031322956, -0.049161024391651154, -0.007803867105394602, -0.03392089158296585, -0.053645163774490356, ...
signSGD: Compressed Optimisation for Non-Convex Problems
https://proceedings.mlr.press/v80/bernstein18a.html
[ "Jeremy Bernstein", "Yu-Xiang Wang", "Kamyar Azizzadenesheli", "Animashree Anandkumar" ]
null
null
Training large neural networks requires distributing learning across multiple workers, where the cost of communicating gradients can be a significant bottleneck. signSGD alleviates this problem by transmitting just the sign of each minibatch stochastic gradient. We prove that it can get the best of both worlds: compres...
[]
null
58
1802.04434
title_snapshot
[ -0.005761210341006517, -0.039228569716215134, 0.004956952296197414, 0.03328252211213112, 0.008346573449671268, 0.08371677249670029, 0.010680037550628185, 0.026568466797471046, -0.031657833606004715, -0.03556203469634056, -0.009812523610889912, -0.018285172060132027, -0.06390006095170975, -...
Distributed Clustering via LSH Based Data Partitioning
https://proceedings.mlr.press/v80/bhaskara18a.html
[ "Aditya Bhaskara", "Maheshakya Wijewardena" ]
null
null
Given the importance of clustering in the analysisof large scale data, distributed algorithms for formulations such as k-means, k-median, etc. have been extensively studied. A successful approach here has been the “reduce and merge” paradigm, in which each machine reduces its input size to {Õ}(k), and this data reducti...
[]
null
59
null
null
[ -0.031752992421388626, -0.0334770604968071, -0.011341880075633526, 0.032251279801130295, 0.05104469507932663, 0.046743039041757584, 0.014129867777228355, -0.007079260889440775, -0.01893998123705387, -0.05084257572889328, 0.018648307770490646, -0.04611540213227272, -0.0512298084795475, 0.03...
Autoregressive Convolutional Neural Networks for Asynchronous Time Series
https://proceedings.mlr.press/v80/binkowski18a.html
[ "Mikolaj Binkowski", "Gautier Marti", "Philippe Donnat" ]
null
null
We propose Significance-Offset Convolutional Neural Network, a deep convolutional network architecture for regression of multivariate asynchronous time series. The model is inspired by standard autoregressive (AR) models and gating mechanisms used in recurrent neural networks. It involves an AR-like weighting system, w...
[]
null
60
1703.04122
title_snapshot
[ 0.021280605345964432, -0.03456370532512665, 0.006803452502936125, 0.02348056621849537, 0.02457301691174507, 0.07663653790950775, 0.033001065254211426, 0.024402040988206863, -0.010654335841536522, -0.04459359124302864, 0.023269150406122208, -0.0028318625409156084, -0.0748765841126442, 0.022...
Adaptive Sampled Softmax with Kernel Based Sampling
https://proceedings.mlr.press/v80/blanc18a.html
[ "Guy Blanc", "Steffen Rendle" ]
null
null
Softmax is the most commonly used output function for multiclass problems and is widely used in areas such as vision, natural language processing, and recommendation. A softmax model has linear costs in the number of classes which makes it too expensive for many real-world problems. A common approach to speed up traini...
[]
null
61
1712.00527
title_snapshot
[ -0.030657244846224785, -0.042428094893693924, 0.0343196801841259, 0.042311299592256546, 0.035980526357889175, 0.016976123675704002, 0.001243041013367474, -0.0064428094774484634, -0.018571726977825165, -0.06656737625598907, -0.030195072293281555, 0.012242899276316166, -0.07098571956157684, ...
Optimizing the Latent Space of Generative Networks
https://proceedings.mlr.press/v80/bojanowski18a.html
[ "Piotr Bojanowski", "Armand Joulin", "David Lopez-Pas", "Arthur Szlam" ]
null
null
Generative Adversarial Networks (GANs) have achieved remarkable results in the task of generating realistic natural images. In most successful applications, GAN models share two common aspects: solving a challenging saddle point optimization problem, interpreted as an adversarial game between a generator and a discrimi...
[]
null
62
1707.05776
title_snapshot
[ -0.00834494736045599, -0.02664790488779545, -0.011392517015337944, 0.054123274981975555, 0.03159942477941513, 0.03834301978349686, -0.000886248832102865, 0.013509857468307018, 0.008862887509167194, -0.05449458956718445, -0.005702861584722996, -0.012563957832753658, -0.0599365159869194, 0.0...
NetGAN: Generating Graphs via Random Walks
https://proceedings.mlr.press/v80/bojchevski18a.html
[ "Aleksandar Bojchevski", "Oleksandr Shchur", "Daniel Zügner", "Stephan Günnemann" ]
null
null
We propose NetGAN - the first implicit generative model for graphs able to mimic real-world networks. We pose the problem of graph generation as learning the distribution of biased random walks over the input graph. The proposed model is based on a stochastic neural network that generates discrete output samples and is...
[]
null
63
1803.00816
title_snapshot
[ 0.004598681349307299, -0.04099036753177643, -0.004874012432992458, 0.035551100969314575, 0.006082477048039436, 0.005918737966567278, 0.0030766665004193783, 0.02875087969005108, -0.003926643170416355, -0.06199530512094498, 0.00043029876542277634, -0.03738599643111229, -0.07024933397769928, ...
A Progressive Batching L-BFGS Method for Machine Learning
https://proceedings.mlr.press/v80/bollapragada18a.html
[ "Raghu Bollapragada", "Jorge Nocedal", "Dheevatsa Mudigere", "Hao-Jun Shi", "Ping Tak Peter Tang" ]
null
null
The standard L-BFGS method relies on gradient approximations that are not dominated by noise, so that search directions are descent directions, the line search is reliable, and quasi-Newton updating yields useful quadratic models of the objective function. All of this appears to call for a full batch approach, but sinc...
[]
null
64
1802.05374
title_snapshot
[ -0.035608671605587006, -0.02066941186785698, -0.007616119459271431, -0.02456582710146904, 0.04550609365105629, 0.06849128007888794, 0.017505688592791557, 0.017332354560494423, -0.0407782644033432, -0.053224433213472366, -0.006281447131186724, 0.002145129721611738, -0.06279164552688599, -0....
Prediction Rule Reshaping
https://proceedings.mlr.press/v80/bonakdarpour18a.html
[ "Matt Bonakdarpour", "Sabyasachi Chatterjee", "Rina Foygel Barber", "John Lafferty" ]
null
null
Two methods are proposed for high-dimensional shape-constrained regression and classification. These methods reshape pre-trained prediction rules to satisfy shape constraints like monotonicity and convexity. The first method can be applied to any pre-trained prediction rule, while the second method deals specifically w...
[]
null
65
1805.06439
title_snapshot
[ -0.011762934736907482, -0.025585556402802467, -0.011332734487950802, 0.012891456484794617, 0.050554294139146805, 0.06930531561374664, 0.03791753947734833, -0.03447754308581352, -0.039774078875780106, -0.05021336302161217, -0.018302861601114273, -0.004279029555618763, -0.08417002856731415, ...
QuantTree: Histograms for Change Detection in Multivariate Data Streams
https://proceedings.mlr.press/v80/boracchi18a.html
[ "Giacomo Boracchi", "Diego Carrera", "Cristiano Cervellera", "Danilo Macciò" ]
null
null
We address the problem of detecting distribution changes in multivariate data streams by means of histograms. Histograms are very general and flexible models, which have been relatively ignored in the change-detection literature as they often require a number of bins that grows unfeasibly with the data dimension. We pr...
[]
null
66
null
null
[ 0.005512619391083717, -0.0576305016875267, -0.03446803614497185, 0.04082280769944191, 0.04404279217123985, 0.05487942695617676, -0.0025869521778076887, -0.008653008379042149, -0.04609829932451248, -0.06298330426216125, -0.04480425640940666, 0.022783078253269196, -0.05805319920182228, -0.00...
Matrix Norms in Data Streams: Faster, Multi-Pass and Row-Order
https://proceedings.mlr.press/v80/braverman18a.html
[ "Vladimir Braverman", "Stephen Chestnut", "Robert Krauthgamer", "Yi Li", "David Woodruff", "Lin Yang" ]
null
null
A central problem in mining massive data streams is characterizing which functions of an underlying frequency vector can be approximated efficiently. Given the prevalence of large scale linear algebra problems in machine learning, recently there has been considerable effort in extending this data stream problem to that...
[]
null
67
1609.05885
title_snapshot
[ -0.009311710484325886, -0.02288789115846157, 0.019259115681052208, 0.011989312246441841, 0.028993993997573853, 0.008826570585370064, 0.024606317281723022, -0.002735347021371126, -0.028610171750187874, -0.017837470397353172, -0.01898818649351597, -0.0001572071632836014, -0.08193888515233994, ...
Predict and Constrain: Modeling Cardinality in Deep Structured Prediction
https://proceedings.mlr.press/v80/brukhim18a.html
[ "Nataly Brukhim", "Amir Globerson" ]
null
null
Many machine learning problems require the prediction of multi-dimensional labels. Such structured prediction models can benefit from modeling dependencies between labels. Recently, several deep learning approaches to structured prediction have been proposed. Here we focus on capturing cardinality constraints in such m...
[]
null
68
1802.04721
title_snapshot
[ -0.005256191827356815, -0.04206487908959389, -0.02193189226090908, 0.02149823307991028, 0.04416794329881668, 0.02485043928027153, 0.005263300612568855, -0.017614595592021942, -0.02169157937169075, -0.022878816351294518, -0.018552083522081375, 0.03070163168013096, -0.06473414599895477, 0.01...
Quasi-Monte Carlo Variational Inference
https://proceedings.mlr.press/v80/buchholz18a.html
[ "Alexander Buchholz", "Florian Wenzel", "Stephan Mandt" ]
null
null
Many machine learning problems involve Monte Carlo gradient estimators. As a prominent example, we focus on Monte Carlo variational inference (MCVI) in this paper. The performance of MCVI crucially depends on the variance of its stochastic gradients. We propose variance reduction by means of Quasi-Monte Carlo (QMC) sam...
[]
null
69
1807.01604
title_snapshot
[ 0.005053242202848196, 0.00183450139593333, 0.0061429147608578205, 0.05273038521409035, 0.042902324348688126, 0.03642076998949051, 0.0216651801019907, -0.003178420476615429, -0.03269163891673088, -0.051179103553295135, -0.0008179516298696399, 0.015257325023412704, -0.0689091607928276, 0.007...
Path-Level Network Transformation for Efficient Architecture Search
https://proceedings.mlr.press/v80/cai18a.html
[ "Han Cai", "Jiacheng Yang", "Weinan Zhang", "Song Han", "Yong Yu" ]
null
null
We introduce a new function-preserving transformation for efficient neural architecture search. This network transformation allows reusing previously trained networks and existing successful architectures that improves sample efficiency. We aim to address the limitation of current network transformation operations that...
[]
null
70
1806.02639
title_snapshot
[ -0.01590346172451973, -0.027350371703505516, 0.011378796771168709, 0.0392567440867424, 0.049970440566539764, 0.03596682846546173, 0.022227603942155838, 0.001988261239603162, -0.0014849381987005472, -0.03931808844208717, -0.00021055068646091968, -0.022007964551448822, -0.049245648086071014, ...
Improved large-scale graph learning through ridge spectral sparsification
https://proceedings.mlr.press/v80/calandriello18a.html
[ "Daniele Calandriello", "Alessandro Lazaric", "Ioannis Koutis", "Michal Valko" ]
null
null
The representation and learning benefits of methods based on graph Laplacians, such as Laplacian smoothing or harmonic function solution for semi-supervised learning (SSL), are empirically and theoretically well supported. Nonetheless, the exact versions of these methods scale poorly with the number of nodes $n$ of the...
[]
null
71
2604.20078
title_snapshot
[ -0.009133861400187016, -0.04446606710553169, 0.025366846472024918, 0.03274093568325043, 0.03960968554019928, 0.03364623337984085, 0.022527696564793587, -0.011757238768041134, -0.02222398668527603, -0.05772016942501068, 0.02124619111418724, -0.01516359206289053, -0.07842475920915604, 0.0150...
Bayesian Coreset Construction via Greedy Iterative Geodesic Ascent
https://proceedings.mlr.press/v80/campbell18a.html
[ "Trevor Campbell", "Tamara Broderick" ]
null
null
Coherent uncertainty quantification is a key strength of Bayesian methods. But modern algorithms for approximate Bayesian posterior inference often sacrifice accurate posterior uncertainty estimation in the pursuit of scalability. This work shows that previous Bayesian coreset construction algorithms—which build a smal...
[]
null
72
1802.01737
title_snapshot
[ -0.0045280009508132935, 0.0002833317848853767, -0.0018273175228387117, 0.048425476998090744, 0.03534611314535141, 0.058823928236961365, 0.013769739307463169, 0.004812315106391907, -0.016249535605311394, -0.06323300302028656, -0.016300179064273834, -0.009735229425132275, -0.07349368184804916,...
Adversarial Learning with Local Coordinate Coding
https://proceedings.mlr.press/v80/cao18a.html
[ "Jiezhang Cao", "Yong Guo", "Qingyao Wu", "Chunhua Shen", "Junzhou Huang", "Mingkui Tan" ]
null
null
Generative adversarial networks (GANs) aim to generate realistic data from some prior distribution (e.g., Gaussian noises). However, such prior distribution is often independent of real data and thus may lose semantic information (e.g., geometric structure or content in images) of data. In practice, the semantic inform...
[]
null
73
1806.04895
title_snapshot
[ -0.00001334843454969814, -0.029044145718216896, -0.015114360488951206, 0.07005082070827484, 0.03249821066856384, 0.02110949717462063, -0.021159861236810684, 0.0027709996793419123, 0.007294533308595419, -0.0632614716887474, -0.023434897884726524, -0.023153407499194145, -0.06577980518341064, ...
Fair and Diverse DPP-Based Data Summarization
https://proceedings.mlr.press/v80/celis18a.html
[ "Elisa Celis", "Vijay Keswani", "Damian Straszak", "Amit Deshpande", "Tarun Kathuria", "Nisheeth Vishnoi" ]
null
null
Sampling methods that choose a subset of the data proportional to its diversity in the feature space are popular for data summarization. However, recent studies have noted the occurrence of bias {–} e.g., under or over representation of a particular gender or ethnicity {–} in such data summarization methods. In this pa...
[]
null
74
1802.04023
title_snapshot
[ -0.01726652681827545, -0.05134433135390282, -0.017696673050522804, 0.06773313134908676, 0.02847248502075672, 0.05058274790644646, 0.01196517888456583, -0.019557176157832146, -0.04278520494699478, -0.035898447036743164, -0.026496846228837967, -0.015055069699883461, -0.1063644140958786, -0.0...
Conditional Noise-Contrastive Estimation of Unnormalised Models
https://proceedings.mlr.press/v80/ceylan18a.html
[ "Ciwan Ceylan", "Michael U. Gutmann" ]
null
null
Many parametric statistical models are not properly normalised and only specified up to an intractable partition function, which renders parameter estimation difficult. Examples of unnormalised models are Gibbs distributions, Markov random fields, and neural network models in unsupervised deep learning. In previous wor...
[]
null
75
1806.03664
title_snapshot
[ -0.005653658881783485, -0.01630636677145958, -0.030709825456142426, 0.020680895075201988, 0.03509237617254257, 0.05646434426307678, 0.033173564821481705, -0.004314749035984278, -0.0301954485476017, -0.04762071743607521, -0.012043232098221779, 0.012125998735427856, -0.06116024777293205, 0.0...
Adversarial Time-to-Event Modeling
https://proceedings.mlr.press/v80/chapfuwa18a.html
[ "Paidamoyo Chapfuwa", "Chenyang Tao", "Chunyuan Li", "Courtney Page", "Benjamin Goldstein", "Lawrence Carin Duke", "Ricardo Henao" ]
null
null
Modern health data science applications leverage abundant molecular and electronic health data, providing opportunities for machine learning to build statistical models to support clinical practice. Time-to-event analysis, also called survival analysis, stands as one of the most representative examples of such statisti...
[]
null
76
1804.03184
title_snapshot
[ 0.012491283938288689, -0.02768212929368019, -0.04881741479039192, 0.0277198888361454, 0.043141499161720276, 0.033770348876714706, 0.034118860960006714, -0.010942059569060802, 0.0028026532381772995, -0.045397527515888214, 0.021622909232974052, -0.008568362332880497, -0.048491254448890686, 0...
Stability and Generalization of Learning Algorithms that Converge to Global Optima
https://proceedings.mlr.press/v80/charles18a.html
[ "Zachary Charles", "Dimitris Papailiopoulos" ]
null
null
We establish novel generalization bounds for learning algorithms that converge to global minima. We derive black-box stability results that only depend on the convergence of a learning algorithm and the geometry around the minimizers of the empirical risk function. The results are shown for non-convex loss functions sa...
[]
null
77
1710.08402
title_snapshot
[ -0.0298627819865942, -0.0068757906556129456, 0.011524220928549767, 0.0318542942404747, 0.03461939096450806, 0.041641756892204285, 0.02499089203774929, -0.0020681333262473345, -0.03310798481106758, -0.035412561148405075, -0.012981472536921501, -0.007848822511732578, -0.08494903147220612, 0....
Learning and Memorization
https://proceedings.mlr.press/v80/chatterjee18a.html
[ "Satrajit Chatterjee" ]
null
null
In the machine learning research community, it is generally believed that there is a tension between memorization and generalization. In this work we examine to what extent this tension exists by exploring if it is possible to generalize by memorizing alone. Although direct memorization with a lookup table obviously do...
[]
null
78
null
null
[ -0.027008377015590668, 0.010567549616098404, -0.014659266918897629, 0.03688686341047287, 0.06217102333903313, -0.009744498878717422, 0.04038652777671814, 0.018331727012991905, -0.04202384501695633, -0.030111799016594887, 0.005275109317153692, 0.020259542390704155, -0.05510809272527695, -0....
On the Theory of Variance Reduction for Stochastic Gradient Monte Carlo
https://proceedings.mlr.press/v80/chatterji18a.html
[ "Niladri Chatterji", "Nicolas Flammarion", "Yian Ma", "Peter Bartlett", "Michael Jordan" ]
null
null
We provide convergence guarantees in Wasserstein distance for a variety of variance-reduction methods: SAGA Langevin diffusion, SVRG Langevin diffusion and control-variate underdamped Langevin diffusion. We analyze these methods under a uniform set of assumptions on the log-posterior distribution, assuming it to be smo...
[]
null
79
1802.05431
title_snapshot
[ -0.02080252207815647, 0.007489309646189213, 0.02106139063835144, 0.05253930762410164, 0.0363122932612896, 0.022517243400216103, 0.04584014415740967, 0.0010811099782586098, -0.017765244469046593, -0.0693395808339119, 0.023058630526065826, 0.01751333475112915, -0.03331131488084793, -0.002278...
Hierarchical Clustering with Structural Constraints
https://proceedings.mlr.press/v80/chatziafratis18a.html
[ "Vaggos Chatziafratis", "Rad Niazadeh", "Moses Charikar" ]
null
null
Hierarchical clustering is a popular unsupervised data analysis method. For many real-world applications, we would like to exploit prior information about the data that imposes constraints on the clustering hierarchy, and is not captured by the set of features available to the algorithm. This gives rise to the problem ...
[]
null
80
1805.09476
title_snapshot
[ -0.006247303448617458, 0.003113897517323494, -0.006447473540902138, 0.03422592952847481, 0.044318974018096924, 0.02442788891494274, 0.013867869973182678, -0.033165834844112396, -0.02800232730805874, -0.025602353736758232, -0.008099270984530449, -0.03076641634106636, -0.07284614443778992, 0...
Hierarchical Deep Generative Models for Multi-Rate Multivariate Time Series
https://proceedings.mlr.press/v80/che18a.html
[ "Zhengping Che", "Sanjay Purushotham", "Guangyu Li", "Bo Jiang", "Yan Liu" ]
null
null
Multi-Rate Multivariate Time Series (MR-MTS) are the multivariate time series observations which come with various sampling rates and encode multiple temporal dependencies. State-space models such as Kalman filters and deep learning models such as deep Markov models are mainly designed for time series data with the sam...
[]
null
81
null
null
[ -0.016220945864915848, -0.01674794778227806, 0.0013193392660468817, 0.04672309011220932, 0.04364002123475075, 0.03958190977573395, 0.03891092538833618, 0.011896755546331406, -0.0031734437216073275, -0.040439046919345856, 0.021419361233711243, 0.03330148011445999, -0.05167738348245621, 0.03...
GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networks
https://proceedings.mlr.press/v80/chen18a.html
[ "Zhao Chen", "Vijay Badrinarayanan", "Chen-Yu Lee", "Andrew Rabinovich" ]
null
null
Deep multitask networks, in which one neural network produces multiple predictive outputs, can offer better speed and performance than their single-task counterparts but are challenging to train properly. We present a gradient normalization (GradNorm) algorithm that automatically balances training in deep multitask mod...
[]
null
82
1711.02257
title_snapshot
[ -0.008260740898549557, -0.029722684994339943, 0.0029938407242298126, 0.022227952256798744, 0.04368599131703377, 0.05745215341448784, 0.018062030896544456, -0.006543816067278385, -0.028210610151290894, -0.041837092489004135, -0.006507352460175753, -0.0015863657463341951, -0.06331513077020645,...
Weakly Submodular Maximization Beyond Cardinality Constraints: Does Randomization Help Greedy?
https://proceedings.mlr.press/v80/chen18b.html
[ "Lin Chen", "Moran Feldman", "Amin Karbasi" ]
null
null
Submodular functions are a broad class of set functions that naturally arise in many machine learning applications. Due to their combinatorial structures, there has been a myriad of algorithms for maximizing such functions under various constraints. Unfortunately, once a function deviates from submodularity (even sligh...
[]
null
83
1707.04347
title_snapshot
[ -0.011420654132962227, -0.04292776435613632, -0.006158137694001198, 0.04435183107852936, 0.042297814041376114, 0.03916516900062561, 0.01688578352332115, -0.034500062465667725, -0.021143224090337753, -0.02602425403892994, -0.01654069684445858, 0.016483495011925697, -0.07335323095321655, 0.0...
Projection-Free Online Optimization with Stochastic Gradient: From Convexity to Submodularity
https://proceedings.mlr.press/v80/chen18c.html
[ "Lin Chen", "Christopher Harshaw", "Hamed Hassani", "Amin Karbasi" ]
null
null
Online optimization has been a successful framework for solving large-scale problems under computational constraints and partial information. Current methods for online convex optimization require either a projection or exact gradient computation at each step, both of which can be prohibitively expensive for large-scal...
[]
null
84
1802.08183
title_snapshot
[ -0.025674927979707718, -0.011644726619124413, 0.028103049844503403, 0.0366104356944561, 0.0348394550383091, 0.0399969257414341, 0.020174620673060417, 0.0011698264861479402, -0.005393621977418661, -0.031500257551670074, -0.018315937370061874, -0.0017139369156211615, -0.05961739644408226, -0...
Continuous-Time Flows for Efficient Inference and Density Estimation
https://proceedings.mlr.press/v80/chen18d.html
[ "Changyou Chen", "Chunyuan Li", "Liqun Chen", "Wenlin Wang", "Yunchen Pu", "Lawrence Carin Duke" ]
null
null
Two fundamental problems in unsupervised learning are efficient inference for latent-variable models and robust density estimation based on large amounts of unlabeled data. Algorithms for the two tasks, such as normalizing flows and generative adversarial networks (GANs), are often developed independently. In this pape...
[]
null
85
1709.01179
title_snapshot
[ 0.02373977191746235, -0.03910282999277115, -0.003055596724152565, 0.06383208185434341, 0.031849008053541183, 0.019328750669956207, 0.004898691084235907, 0.004577754531055689, 0.017547134310007095, -0.06302590668201447, 0.014491509646177292, -0.033771052956581116, -0.06380073726177216, 0.01...
Scalable Bilinear Pi Learning Using State and Action Features
https://proceedings.mlr.press/v80/chen18e.html
[ "Yichen Chen", "Lihong Li", "Mengdi Wang" ]
null
null
Approximate linear programming (ALP) represents one of the major algorithmic families to solve large-scale Markov decision processes (MDP). In this work, we study a primal-dual formulation of the ALP, and develop a scalable, model-free algorithm called bilinear $\pi$ learning for reinforcement learning when a sampling ...
[]
null
86
null
null
[ -0.05116851627826691, -0.01771567016839981, -0.009982415474951267, 0.01070543471723795, 0.020897535607218742, 0.027423344552516937, 0.006191940512508154, -0.014706638641655445, -0.003268406493589282, -0.043648287653923035, 0.003035055473446846, -0.013374035246670246, -0.0696486085653305, 0...
Stein Points
https://proceedings.mlr.press/v80/chen18f.html
[ "Wilson Ye Chen", "Lester Mackey", "Jackson Gorham", "Francois-Xavier Briol", "Chris Oates" ]
null
null
An important task in computational statistics and machine learning is to approximate a posterior distribution $p(x)$ with an empirical measure supported on a set of representative points $\{x_i\}_{i=1}^n$. This paper focuses on methods where the selection of points is essentially deterministic, with an emphasis on achi...
[]
null
87
1803.10161
title_snapshot
[ -0.046410802751779556, 0.00957256555557251, 0.017329279333353043, 0.020474934950470924, 0.030307458713650703, 0.060390815138816833, 0.01960594952106476, -0.02709224820137024, -0.03151905536651611, -0.051520902663469315, -0.013857711106538773, -0.02365727350115776, -0.056524571031332016, -0...
Learning K-way D-dimensional Discrete Codes for Compact Embedding Representations
https://proceedings.mlr.press/v80/chen18g.html
[ "Ting Chen", "Martin Renqiang Min", "Yizhou Sun" ]
null
null
Conventional embedding methods directly associate each symbol with a continuous embedding vector, which is equivalent to applying a linear transformation based on a “one-hot” encoding of the discrete symbols. Despite its simplicity, such approach yields the number of parameters that grows linearly with the vocabulary s...
[]
null
88
1806.09464
title_snapshot
[ -0.010097461752593517, -0.03320877254009247, -0.0034591183066368103, 0.05051103234291077, 0.03773565962910652, 0.036860864609479904, 0.01901000365614891, -0.0009327212464995682, 0.016857216134667397, -0.04149571433663368, -0.014445275068283081, -0.03322455286979675, -0.05262034758925438, 0...
PixelSNAIL: An Improved Autoregressive Generative Model
https://proceedings.mlr.press/v80/chen18h.html
[ "XI Chen", "Nikhil Mishra", "Mostafa Rohaninejad", "Pieter Abbeel" ]
null
null
Autoregressive generative models achieve the best results in density estimation tasks involving high dimensional data, such as images or audio. They pose density estimation as a sequence modeling task, where a recurrent neural network (RNN) models the conditional distribution over the next element conditioned on all pr...
[]
null
89
1712.09763
title_snapshot
[ 0.014110961928963661, -0.051166076213121414, -0.02248510532081127, 0.06362731754779816, 0.029468504711985588, 0.06512833386659622, 0.018031563609838486, 0.032003823667764664, -0.04621574282646179, -0.051880888640880585, -0.0072160582058131695, -0.00902589876204729, -0.08223719149827957, 0....
Dynamical Isometry and a Mean Field Theory of RNNs: Gating Enables Signal Propagation in Recurrent Neural Networks
https://proceedings.mlr.press/v80/chen18i.html
[ "Minmin Chen", "Jeffrey Pennington", "Samuel Schoenholz" ]
null
null
Recurrent neural networks have gained widespread use in modeling sequence data across various domains. While many successful recurrent architectures employ a notion of gating, the exact mechanism that enables such remarkable performance is not well understood. We develop a theory for signal propagation in recurrent net...
[]
null
90
1806.05394
title_snapshot
[ -0.041674524545669556, -0.018001984804868698, 0.007444897200912237, 0.028399184346199036, 0.035477008670568466, 0.034857627004384995, 0.05232622101902962, 0.021656828001141548, -0.06293340772390366, -0.030655456706881523, 0.031184379011392593, 0.003684569150209427, -0.05630119517445564, 0....
Learning to Explain: An Information-Theoretic Perspective on Model Interpretation
https://proceedings.mlr.press/v80/chen18j.html
[ "Jianbo Chen", "Le Song", "Martin Wainwright", "Michael Jordan" ]
null
null
We introduce instancewise feature selection as a methodology for model interpretation. Our method is based on learning a function to extract a subset of features that are most informative for each given example. This feature selector is trained to maximize the mutual information between selected features and the respon...
[]
null
91
1802.07814
title_snapshot
[ -0.03287643939256668, 0.011038878001272678, -0.030538048595190048, 0.05874441936612129, 0.02813579887151718, 0.053217265754938126, 0.015563074499368668, -0.03524983301758766, -0.02021697908639908, -0.011701513081789017, -0.027847839519381523, 0.0574922114610672, -0.07805388420820236, 0.005...
Variational Inference and Model Selection with Generalized Evidence Bounds
https://proceedings.mlr.press/v80/chen18k.html
[ "Liqun Chen", "Chenyang Tao", "Ruiyi Zhang", "Ricardo Henao", "Lawrence Carin Duke" ]
null
null
Recent advances on the scalability and flexibility of variational inference have made it successful at unravelling hidden patterns in complex data. In this work we propose a new variational bound formulation, yielding an estimator that extends beyond the conventional variational bound. It naturally subsumes the importa...
[]
null
92
null
null
[ -0.04253527149558067, 0.012561548501253128, -0.0006884405156597495, 0.03395789861679077, 0.04816785827279091, 0.03292826563119888, 0.03594585135579109, -0.012291104532778263, -0.03811344504356384, -0.041999079287052155, -0.006682717241346836, 0.03337977081537247, -0.08097289502620697, 0.00...
DRACO: Byzantine-resilient Distributed Training via Redundant Gradients
https://proceedings.mlr.press/v80/chen18l.html
[ "Lingjiao Chen", "Hongyi Wang", "Zachary Charles", "Dimitris Papailiopoulos" ]
null
null
Distributed model training is vulnerable to byzantine system failures and adversarial compute nodes, i.e., nodes that use malicious updates to corrupt the global model stored at a parameter server (PS). To guarantee some form of robustness, recent work suggests using variants of the geometric median as an aggregation r...
[]
null
93
1803.09877
title_snapshot
[ -0.004098176956176758, -0.043995294719934464, -0.045020200312137604, 0.09450160712003708, 0.025836586952209473, 0.04434334859251976, 0.03646521270275116, -0.014127802103757858, -0.005956381559371948, -0.05595496669411659, 0.005583807826042175, 0.00048235885333269835, -0.0729953944683075, 0...
SADAGRAD: Strongly Adaptive Stochastic Gradient Methods
https://proceedings.mlr.press/v80/chen18m.html
[ "Zaiyi Chen", "Yi Xu", "Enhong Chen", "Tianbao Yang" ]
null
null
Although the convergence rates of existing variants of ADAGRAD have a better dependence on the number of iterations under the strong convexity condition, their iteration complexities have a explicitly linear dependence on the dimensionality of the problem. To alleviate this bad dependence, we propose a simple yet novel...
[]
null
94
null
null
[ -0.037527311593294144, -0.02056058496236801, 0.008266700431704521, 0.03598932921886444, 0.030309857800602913, 0.05866575241088867, 0.03198548033833504, -0.002487985650077462, -0.022724226117134094, -0.041919391602277756, -0.01935374177992344, -0.008294282481074333, -0.05742594227194786, -0...
Covariate Adjusted Precision Matrix Estimation via Nonconvex Optimization
https://proceedings.mlr.press/v80/chen18n.html
[ "Jinghui Chen", "Pan Xu", "Lingxiao Wang", "Jian Ma", "Quanquan Gu" ]
null
null
We propose a nonconvex estimator for the covariate adjusted precision matrix estimation problem in the high dimensional regime, under sparsity constraints. To solve this estimator, we propose an alternating gradient descent algorithm with hard thresholding. Compared with existing methods along this line of research, wh...
[]
null
95
null
null
[ -0.026133501902222633, -0.011183597147464752, -0.0151857053861022, -0.00041742980829440057, 0.023059310391545296, 0.05028480663895607, 0.045137859880924225, -0.0074387285858392715, -0.039812520146369934, -0.02547813020646572, -0.008824639022350311, 0.0032299067825078964, -0.06399616599082947...
End-to-End Learning for the Deep Multivariate Probit Model
https://proceedings.mlr.press/v80/chen18o.html
[ "Di Chen", "Yexiang Xue", "Carla Gomes" ]
null
null
The multivariate probit model (MVP) is a popular classic model for studying binary responses of multiple entities. Nevertheless, the computational challenge of learning the MVP model, given that its likelihood involves integrating over a multidimensional constrained space of latent variables, significantly limits its a...
[]
null
96
1803.08591
title_snapshot
[ -0.019264720380306244, -0.042744845151901245, -0.0010272335493937135, 0.04258427023887634, 0.024136677384376526, 0.03164594992995262, 0.026798468083143234, -0.002545527881011367, -0.022957313805818558, -0.034796468913555145, 0.00008303491631522775, -0.00023614698147866875, -0.064224019646644...
Stochastic Training of Graph Convolutional Networks with Variance Reduction
https://proceedings.mlr.press/v80/chen18p.html
[ "Jianfei Chen", "Jun Zhu", "Le Song" ]
null
null
Graph convolutional networks (GCNs) are powerful deep neural networks for graph-structured data. However, GCN computes the representation of a node recursively from its neighbors, making the receptive field size grow exponentially with the number of layers. Previous attempts on reducing the receptive field size by subs...
[]
null
97
1710.10568
title_snapshot
[ -0.0041466462425887585, -0.03268871083855629, 0.014108406379818916, 0.04457511007785797, 0.027471693232655525, 0.031356364488601685, 0.0423765629529953, 0.038596563041210175, -0.014474861323833466, -0.05439577251672745, 0.013518567197024822, -0.028248202055692673, -0.07683083415031433, 0.0...
Extreme Learning to Rank via Low Rank Assumption
https://proceedings.mlr.press/v80/cheng18a.html
[ "Minhao Cheng", "Ian Davidson", "Cho-Jui Hsieh" ]
null
null
We consider the setting where we wish to perform ranking for hundreds of thousands of users which is common in recommender systems and web search ranking. Learning a single ranking function is unlikely to capture the variability across all users while learning a ranking function for each person is time-consuming and re...
[]
null
98
null
null
[ -0.02443986013531685, -0.04088618606328964, 0.03190746158361435, 0.03643010929226875, 0.030792275443673134, 0.012900067493319511, 0.02703217603266239, -0.025465765967965126, -0.00885769072920084, -0.01266972254961729, -0.01309389527887106, 0.016774915158748627, -0.05479685217142105, 0.0083...
Learning a Mixture of Two Multinomial Logits
https://proceedings.mlr.press/v80/chierichetti18a.html
[ "Flavio Chierichetti", "Ravi Kumar", "Andrew Tomkins" ]
null
null
The classical Multinomial Logit (MNL) is a behavioral model for user choice. In this model, a user is offered a slate of choices (a subset of a finite universe of $n$ items), and selects exactly one item from the slate, each with probability proportional to its (positive) weight. Given a set of observed slates and choi...
[]
null
99
null
null
[ -0.029910147190093994, -0.028513103723526, -0.01845056563615799, 0.0359257236123085, 0.03326145559549332, 0.018217680975794792, 0.004307679831981659, 0.017352504655718803, -0.04257752001285553, -0.028816012665629387, -0.027338329702615738, 0.008796567097306252, -0.06970342248678207, -0.033...
Structured Evolution with Compact Architectures for Scalable Policy Optimization
https://proceedings.mlr.press/v80/choromanski18a.html
[ "Krzysztof Choromanski", "Mark Rowland", "Vikas Sindhwani", "Richard Turner", "Adrian Weller" ]
null
null
We present a new method of blackbox optimization via gradient approximation with the use of structured random orthogonal matrices, providing more accurate estimators than baselines and with provable theoretical guarantees. We show that this algorithm can be successfully applied to learn better quality compact policies ...
[]
null
100
1804.02395
title_snapshot
[ -0.014689408242702484, -0.04816630855202675, -0.009002219885587692, 0.04881306737661362, 0.043307799845933914, 0.07177116721868515, 0.005093025509268045, -0.020857486873865128, -0.013130788691341877, -0.03346388041973114, 0.012824217788875103, -0.002358238445594907, -0.092483751475811, -0....