title
stringlengths
12
143
paper_url
stringlengths
44
58
authors
listlengths
1
13
type
stringclasses
0 values
primary_area
stringclasses
0 values
abstract
large_stringlengths
288
4.43k
keywords
listlengths
0
0
TL;DR
large_stringclasses
0 values
submission_number
int64
1
773
arxiv_id
stringlengths
10
10
arxiv_id_source
stringclasses
2 values
embedding
listlengths
768
768
AReS and MaRS Adversarial and MMD-Minimizing Regression for SDEs
https://proceedings.mlr.press/v97/abbati19a.html
[ "Gabriele Abbati", "Philippe Wenk", "Michael A. Osborne", "Andreas Krause", "Bernhard Schölkopf", "Stefan Bauer" ]
null
null
Stochastic differential equations are an important modeling class in many disciplines. Consequently, there exist many methods relying on various discretization and numerical integration schemes. In this paper, we propose a novel, probabilistic model for estimating the drift and diffusion given noisy observations of the...
[]
null
1
1902.08480
title_snapshot
[ -0.01475281361490488, 0.02767331153154373, -0.010267019271850586, 0.05629995837807655, 0.05450527369976044, 0.04201443865895271, 0.03370289131999016, 0.006314247380942106, -0.026731790974736214, -0.06461016833782196, 0.00764450803399086, -0.007380622439086437, -0.04433220252394676, -0.0096...
Dynamic Weights in Multi-Objective Deep Reinforcement Learning
https://proceedings.mlr.press/v97/abels19a.html
[ "Axel Abels", "Diederik Roijers", "Tom Lenaerts", "Ann Nowé", "Denis Steckelmacher" ]
null
null
Many real-world decision problems are characterized by multiple conflicting objectives which must be balanced based on their relative importance. In the dynamic weights setting the relative importance changes over time and specialized algorithms that deal with such change, such as a tabular Reinforcement Learning (RL) ...
[]
null
2
1809.07803
title_snapshot
[ -0.06123773753643036, -0.013469628989696503, -0.006091605871915817, 0.034519731998443604, 0.028802571818232536, 0.031595628708601, -0.006737292744219303, 0.007433074526488781, -0.05597781017422676, -0.05052630230784416, -0.026796218007802963, 0.018374165520071983, -0.07239469140768051, -0....
MixHop: Higher-Order Graph Convolutional Architectures via Sparsified Neighborhood Mixing
https://proceedings.mlr.press/v97/abu-el-haija19a.html
[ "Sami Abu-El-Haija", "Bryan Perozzi", "Amol Kapoor", "Nazanin Alipourfard", "Kristina Lerman", "Hrayr Harutyunyan", "Greg Ver Steeg", "Aram Galstyan" ]
null
null
Existing popular methods for semi-supervised learning with Graph Neural Networks (such as the Graph Convolutional Network) provably cannot learn a general class of neighborhood mixing relationships. To address this weakness, we propose a new model, MixHop, that can learn these relationships, including difference operat...
[]
null
3
1905.00067
title_snapshot
[ 0.04124806076288223, -0.04907359555363655, 0.0012405433226376772, 0.05263511463999748, 0.018444178625941277, 0.03627140447497368, 0.019593019038438797, 0.004926078487187624, -0.0038921176455914974, -0.06383872777223587, 0.0034328403417021036, -0.012852258048951626, -0.07533838599920273, 0....
Communication-Constrained Inference and the Role of Shared Randomness
https://proceedings.mlr.press/v97/acharya19a.html
[ "Jayadev Acharya", "Clement Canonne", "Himanshu Tyagi" ]
null
null
A central server needs to perform statistical inference based on samples that are distributed over multiple users who can each send a message of limited length to the center. We study problems of distribution learning and identity testing in this distributed inference setting and examine the role of shared randomness a...
[]
null
4
1905.08302
title_judge
[ 0.0036378460936248302, -0.0038676925469189882, -0.03272268548607826, 0.0800064280629158, 0.03515314310789108, 0.008345500566065311, 0.03516628220677376, 0.007164329290390015, -0.0062719425186514854, -0.04561731591820717, 0.025469446554780006, -0.023213587701320648, -0.06913220882415771, 0....
Distributed Learning with Sublinear Communication
https://proceedings.mlr.press/v97/acharya19b.html
[ "Jayadev Acharya", "Chris De Sa", "Dylan Foster", "Karthik Sridharan" ]
null
null
In distributed statistical learning, $N$ samples are split across $m$ machines and a learner wishes to use minimal communication to learn as well as if the examples were on a single machine. This model has received substantial interest in machine learning due to its scalability and potential for parallel speedup. Howev...
[]
null
5
1902.11259
title_snapshot
[ -0.021587898954749107, -0.03199276328086853, -0.0012688988354057074, 0.012993289157748222, 0.0276494100689888, 0.020025186240673065, 0.0534999780356884, -0.014428995549678802, -0.004912131931632757, -0.05558633431792259, 0.009769429452717304, -0.0189049169421196, -0.07537578791379929, 0.02...
Communication Complexity in Locally Private Distribution Estimation and Heavy Hitters
https://proceedings.mlr.press/v97/acharya19c.html
[ "Jayadev Acharya", "Ziteng Sun" ]
null
null
We consider the problems of distribution estimation, and heavy hitter (frequency) estimation under privacy, and communication constraints. While the constraints have been studied separately, optimal schemes for one are sub-optimal for the other. We propose a sample-optimal $\eps$-locally differentially private (LDP) sc...
[]
null
6
1905.11888
title_snapshot
[ -0.016130002215504646, 0.008072704076766968, 0.0022999283391982317, 0.04908524081110954, 0.0366847924888134, 0.028917308896780014, 0.021175777539610863, -0.02164597064256668, -0.010404601693153381, -0.043140798807144165, 0.032812103629112244, -0.011841246858239174, -0.054935261607170105, -...
Learning Models from Data with Measurement Error: Tackling Underreporting
https://proceedings.mlr.press/v97/adams19a.html
[ "Roy Adams", "Yuelong Ji", "Xiaobin Wang", "Suchi Saria" ]
null
null
Measurement error in observational datasets can lead to systematic bias in inferences based on these datasets. As studies based on observational data are increasingly used to inform decisions with real-world impact, it is critical that we develop a robust set of techniques for analyzing and adjusting for these biases. ...
[]
null
7
1901.09060
title_snapshot
[ 0.0037026198115199804, 0.01595727913081646, -0.04886854439973831, 0.01095893606543541, 0.051226649433374405, 0.03163708373904228, 0.06759265065193176, 0.004640278872102499, -0.01882386952638626, -0.03743262588977814, 0.016706207767128944, -0.009080409072339535, -0.06695521622896194, -0.020...
TibGM: A Transferable and Information-Based Graphical Model Approach for Reinforcement Learning
https://proceedings.mlr.press/v97/adel19a.html
[ "Tameem Adel", "Adrian Weller" ]
null
null
One of the challenges to reinforcement learning (RL) is scalable transferability among complex tasks. Incorporating a graphical model (GM), along with the rich family of related methods, as a basis for RL frameworks provides potential to address issues such as transferability, generalisation and exploration. Here we pr...
[]
null
8
null
null
[ -0.03973774239420891, -0.028840720653533936, -0.0003902135940734297, 0.010376001708209515, 0.029929162934422493, -0.004094300325959921, 0.017187807708978653, 0.003675706684589386, -0.017560606822371483, -0.03998705744743347, -0.008465657941997051, 0.05053151771426201, -0.08467479050159454, ...
PAC Learnability of Node Functions in Networked Dynamical Systems
https://proceedings.mlr.press/v97/adiga19a.html
[ "Abhijin Adiga", "Chris J Kuhlman", "Madhav Marathe", "S Ravi", "Anil Vullikanti" ]
null
null
We consider the PAC learnability of the local functions at the vertices of a discrete networked dynamical system, assuming that the underlying network is known. Our focus is on the learnability of threshold functions. We show that several variants of threshold functions are PAC learnable and provide tight bounds on the...
[]
null
9
null
null
[ -0.03356032818555832, -0.03384532406926155, -0.015024405904114246, 0.06329545378684998, 0.03417431190609932, 0.028963129967451096, 0.016521872952580452, -0.007647205609828234, -0.01691633090376854, -0.02353443205356598, 0.03618233650922775, 0.0066099269315600395, -0.06770040839910507, 0.00...
Static Automatic Batching In TensorFlow
https://proceedings.mlr.press/v97/agarwal19a.html
[ "Ashish Agarwal" ]
null
null
Dynamic neural networks are becoming increasingly common, and yet it is hard to implement them efficiently. On-the-fly operation batching for such models is sub-optimal and suffers from run time overheads, while writing manually batched versions can be hard and error-prone. To address this we extend TensorFlow with pfo...
[]
null
10
null
null
[ -0.027163801714777946, -0.04578056558966637, -0.013532647863030434, 0.0036484613083302975, 0.017245884984731674, 0.05002356320619583, 0.02000141330063343, 0.02031344547867775, -0.04729184880852699, -0.04810208082199097, -0.009427351877093315, -0.027539731934666634, -0.08110728859901428, 0....
Efficient Full-Matrix Adaptive Regularization
https://proceedings.mlr.press/v97/agarwal19b.html
[ "Naman Agarwal", "Brian Bullins", "Xinyi Chen", "Elad Hazan", "Karan Singh", "Cyril Zhang", "Yi Zhang" ]
null
null
Adaptive regularization methods pre-multiply a descent direction by a preconditioning matrix. Due to the large number of parameters of machine learning problems, full-matrix preconditioning methods are prohibitively expensive. We show how to modify full-matrix adaptive regularization in order to make it practical and e...
[]
null
11
1806.02958
title_snapshot
[ -0.0425356961786747, -0.04077911376953125, 0.04025338590145111, 0.03193371370434761, 0.03252701088786125, 0.03538951650261879, 0.02952929586172104, -0.001909688115119934, -0.04351881891489029, -0.05929252877831459, -0.020151372998952866, 0.010660850442945957, -0.04253213480114937, -0.01625...
Online Control with Adversarial Disturbances
https://proceedings.mlr.press/v97/agarwal19c.html
[ "Naman Agarwal", "Brian Bullins", "Elad Hazan", "Sham Kakade", "Karan Singh" ]
null
null
We study the control of linear dynamical systems with adversarial disturbances, as opposed to statistical noise. We present an efficient algorithm that achieves nearly-tight regret bounds in this setting. Our result generalizes upon previous work in two main aspects: the algorithm can accommodate adversarial noise in t...
[]
null
12
1902.08721
title_snapshot
[ -0.059290993958711624, -0.018718378618359566, -0.02134128101170063, 0.040829699486494064, 0.027293428778648376, 0.018152903765439987, 0.022835932672023773, 0.006047005765140057, -0.03302175924181938, -0.040272582322359085, -0.03473946824669838, -0.006779992952942848, -0.07691144198179245, ...
Fair Regression: Quantitative Definitions and Reduction-Based Algorithms
https://proceedings.mlr.press/v97/agarwal19d.html
[ "Alekh Agarwal", "Miroslav Dudik", "Zhiwei Steven Wu" ]
null
null
In this paper, we study the prediction of a real-valued target, such as a risk score or recidivism rate, while guaranteeing a quantitative notion of fairness with respect to a protected attribute such as gender or race. We call this class of problems fair regression. We propose general schemes for fair regression under...
[]
null
13
1905.12843
title_snapshot
[ -0.03760555386543274, -0.017087290063500404, -0.02962334081530571, 0.017457446083426476, 0.03710746765136719, 0.045691464096307755, 0.03541741147637367, -0.03388527035713196, -0.053169120103120804, -0.016321981325745583, -0.007842381484806538, 0.024585802108049393, -0.08208519220352173, -0...
Learning to Generalize from Sparse and Underspecified Rewards
https://proceedings.mlr.press/v97/agarwal19e.html
[ "Rishabh Agarwal", "Chen Liang", "Dale Schuurmans", "Mohammad Norouzi" ]
null
null
We consider the problem of learning from sparse and underspecified rewards, where an agent receives a complex input, such as a natural language instruction, and needs to generate a complex response, such as an action sequence, while only receiving binary success-failure feedback. Such success-failure rewards are often ...
[]
null
14
1902.07198
title_snapshot
[ -0.010337923653423786, -0.03554059565067291, 0.019207540899515152, 0.04866225644946098, 0.0342785082757473, 0.002307645743712783, 0.040004435926675797, 0.00041403519571758807, -0.04671498015522957, -0.012819267809391022, -0.023632556200027466, 0.04299165681004524, -0.058395400643348694, -0...
The Kernel Interaction Trick: Fast Bayesian Discovery of Pairwise Interactions in High Dimensions
https://proceedings.mlr.press/v97/agrawal19a.html
[ "Raj Agrawal", "Brian Trippe", "Jonathan Huggins", "Tamara Broderick" ]
null
null
Discovering interaction effects on a response of interest is a fundamental problem faced in biology, medicine, economics, and many other scientific disciplines. In theory, Bayesian methods for discovering pairwise interactions enjoy many benefits such as coherent uncertainty quantification, the ability to incorporate b...
[]
null
15
1905.06501
title_snapshot
[ -0.022739779204130173, -0.0017288693925365806, 0.011078648269176483, 0.032940663397312164, 0.029488494619727135, 0.00620168587192893, 0.03567599877715111, -0.035668645054101944, 0.009851396083831787, -0.046482622623443604, 0.005579364486038685, 0.036755260080099106, -0.07500413805246353, -...
Understanding the Impact of Entropy on Policy Optimization
https://proceedings.mlr.press/v97/ahmed19a.html
[ "Zafarali Ahmed", "Nicolas Le Roux", "Mohammad Norouzi", "Dale Schuurmans" ]
null
null
Entropy regularization is commonly used to improve policy optimization in reinforcement learning. It is believed to help with exploration by encouraging the selection of more stochastic policies. In this work, we analyze this claim using new visualizations of the optimization landscape based on randomly perturbing the ...
[]
null
16
1811.11214
title_snapshot
[ -0.02288273721933365, -0.04116581752896309, -0.0013306362088769674, 0.045741431415081024, 0.045938681811094284, 0.032666902989149094, 0.01655091531574726, -0.02985502779483795, -0.04442068934440613, -0.04763878509402275, -0.017712241038680077, 0.013647621497511864, -0.04940979182720184, -0...
Fairwashing: the risk of rationalization
https://proceedings.mlr.press/v97/aivodji19a.html
[ "Ulrich Aivodji", "Hiromi Arai", "Olivier Fortineau", "Sébastien Gambs", "Satoshi Hara", "Alain Tapp" ]
null
null
Black-box explanation is the problem of explaining how a machine learning model – whose internal logic is hidden to the auditor and generally complex – produces its outcomes. Current approaches for solving this problem include model explanation, outcome explanation as well as model inspection. While these techniques ca...
[]
null
17
1901.09749
title_snapshot
[ -0.029597414657473564, -0.02423294447362423, -0.062270455062389374, 0.035930946469306946, 0.020250918343663216, 0.018328329548239708, 0.0035645479802042246, -0.001476317411288619, -0.03699144721031189, -0.019089382141828537, -0.010720538906753063, 0.030078595504164696, -0.07437784969806671, ...
Adaptive Stochastic Natural Gradient Method for One-Shot Neural Architecture Search
https://proceedings.mlr.press/v97/akimoto19a.html
[ "Youhei Akimoto", "Shinichi Shirakawa", "Nozomu Yoshinari", "Kento Uchida", "Shota Saito", "Kouhei Nishida" ]
null
null
High sensitivity of neural architecture search (NAS) methods against their input such as step-size (i.e., learning rate) and search space prevents practitioners from applying them out-of-the-box to their own problems, albeit its purpose is to automate a part of tuning process. Aiming at a fast, robust, and widely-appli...
[]
null
18
1905.08537
title_snapshot
[ -0.020703408867120743, -0.015107518993318081, -0.008613206446170807, 0.02158941887319088, 0.03743726760149002, 0.07676473259925842, 0.045730654150247574, 0.00038399043842218816, -0.02918718196451664, -0.05741269886493683, -0.0070185880176723, -0.027796870097517967, -0.03380810469388962, -0...
Projections for Approximate Policy Iteration Algorithms
https://proceedings.mlr.press/v97/akrour19a.html
[ "Riad Akrour", "Joni Pajarinen", "Jan Peters", "Gerhard Neumann" ]
null
null
Approximate policy iteration is a class of reinforcement learning (RL) algorithms where the policy is encoded using a function approximator and which has been especially prominent in RL with continuous action spaces. In this class of RL algorithms, ensuring increase of the policy return during policy update often requi...
[]
null
19
null
null
[ -0.04400923103094101, -0.04621373862028122, -0.006419236771762371, 0.04314235597848892, 0.04382247105240822, 0.02520717680454254, 0.00391204422339797, -0.03246338665485382, -0.03686324506998062, -0.006859214045107365, -0.029338739812374115, 0.011057017371058464, -0.07310944050550461, -0.01...
Validating Causal Inference Models via Influence Functions
https://proceedings.mlr.press/v97/alaa19a.html
[ "Ahmed Alaa", "Mihaela Van Der Schaar" ]
null
null
The problem of estimating causal effects of treatments from observational data falls beyond the realm of supervised learning {—} because counterfactual data is inaccessible, we can never observe the true causal effects. In the absence of "supervision", how can we evaluate the performance of causal inference methods? In...
[]
null
20
null
null
[ -0.0007720402209088206, -0.018563684076070786, -0.020588146522641182, 0.019881580024957657, 0.038256287574768066, 0.027090592309832573, 0.03450150415301323, -0.0056958734057843685, -0.004537667613476515, -0.02606978267431259, 0.005440508481115103, 0.06701138615608215, -0.05494000390172005, ...
Multi-objective training of Generative Adversarial Networks with multiple discriminators
https://proceedings.mlr.press/v97/albuquerque19a.html
[ "Isabela Albuquerque", "Joao Monteiro", "Thang Doan", "Breandan Considine", "Tiago Falk", "Ioannis Mitliagkas" ]
null
null
Recent literature has demonstrated promising results for training Generative Adversarial Networks by employing a set of discriminators, in contrast to the traditional game involving one generator against a single adversary. Such methods perform single-objective optimization on some simple consolidation of the losses, e...
[]
null
21
1901.08680
title_snapshot
[ -0.02659333497285843, -0.032481662929058075, 0.01795300655066967, 0.038121648132801056, -0.008874399587512016, 0.035303473472595215, 0.011153753846883774, -0.0008420497179031372, -0.022983504459261894, -0.054805222898721695, -0.016983933746814728, 0.02095523662865162, -0.06890363246202469, ...
Graph Element Networks: adaptive, structured computation and memory
https://proceedings.mlr.press/v97/alet19a.html
[ "Ferran Alet", "Adarsh Keshav Jeewajee", "Maria Bauza Villalonga", "Alberto Rodriguez", "Tomas Lozano-Perez", "Leslie Kaelbling" ]
null
null
We explore the use of graph neural networks (GNNs) to model spatial processes in which there is no a priori graphical structure. Similar to finite element analysis, we assign nodes of a GNN to spatial locations and use a computational process defined on the graph to model the relationship between an initial function de...
[]
null
22
1904.09019
title_snapshot
[ -0.016542546451091766, 0.02164691686630249, 0.001327596022747457, 0.04057066887617111, 0.035557836294174194, 0.04027222841978073, 0.0022367953788489103, 0.030397487804293633, -0.03727591410279274, -0.05947359278798103, 0.0018884413875639439, -0.03716161474585533, -0.059930916875600815, 0.0...
Analogies Explained: Towards Understanding Word Embeddings
https://proceedings.mlr.press/v97/allen19a.html
[ "Carl Allen", "Timothy Hospedales" ]
null
null
Word embeddings generated by neural network methods such as word2vec (W2V) are well known to exhibit seemingly linear behaviour, e.g. the embeddings of analogy “woman is to queen as man is to king” approximately describe a parallelogram. This property is particularly intriguing since the embeddings are not trained to a...
[]
null
23
1901.09813
title_snapshot
[ -0.00987306796014309, -0.011488127522170544, 0.0005684912321157753, 0.0331781730055809, 0.028152991086244583, 0.06270790100097656, 0.030779067426919937, -0.007321174722164869, -0.0014028623700141907, -0.01855260506272316, 0.0011407192796468735, -0.000033027492463588715, -0.05135674029588699,...
Infinite Mixture Prototypes for Few-shot Learning
https://proceedings.mlr.press/v97/allen19b.html
[ "Kelsey Allen", "Evan Shelhamer", "Hanul Shin", "Joshua Tenenbaum" ]
null
null
We propose infinite mixture prototypes to adaptively represent both simple and complex data distributions for few-shot learning. Infinite mixture prototypes combine deep representation learning with Bayesian nonparametrics, representing each class by a set of clusters, unlike existing prototypical methods that represen...
[]
null
24
1902.04552
title_snapshot
[ -0.005830499343574047, -0.03140658512711525, -0.02726820670068264, 0.0490240715444088, 0.03606041148304939, 0.04214606434106827, 0.024453986436128616, -0.008028600364923477, -0.0222258772701025, -0.0254032202064991, -0.030528971925377846, 0.00718291662633419, -0.049989305436611176, 0.03352...
A Convergence Theory for Deep Learning via Over-Parameterization
https://proceedings.mlr.press/v97/allen-zhu19a.html
[ "Zeyuan Allen-Zhu", "Yuanzhi Li", "Zhao Song" ]
null
null
Deep neural networks (DNNs) have demonstrated dominating performance in many fields; since AlexNet, networks used in practice are going wider and deeper. On the theoretical side, a long line of works have been focusing on why we can train neural networks when there is only one hidden layer. The theory of multi-layer ne...
[]
null
25
1811.03962
title_snapshot
[ -0.033369407057762146, -0.019056767225265503, -0.00206159520894289, 0.04771957919001579, 0.0331660695374012, 0.043386060744524, 0.01042701955884695, 0.01657194457948208, -0.03135935589671135, -0.04764215275645256, -0.0020170279312878847, -0.036673497408628464, -0.054563023149967194, 0.0042...
Asynchronous Batch Bayesian Optimisation with Improved Local Penalisation
https://proceedings.mlr.press/v97/alvi19a.html
[ "Ahsan Alvi", "Binxin Ru", "Jan-Peter Calliess", "Stephen Roberts", "Michael A. Osborne" ]
null
null
Batch Bayesian optimisation (BO) has been successfully applied to hyperparameter tuning using parallel computing, but it is wasteful of resources: workers that complete jobs ahead of others are left idle. We address this problem by developing an approach, Penalising Locally for Asynchronous Bayesian Optimisation on K W...
[]
null
26
1901.10452
title_snapshot
[ -0.015761438757181168, -0.016545286402106285, -0.003979539033025503, -0.00342151103541255, 0.011339429765939713, 0.05580112338066101, 0.04780634492635727, -0.0026426564436405897, -0.022918174043297768, -0.022926008328795433, -0.004437388852238655, -0.01471610739827156, -0.057618819177150726,...
Bounding User Contributions: A Bias-Variance Trade-off in Differential Privacy
https://proceedings.mlr.press/v97/amin19a.html
[ "Kareem Amin", "Alex Kulesza", "Andres Munoz", "Sergei Vassilvtiskii" ]
null
null
Differentially private learning algorithms protect individual participants in the training dataset by guaranteeing that their presence does not significantly change the resulting model. In order to make this promise, such algorithms need to know the maximum contribution that can be made by a single user: the more data ...
[]
null
27
null
null
[ 0.0037752517964690924, -0.009676460176706314, -0.012873259373009205, 0.0619286373257637, 0.04935234785079956, 0.005086956545710564, 0.04885224997997284, -0.035238541662693024, -0.027776718139648438, -0.020548515021800995, 0.0021485830657184124, -0.017222439870238304, -0.06943067908287048, ...
Explaining Deep Neural Networks with a Polynomial Time Algorithm for Shapley Value Approximation
https://proceedings.mlr.press/v97/ancona19a.html
[ "Marco Ancona", "Cengiz Oztireli", "Markus Gross" ]
null
null
The problem of explaining the behavior of deep neural networks has recently gained a lot of attention. While several attribution methods have been proposed, most come without strong theoretical foundations, which raises questions about their reliability. On the other hand, the literature on cooperative game theory sugg...
[]
null
28
1903.10992
title_judge
[ -0.03908690810203552, -0.027491915971040726, -0.018021270632743835, 0.05609423294663429, 0.025210879743099213, 0.021734118461608887, -0.005921620409935713, 0.0009492371464148164, 0.005239806603640318, -0.04164913669228554, -0.011265575885772705, 0.01631290838122368, -0.06669192016124725, 0...
Scaling Up Ordinal Embedding: A Landmark Approach
https://proceedings.mlr.press/v97/anderton19a.html
[ "Jesse Anderton", "Javed Aslam" ]
null
null
Ordinal Embedding is the problem of placing n objects into R^d to satisfy constraints like "object a is closer to b than to c." It can accommodate data that embeddings from features or distances cannot, but is a more difficult problem. We propose a novel landmark-based method as a partial solution. At small to medium s...
[]
null
29
null
null
[ -0.015047959983348846, -0.020065154880285263, 0.03964301198720932, 0.01119279209524393, 0.02765261009335518, 0.037527523934841156, 0.022238347679376602, -0.01061997376382351, -0.023959366604685783, -0.03986331447958946, -0.019447730854153633, 0.002251907717436552, -0.07437661290168762, -0....
Sorting Out Lipschitz Function Approximation
https://proceedings.mlr.press/v97/anil19a.html
[ "Cem Anil", "James Lucas", "Roger Grosse" ]
null
null
Training neural networks under a strict Lipschitz constraint is useful for provable adversarial robustness, generalization bounds, interpretable gradients, and Wasserstein distance estimation. By the composition property of Lipschitz functions, it suffices to ensure that each individual affine transformation or nonline...
[]
null
30
1811.05381
title_snapshot
[ -0.06185542419552803, -0.007945425808429718, 0.005749076139181852, 0.038683027029037476, 0.008609121665358543, 0.029329214245080948, 0.02905832603573799, -0.007587534841150045, -0.024648016318678856, -0.031181320548057556, -0.011790946125984192, -0.0006135163130238652, -0.06065629795193672, ...
Sparse Multi-Channel Variational Autoencoder for the Joint Analysis of Heterogeneous Data
https://proceedings.mlr.press/v97/antelmi19a.html
[ "Luigi Antelmi", "Nicholas Ayache", "Philippe Robert", "Marco Lorenzi" ]
null
null
Interpretable modeling of heterogeneous data channels is essential in medical applications, for example when jointly analyzing clinical scores and medical images. Variational Autoencoders (VAE) are powerful generative models that learn representations of complex data. The flexibility of VAE may come at the expense of l...
[]
null
31
null
null
[ 0.015115177258849144, 0.007061624433845282, -0.019275708124041557, 0.020346730947494507, 0.03930108621716499, 0.05608876049518585, 0.05085361748933792, -0.005518944468349218, -0.027796991169452667, -0.055170804262161255, -0.0038109607994556427, 0.0025623179972171783, -0.05756214261054993, ...
Unsupervised Label Noise Modeling and Loss Correction
https://proceedings.mlr.press/v97/arazo19a.html
[ "Eric Arazo", "Diego Ortego", "Paul Albert", "Noel O’Connor", "Kevin Mcguinness" ]
null
null
Despite being robust to small amounts of label noise, convolutional neural networks trained with stochastic gradient methods have been shown to easily fit random labels. When there are a mixture of correct and mislabelled targets, networks tend to fit the former before the latter. This suggests using a suitable two-com...
[]
null
32
1904.11238
title_snapshot
[ 0.04420575127005577, -0.0194106362760067, -0.0405062772333622, 0.04379773512482643, 0.02147842012345791, 0.035503171384334564, 0.009301247075200081, 0.01978721283376217, -0.016588184982538223, -0.05583423748612404, -0.007545450236648321, 0.015206938609480858, -0.06483621895313263, -0.00854...
Fine-Grained Analysis of Optimization and Generalization for Overparameterized Two-Layer Neural Networks
https://proceedings.mlr.press/v97/arora19a.html
[ "Sanjeev Arora", "Simon Du", "Wei Hu", "Zhiyuan Li", "Ruosong Wang" ]
null
null
Recent works have cast some light on the mystery of why deep nets fit any data and generalize despite being very overparametrized. This paper analyzes training and generalization for a simple 2-layer ReLU net with random initialization, and provides the following improvements over recent works: (i) Using a tighter char...
[]
null
33
1901.08584
title_snapshot
[ -0.027994800359010696, -0.031511325389146805, 0.009563660249114037, 0.04895868897438049, 0.04588457942008972, 0.03363827243447304, 0.022636696696281433, 0.0024052592925727367, -0.02018953301012516, -0.04034588113427162, 0.005155140534043312, -0.014245038852095604, -0.05237139016389847, -0....
Distributed Weighted Matching via Randomized Composable Coresets
https://proceedings.mlr.press/v97/assadi19a.html
[ "Sepehr Assadi", "Mohammadhossein Bateni", "Vahab Mirrokni" ]
null
null
Maximum weight matching is one of the most fundamental combinatorial optimization problems with a wide range of applications in data mining and bioinformatics. Developing distributed weighted matching algorithms has been challenging due to the sequential nature of efficient algorithms for this problem. In this paper, w...
[]
null
34
1906.01993
title_snapshot
[ -0.01930774562060833, -0.047239404171705246, -0.014610439538955688, 0.050514183938503265, 0.048677872866392136, 0.05801847204566002, 0.008184782229363918, 0.016747020184993744, -0.007916850037872791, -0.07516557723283768, 0.012897332198917866, -0.05894263833761215, -0.0854407250881195, -0....
Stochastic Gradient Push for Distributed Deep Learning
https://proceedings.mlr.press/v97/assran19a.html
[ "Mahmoud Assran", "Nicolas Loizou", "Nicolas Ballas", "Mike Rabbat" ]
null
null
Distributed data-parallel algorithms aim to accelerate the training of deep neural networks by parallelizing the computation of large mini-batch gradient updates across multiple nodes. Approaches that synchronize nodes using exact distributed averaging (e.g., via AllReduce) are sensitive to stragglers and communication...
[]
null
35
1811.10792
title_snapshot
[ -0.01196060050278902, -0.0464484803378582, -0.01982865482568741, 0.051851145923137665, 0.020633036270737648, 0.06096063554286957, 0.04237683489918709, 0.009597637690603733, -0.03448545187711716, -0.03779999539256096, 0.012085480615496635, -0.03947386518120766, -0.06883405894041061, -0.0098...
Bayesian Optimization of Composite Functions
https://proceedings.mlr.press/v97/astudillo19a.html
[ "Raul Astudillo", "Peter Frazier" ]
null
null
We consider optimization of composite objective functions, i.e., of the form $f(x)=g(h(x))$, where $h$ is a black-box derivative-free expensive-to-evaluate function with vector-valued outputs, and $g$ is a cheap-to-evaluate real-valued function. While these problems can be solved with standard Bayesian optimization, we...
[]
null
36
1906.01537
title_snapshot
[ -0.01617908664047718, 0.024720318615436554, 0.019122691825032234, 0.05501163750886917, 0.02694694697856903, 0.05633172020316124, 0.01098536979407072, -0.0055684796534478664, -0.0065635680221021175, -0.03736089915037155, -0.012946001254022121, 0.011888420209288597, -0.05482219532132149, -0....
Linear-Complexity Data-Parallel Earth Mover’s Distance Approximations
https://proceedings.mlr.press/v97/atasu19a.html
[ "Kubilay Atasu", "Thomas Mittelholzer" ]
null
null
The Earth Mover’s Distance (EMD) is a state-of-the art metric for comparing discrete probability distributions, but its high distinguishability comes at a high cost in computational complexity. Even though linear-complexity approximation algorithms have been proposed to improve its scalability, these algorithms are eit...
[]
null
37
1812.02091
title_judge
[ -0.01378539577126503, -0.014136810787022114, -0.007533992175012827, 0.03769529610872269, 0.030040252953767776, 0.05845164880156517, 0.013195965439081192, 0.029885675758123398, -0.04108109325170517, -0.02696789614856243, -0.0026349364779889584, -0.019763654097914696, -0.06680940091609955, 0...
Benefits and Pitfalls of the Exponential Mechanism with Applications to Hilbert Spaces and Functional PCA
https://proceedings.mlr.press/v97/awan19a.html
[ "Jordan Awan", "Ana Kenney", "Matthew Reimherr", "Aleksandra Slavković" ]
null
null
The exponential mechanism is a fundamental tool of Differential Privacy (DP) due to its strong privacy guarantees and flexibility. We study its extension to settings with summaries based on infinite dimensional outputs such as with functional data analysis, shape analysis, and nonparametric statistics. We show that the...
[]
null
38
1901.10864
title_snapshot
[ -0.01191595010459423, 0.004251785576343536, 0.005547673441469669, 0.0582570843398571, 0.048536013811826706, 0.023915620520710945, 0.03486467897891998, -0.02982613816857338, -0.0021254788152873516, -0.04740659520030022, 0.0025043741334229708, -0.017622986808419228, -0.0563710480928421, -0.0...
Feature Grouping as a Stochastic Regularizer for High-Dimensional Structured Data
https://proceedings.mlr.press/v97/aydore19a.html
[ "Sergul Aydore", "Bertrand Thirion", "Gael Varoquaux" ]
null
null
In many applications where collecting data is expensive, for example neuroscience or medical imaging, the sample size is typically small compared to the feature dimension. These datasets call for intelligent regularization that exploits known structure, such as correlations between the features arising from the measure...
[]
null
39
1807.11718
title_snapshot
[ -0.010308043099939823, -0.02335439994931221, 0.01888061687350273, 0.03296799212694168, 0.03259573504328728, 0.04502082243561745, 0.026539990678429604, -0.01555748749524355, -0.02729935757815838, -0.04834676906466484, 0.004232704173773527, -0.021255865693092346, -0.057899359613657, 0.016630...
Beyond the Chinese Restaurant and Pitman-Yor processes: Statistical Models with double power-law behavior
https://proceedings.mlr.press/v97/ayed19a.html
[ "Fadhel Ayed", "Juho Lee", "Francois Caron" ]
null
null
Bayesian nonparametric approaches, in particular the Pitman-Yor process and the associated two-parameter Chinese Restaurant process, have been successfully used in applications where the data exhibit a power-law behavior. Examples include natural language processing, natural images or networks. There is also growing em...
[]
null
40
1902.04714
title_snapshot
[ -0.017660409212112427, 0.012768448330461979, -0.039055489003658295, 0.012548888102173805, 0.03353333845734596, 0.03222781419754028, 0.01592179760336876, 0.03523753955960274, 0.0038500039372593164, -0.03291258588433266, -0.0025874844286590815, 0.003308885730803013, -0.05997190251946449, -0....
Scalable Fair Clustering
https://proceedings.mlr.press/v97/backurs19a.html
[ "Arturs Backurs", "Piotr Indyk", "Krzysztof Onak", "Baruch Schieber", "Ali Vakilian", "Tal Wagner" ]
null
null
We study the fair variant of the classic k-median problem introduced by (Chierichetti et al., NeurIPS 2017) in which the points are colored, and the goal is to minimize the same average distance objective as in the standard $k$-median problem while ensuring that all clusters have an “approximately equal” number of poin...
[]
null
41
1902.03519
title_snapshot
[ -0.009186210110783577, -0.0285185556858778, 0.013395760208368301, 0.04350198805332184, 0.03832319378852844, 0.045482657849788666, -0.01387955341488123, 0.007324937731027603, -0.0652289092540741, -0.0465562641620636, -0.0013641887344419956, -0.06107863783836365, -0.08105087280273438, -0.006...
Entropic GANs meet VAEs: A Statistical Approach to Compute Sample Likelihoods in GANs
https://proceedings.mlr.press/v97/balaji19a.html
[ "Yogesh Balaji", "Hamed Hassani", "Rama Chellappa", "Soheil Feizi" ]
null
null
Building on the success of deep learning, two modern approaches to learn a probability model from the data are Generative Adversarial Networks (GANs) and Variational AutoEncoders (VAEs). VAEs consider an explicit probability model for the data and compute a generative distribution by maximizing a variational lower-boun...
[]
null
42
1810.04147
title_snapshot
[ 0.0034943351056426764, -0.00717892637476325, -0.004978072829544544, 0.048962999135255814, 0.030699562281370163, 0.039318181574344635, 0.034110985696315765, -0.004772151820361614, -0.015893330797553062, -0.06757859140634537, -0.04514158517122269, 0.017097096890211105, -0.053217943757772446, ...
Provable Guarantees for Gradient-Based Meta-Learning
https://proceedings.mlr.press/v97/balcan19a.html
[ "Maria-Florina Balcan", "Mikhail Khodak", "Ameet Talwalkar" ]
null
null
We study the problem of meta-learning through the lens of online convex optimization, developing a meta-algorithm bridging the gap between popular gradient-based meta-learning and classical regularization-based multi-task transfer methods. Our method is the first to simultaneously satisfy good sample efficiency guarant...
[]
null
43
1902.10644
title_snapshot
[ -0.014501035213470459, -0.017509520053863525, -0.0005523320869542658, 0.0422002412378788, 0.04365403205156326, 0.02887272648513317, 0.039131924510002136, 0.005184888374060392, -0.018515000119805336, -0.016430361196398735, -0.030067427083849907, 0.016506098210811615, -0.05695938691496849, -...
Open-ended learning in symmetric zero-sum games
https://proceedings.mlr.press/v97/balduzzi19a.html
[ "David Balduzzi", "Marta Garnelo", "Yoram Bachrach", "Wojciech Czarnecki", "Julien Perolat", "Max Jaderberg", "Thore Graepel" ]
null
null
Zero-sum games such as chess and poker are, abstractly, functions that evaluate pairs of agents, for example labeling them ‘winner’ and ‘loser’. If the game is approximately transitive, then self-play generates sequences of agents of increasing strength. However, nontransitive games, such as rock-paper-scissors, can ex...
[]
null
44
1901.08106
title_snapshot
[ -0.050890807062387466, -0.018562057986855507, 0.004262853413820267, 0.02406257763504982, 0.02358897030353546, 0.03058573603630066, 0.004994646180421114, 0.03399822115898132, -0.024100542068481445, -0.0342889279127121, -0.0010832910193130374, 0.018178628757596016, -0.0856068879365921, -0.01...
Concrete Autoencoders: Differentiable Feature Selection and Reconstruction
https://proceedings.mlr.press/v97/balin19a.html
[ "Muhammed Fatih Balın", "Abubakar Abid", "James Zou" ]
null
null
We introduce the concrete autoencoder, an end-to-end differentiable method for global feature selection, which efficiently identifies a subset of the most informative features and simultaneously learns a neural network to reconstruct the input data from the selected features. Our method is unsupervised, and is based on...
[]
null
45
1901.09346
title_judge
[ 0.02352130599319935, -0.02082272246479988, -0.021437764167785645, 0.03483206033706665, 0.07562305778265, 0.06664722412824631, 0.04716595634818077, -0.010341878049075603, 0.008867545053362846, -0.05336013808846474, 0.0028061664197593927, -0.017994070425629616, -0.07118147611618042, 0.004949...
HOList: An Environment for Machine Learning of Higher Order Logic Theorem Proving
https://proceedings.mlr.press/v97/bansal19a.html
[ "Kshitij Bansal", "Sarah Loos", "Markus Rabe", "Christian Szegedy", "Stewart Wilcox" ]
null
null
We present an environment, benchmark, and deep learning driven automated theorem prover for higher-order logic. Higher-order interactive theorem provers enable the formalization of arbitrary mathematical theories and thereby present an interesting challenge for deep learning. We provide an open-source framework based o...
[]
null
46
1904.03241
title_judge
[ -0.04054943099617958, 0.028634410351514816, -0.002473343862220645, 0.03328419476747513, 0.04249771311879158, -0.002689594170078635, 0.007681539747864008, -0.0036469537299126387, -0.0382574200630188, 0.013735687360167503, 0.004167989827692509, 0.049035556614398956, -0.06930488348007202, -0....
Structured agents for physical construction
https://proceedings.mlr.press/v97/bapst19a.html
[ "Victor Bapst", "Alvaro Sanchez-Gonzalez", "Carl Doersch", "Kimberly Stachenfeld", "Pushmeet Kohli", "Peter Battaglia", "Jessica Hamrick" ]
null
null
Physical construction—the ability to compose objects, subject to physical dynamics, to serve some function—is fundamental to human intelligence. We introduce a suite of challenging physical construction tasks inspired by how children play with blocks, such as matching a target configuration, stacking blocks to connect ...
[]
null
47
1904.03177
title_snapshot
[ 0.02420433983206749, -0.0059991395100951195, -0.013640444725751877, 0.05388571321964264, 0.06585676223039627, 0.03196099400520325, 0.0039736186154186726, 0.0010312228696420789, -0.00847160816192627, -0.047232866287231445, -0.04702897369861603, 0.0028998227789998055, -0.07990618795156479, -...
Learning to Route in Similarity Graphs
https://proceedings.mlr.press/v97/baranchuk19a.html
[ "Dmitry Baranchuk", "Dmitry Persiyanov", "Anton Sinitsin", "Artem Babenko" ]
null
null
Recently similarity graphs became the leading paradigm for efficient nearest neighbor search, outperforming traditional tree-based and LSH-based methods. Similarity graphs perform the search via greedy routing: a query traverses the graph and in each vertex moves to the adjacent vertex that is the closest to this query...
[]
null
48
1905.10987
title_snapshot
[ -0.008096657693386078, -0.0281866192817688, 0.008061422035098076, 0.05159163847565651, 0.043465640395879745, 0.02897181361913681, 0.0159535463899374, 0.006615323945879936, -0.0056420788168907166, -0.06213158369064331, 0.016488416120409966, -0.038736049085855484, -0.06975706666707993, 0.011...
A Personalized Affective Memory Model for Improving Emotion Recognition
https://proceedings.mlr.press/v97/barros19a.html
[ "Pablo Barros", "German Parisi", "Stefan Wermter" ]
null
null
Recent models of emotion recognition strongly rely on supervised deep learning solutions for the distinction of general emotion expressions. However, they are not reliable when recognizing online and personalized facial expressions, e.g., for person-specific affective understanding. In this paper, we present a neural m...
[]
null
49
1904.12632
title_judge
[ -0.008542847819626331, 0.005972254555672407, -0.011885798536241055, 0.03744407370686531, 0.027598831802606583, 0.03544187918305397, 0.029600732028484344, 0.00948408804833889, -0.014873501844704151, -0.03793366253376007, -0.03793651983141899, 0.0018235406605526805, -0.04155464470386505, -0....
Scale-free adaptive planning for deterministic dynamics & discounted rewards
https://proceedings.mlr.press/v97/bartlett19a.html
[ "Peter Bartlett", "Victor Gabillon", "Jennifer Healey", "Michal Valko" ]
null
null
We address the problem of planning in an environment with deterministic dynamics and stochastic discounted rewards under a limited numerical budget where the ranges of both rewards and noise are unknown. We introduce PlaTypOOS, an adaptive, robust, and efficient alternative to the OLOP (open-loop optimistic planning) a...
[]
null
50
2604.18312
title_snapshot
[ -0.03526628389954567, -0.015449387952685356, -0.004678144119679928, 0.039804644882678986, 0.0498565249145031, 0.05702313035726547, -0.00461507448926568, 0.01110968366265297, -0.06442362815141678, -0.05066641792654991, -0.01565994694828987, 0.008223196491599083, -0.0658402144908905, -0.0188...
Pareto Optimal Streaming Unsupervised Classification
https://proceedings.mlr.press/v97/basu19a.html
[ "Soumya Basu", "Steven Gutstein", "Brent Lance", "Sanjay Shakkottai" ]
null
null
We study an online and streaming unsupervised classification system. Our setting consists of a collection of classifiers (with unknown confusion matrices) each of which can classify one sample per unit time, and which are accessed by a stream of unlabeled samples. Each sample is dispatched to one or more classifiers, a...
[]
null
51
null
null
[ -0.014313728548586369, -0.024111930280923843, 0.003946197684854269, 0.03501194342970848, 0.025146542116999626, 0.03749094903469086, -0.01197418849915266, 0.0028729059267789125, -0.03370725363492966, -0.04846012964844704, -0.03410349786281586, 0.0006512415129691362, -0.08143986016511917, -0...
Categorical Feature Compression via Submodular Optimization
https://proceedings.mlr.press/v97/bateni19a.html
[ "Mohammadhossein Bateni", "Lin Chen", "Hossein Esfandiari", "Thomas Fu", "Vahab Mirrokni", "Afshin Rostamizadeh" ]
null
null
In the era of big data, learning from categorical features with very large vocabularies (e.g., 28 million for the Criteo click prediction dataset) has become a practical challenge for machine learning researchers and practitioners. We design a highly-scalable vocabulary compression algorithm that seeks to maximize the ...
[]
null
52
1904.13389
title_snapshot
[ -0.026331180706620216, -0.024854570627212524, -0.01283295638859272, 0.04641370102763176, 0.04481286182999611, 0.04596252739429474, 0.018279584124684334, -0.013416420668363571, -0.011013709008693695, -0.0172300785779953, -0.02198600582778454, 0.010183366015553474, -0.06301505118608475, -0.0...
Noise2Self: Blind Denoising by Self-Supervision
https://proceedings.mlr.press/v97/batson19a.html
[ "Joshua Batson", "Loic Royer" ]
null
null
We propose a general framework for denoising high-dimensional measurements which requires no prior on the signal, no estimate of the noise, and no clean training data. The only assumption is that the noise exhibits statistical independence across different dimensions of the measurement, while the true signal exhibits s...
[]
null
53
1901.11365
title_snapshot
[ -0.0005689770332537591, 0.0036179472226649523, 0.004293454345315695, 0.013671920634806156, 0.01862451620399952, 0.04627919942140579, 0.04853527620434761, -0.017967989668250084, -0.023071786388754845, -0.06419631093740463, 0.005057281348854303, 0.011689159087836742, -0.07254143804311752, 0....
Efficient optimization of loops and limits with randomized telescoping sums
https://proceedings.mlr.press/v97/beatson19a.html
[ "Alex Beatson", "Ryan P Adams" ]
null
null
We consider optimization problems in which the objective requires an inner loop with many steps or is the limit of a sequence of increasingly costly approximations. Meta-learning, training recurrent neural networks, and optimization of the solutions to differential equations are all examples of optimization problems wi...
[]
null
54
1905.07006
title_snapshot
[ 0.011256976053118706, 0.006288034841418266, 0.005164484027773142, 0.040332213044166565, 0.034543998539447784, 0.05239594355225563, 0.03502881899476051, 0.005676646251231432, -0.030886422842741013, -0.021822519600391388, -0.02201125957071781, 0.008229995146393776, -0.07447424530982971, -0.0...
Recurrent Kalman Networks: Factorized Inference in High-Dimensional Deep Feature Spaces
https://proceedings.mlr.press/v97/becker19a.html
[ "Philipp Becker", "Harit Pandya", "Gregor Gebhardt", "Cheng Zhao", "C. James Taylor", "Gerhard Neumann" ]
null
null
In order to integrate uncertainty estimates into deep time-series modelling, Kalman Filters (KFs) (Kalman et al., 1960) have been integrated with deep learning models, however, such approaches typically rely on approximate inference tech- niques such as variational inference which makes learning more complex and often ...
[]
null
55
1905.07357
title_snapshot
[ -0.000032754858693806455, -0.03169810026884079, -0.02381422184407711, 0.03835428133606911, 0.04612426087260246, 0.05124573037028313, 0.035269636660814285, 0.017248988151550293, -0.012525648809969425, -0.04640384763479233, 0.002547508804127574, -0.028055235743522644, -0.0558166429400444, 0....
Switching Linear Dynamics for Variational Bayes Filtering
https://proceedings.mlr.press/v97/becker-ehmck19a.html
[ "Philip Becker-Ehmck", "Jan Peters", "Patrick Van Der Smagt" ]
null
null
System identification of complex and nonlinear systems is a central problem for model predictive control and model-based reinforcement learning. Despite their complexity, such systems can often be approximated well by a set of linear dynamical systems if broken into appropriate subsequences. This mechanism not only hel...
[]
null
56
1905.12434
title_snapshot
[ -0.030323227867484093, 0.028398865833878517, -0.01470055989921093, 0.03393479436635971, 0.05631637200713158, 0.024236438795924187, 0.0131635507568717, -0.005484637338668108, -0.06272774189710617, -0.04486490413546562, -0.021283697336912155, 0.012328398413956165, -0.0655164122581482, 0.0023...
Active Learning for Probabilistic Structured Prediction of Cuts and Matchings
https://proceedings.mlr.press/v97/behpour19a.html
[ "Sima Behpour", "Anqi Liu", "Brian Ziebart" ]
null
null
Active learning methods, like uncertainty sampling, combined with probabilistic prediction techniques have achieved success in various problems like image classification and text classification. For more complex multivariate prediction tasks, the relationships between labels play an important role in designing structur...
[]
null
57
null
null
[ 0.026927800849080086, -0.015452384017407894, -0.02210851199924946, 0.028963197022676468, 0.03169157728552818, 0.026734476909041405, -0.006198063958436251, -0.030579715967178345, -0.0015940888551995158, -0.046290166676044464, -0.03977743536233902, 0.018952427431941032, -0.0788193792104721, ...
Invertible Residual Networks
https://proceedings.mlr.press/v97/behrmann19a.html
[ "Jens Behrmann", "Will Grathwohl", "Ricky T. Q. Chen", "David Duvenaud", "Joern-Henrik Jacobsen" ]
null
null
We show that standard ResNet architectures can be made invertible, allowing the same model to be used for classification, density estimation, and generation. Typically, enforcing invertibility requires partitioning dimensions or restricting network architectures. In contrast, our approach only requires adding a simple ...
[]
null
58
1811.00995
title_snapshot
[ 0.03306695073843002, -0.028573865070939064, -0.0031767713371664286, 0.0376901738345623, 0.044901393353939056, 0.04314771294593811, 0.016713552176952362, -0.02426246926188469, -0.0313882939517498, -0.05007356032729149, 0.015438004396855831, -0.012775581330060959, -0.04154142737388611, -0.00...
Greedy Layerwise Learning Can Scale To ImageNet
https://proceedings.mlr.press/v97/belilovsky19a.html
[ "Eugene Belilovsky", "Michael Eickenberg", "Edouard Oyallon" ]
null
null
Shallow supervised 1-hidden layer neural networks have a number of favorable properties that make them easier to interpret, analyze, and optimize than their deep counterparts, but lack their representational power. Here we use 1-hidden layer learning problems to sequentially build deep networks layer by layer, which ca...
[]
null
59
1812.11446
title_snapshot
[ -0.010008953511714935, -0.04953819513320923, 0.008382861502468586, 0.023044932633638382, 0.04949670657515526, 0.01598530448973179, -0.007859580218791962, -0.011098427698016167, -0.024655908346176147, -0.04836496338248253, 0.001824231119826436, -0.013769133947789669, -0.05270364135503769, 0...
Overcoming Multi-model Forgetting
https://proceedings.mlr.press/v97/benyahia19a.html
[ "Yassine Benyahia", "Kaicheng Yu", "Kamil Bennani Smires", "Martin Jaggi", "Anthony C. Davison", "Mathieu Salzmann", "Claudiu Musat" ]
null
null
We identify a phenomenon, which we refer to as multi-model forgetting, that occurs when sequentially training multiple deep networks with partially-shared parameters; the performance of previously-trained models degrades as one optimizes a subsequent one, due to the overwriting of shared parameters. To overcome this, w...
[]
null
60
1902.08232
title_snapshot
[ -0.01998542994260788, -0.002721064258366823, -0.0074097090400755405, 0.03876687213778496, 0.030497707426548004, 0.04096151888370514, 0.007190543692559004, 0.0014531940687447786, -0.04754502698779106, -0.032675378024578094, 0.009417552500963211, 0.015020898543298244, -0.04664923995733261, -...
Optimal Kronecker-Sum Approximation of Real Time Recurrent Learning
https://proceedings.mlr.press/v97/benzing19a.html
[ "Frederik Benzing", "Marcelo Matheus Gauy", "Asier Mujika", "Anders Martinsson", "Angelika Steger" ]
null
null
One of the central goals of Recurrent Neural Networks (RNNs) is to learn long-term dependencies in sequential data. Nevertheless, the most popular training method, Truncated Backpropagation through Time (TBPTT), categorically forbids learning dependencies beyond the truncation horizon. In contrast, the online training ...
[]
null
61
1902.03993
title_snapshot
[ -0.026882229372859, -0.06054502725601196, 0.00082791579188779, 0.03510192781686783, 0.05612903833389282, 0.0417475625872612, 0.020002810284495354, 0.03655422851443291, -0.04276265576481819, -0.011357646435499191, -0.028790496289730072, 0.007132495287805796, -0.05575721710920334, -0.0090582...
Adversarially Learned Representations for Information Obfuscation and Inference
https://proceedings.mlr.press/v97/bertran19a.html
[ "Martin Bertran", "Natalia Martinez", "Afroditi Papadaki", "Qiang Qiu", "Miguel Rodrigues", "Galen Reeves", "Guillermo Sapiro" ]
null
null
Data collection and sharing are pervasive aspects of modern society. This process can either be voluntary, as in the case of a person taking a facial image to unlock his/her phone, or incidental, such as traffic cameras collecting videos on pedestrians. An undesirable side effect of these processes is that shared data ...
[]
null
62
null
null
[ -0.0070071653462946415, -0.010209995321929455, -0.0087268752977252, 0.08108989894390106, 0.03685953840613365, 0.010034442879259586, 0.05026959627866745, -0.0283069871366024, -0.010764902457594872, -0.019186586141586304, -0.009044676087796688, -0.012655864469707012, -0.06589929014444351, -0...
Bandit Multiclass Linear Classification: Efficient Algorithms for the Separable Case
https://proceedings.mlr.press/v97/beygelzimer19a.html
[ "Alina Beygelzimer", "David Pal", "Balazs Szorenyi", "Devanathan Thiruvenkatachari", "Chen-Yu Wei", "Chicheng Zhang" ]
null
null
We study the problem of efficient online multiclass linear classification with bandit feedback, where all examples belong to one of $K$ classes and lie in the $d$-dimensional Euclidean space. Previous works have left open the challenge of designing efficient algorithms with finite mistake bounds when the data is linear...
[]
null
63
1902.02244
title_snapshot
[ -0.00703944219276309, 0.004704952705651522, -0.0002339900383958593, 0.0420709028840065, 0.033342014998197556, 0.028505373746156693, 0.02205945923924446, -0.009175742045044899, -0.01492919772863388, -0.047259461134672165, -0.025180058553814888, 0.0016625041607767344, -0.08638948947191238, -...
Analyzing Federated Learning through an Adversarial Lens
https://proceedings.mlr.press/v97/bhagoji19a.html
[ "Arjun Nitin Bhagoji", "Supriyo Chakraborty", "Prateek Mittal", "Seraphin Calo" ]
null
null
Federated learning distributes model training among a multitude of agents, who, guided by privacy concerns, perform training using their local data but share only model parameter updates, for iterative aggregation at the server to train an overall global model. In this work, we explore how the federated learning settin...
[]
null
64
1811.12470
title_snapshot
[ -0.010251840576529503, -0.04152347892522812, -0.021341418847441673, 0.06458230316638947, 0.032792557030916214, 0.005346059799194336, 0.038748543709516525, -0.031694360077381134, -0.01605292595922947, -0.0362338125705719, -0.017693061381578445, -0.0055977217853069305, -0.06006116420030594, ...
Optimal Continuous DR-Submodular Maximization and Applications to Provable Mean Field Inference
https://proceedings.mlr.press/v97/bian19a.html
[ "Yatao Bian", "Joachim Buhmann", "Andreas Krause" ]
null
null
Mean field inference for discrete graphical models is generally a highly nonconvex problem, which also holds for the class of probabilistic log-submodular models. Existing optimization methods, e.g., coordinate ascent algorithms, typically only find local optima. In this work we propose provable mean filed methods for ...
[]
null
65
1805.07482
title_judge
[ -0.01665213704109192, 0.015687773004174232, 0.0005144107853993773, 0.047585681080818176, 0.0547490268945694, 0.03112773597240448, 0.04870561510324478, -0.010617603547871113, -0.005041572265326977, -0.040372319519519806, 0.01128786988556385, 0.017804717645049095, -0.0701000988483429, -0.008...
More Efficient Off-Policy Evaluation through Regularized Targeted Learning
https://proceedings.mlr.press/v97/bibaut19a.html
[ "Aurelien Bibaut", "Ivana Malenica", "Nikos Vlassis", "Mark Van Der Laan" ]
null
null
We study the problem of off-policy evaluation (OPE) in Reinforcement Learning (RL), where the aim is to estimate the performance of a new policy given historical data that may have been generated by a different policy, or policies. In particular, we introduce a novel doubly-robust estimator for the OPE problem in RL, b...
[]
null
66
1912.06292
title_snapshot
[ -0.010794303379952908, -0.03270742669701576, 0.005842551589012146, 0.012883752584457397, 0.03797508031129837, 0.017934221774339676, 0.021815398707985878, 0.00047314297989942133, -0.03636391460895538, -0.020520713180303574, -0.004056460689753294, 0.022826621308922768, -0.08706573396921158, ...
A Kernel Perspective for Regularizing Deep Neural Networks
https://proceedings.mlr.press/v97/bietti19a.html
[ "Alberto Bietti", "Grégoire Mialon", "Dexiong Chen", "Julien Mairal" ]
null
null
We propose a new point of view for regularizing deep neural networks by using the norm of a reproducing kernel Hilbert space (RKHS). Even though this norm cannot be computed, it admits upper and lower approximations leading to various practical strategies. Specifically, this perspective (i) provides a common umbrella f...
[]
null
67
1810.00363
title_snapshot
[ -0.013849678449332714, -0.03439042717218399, -0.0018925677286460996, 0.03962751105427742, 0.03372262045741081, 0.026988772675395012, 0.03874288871884346, -0.029949666932225227, -0.033583056181669235, -0.04904776066541672, -0.037613119930028915, 0.013577000238001347, -0.057821277529001236, ...
Rethinking Lossy Compression: The Rate-Distortion-Perception Tradeoff
https://proceedings.mlr.press/v97/blau19a.html
[ "Yochai Blau", "Tomer Michaeli" ]
null
null
Lossy compression algorithms are typically designed and analyzed through the lens of Shannon’s rate-distortion theory, where the goal is to achieve the lowest possible distortion (e.g., low MSE or high SSIM) at any given bit rate. However, in recent years, it has become increasingly accepted that "low distortion" is no...
[]
null
68
1901.07821
title_snapshot
[ -0.01841619424521923, -0.002348467940464616, -0.01676510088145733, 0.03704024851322174, 0.04518221691250801, 0.027541479095816612, 0.006849551107734442, 0.018863825127482414, -0.03743354603648186, -0.07638508081436157, 0.006641945336014032, 0.036705609411001205, -0.03342435508966446, -0.00...
Correlated bandits or: How to minimize mean-squared error online
https://proceedings.mlr.press/v97/boda19a.html
[ "Vinay Praneeth Boda", "Prashanth L.A." ]
null
null
While the objective in traditional multi-armed bandit problems is to find the arm with the highest mean, in many settings, finding an arm that best captures information about other arms is of interest. This objective, however, requires learning the underlying correlation structure and not just the means. Sensors placem...
[]
null
69
1902.02953
title_snapshot
[ -0.0007958441274240613, 0.015727365389466286, 0.00558610912412405, 0.03263813629746437, 0.03894027695059776, 0.04569762200117111, 0.04347192123532295, -0.0007825939101167023, -0.026272526010870934, -0.05435894802212715, 0.0015371196204796433, -0.016109973192214966, -0.055936139076948166, -...
Adversarial Attacks on Node Embeddings via Graph Poisoning
https://proceedings.mlr.press/v97/bojchevski19a.html
[ "Aleksandar Bojchevski", "Stephan Günnemann" ]
null
null
The goal of network representation learning is to learn low-dimensional node embeddings that capture the graph structure and are useful for solving downstream tasks. However, despite the proliferation of such methods, there is currently no study of their robustness to adversarial attacks. We provide the first adversari...
[]
null
70
1809.01093
title_snapshot
[ -0.010591383092105389, -0.027694495394825935, -0.0068290987983345985, 0.05137762799859047, 0.027362726628780365, 0.007188780233263969, 0.049512770026922226, -0.015191537328064442, 0.004139351658523083, -0.034019939601421356, 0.004820601083338261, -0.044410280883312225, -0.07131027430295944, ...
Online Variance Reduction with Mixtures
https://proceedings.mlr.press/v97/borsos19a.html
[ "Zalán Borsos", "Sebastian Curi", "Kfir Yehuda Levy", "Andreas Krause" ]
null
null
Adaptive importance sampling for stochastic optimization is a promising approach that offers improved convergence through variance reduction. In this work, we propose a new framework for variance reduction that enables the use of mixtures over predefined sampling distributions, which can naturally encode prior knowledg...
[]
null
71
1903.12416
title_snapshot
[ -0.0067610591650009155, -0.0001266004255739972, 0.026961905881762505, 0.03892974928021431, 0.026688227429986, 0.05727788060903549, 0.02743087336421013, -0.001521410420536995, -0.030993230640888214, -0.05458996072411537, -0.007041835691779852, -0.003907877951860428, -0.04594702646136284, 0....
Compositional Fairness Constraints for Graph Embeddings
https://proceedings.mlr.press/v97/bose19a.html
[ "Avishek Bose", "William Hamilton" ]
null
null
Learning high-quality node embeddings is a key building block for machine learning models that operate on graph data, such as social networks and recommender systems. However, existing graph embedding techniques are unable to cope with fairness constraints, e.g., ensuring that the learned representations do not correla...
[]
null
72
1905.10674
title_snapshot
[ 0.00895768217742443, -0.03375222161412239, -0.014435198158025742, 0.05805632099509239, 0.013434008695185184, 0.02208293229341507, 0.01996062695980072, -0.006494542118161917, -0.026645421981811523, -0.02446661703288555, -0.019998691976070404, -0.006841006223112345, -0.07415267080068588, -0....
Unreproducible Research is Reproducible
https://proceedings.mlr.press/v97/bouthillier19a.html
[ "Xavier Bouthillier", "César Laurent", "Pascal Vincent" ]
null
null
The apparent contradiction in the title is a wordplay on the different meanings attributed to the word reproducible across different scientific fields. What we imply is that unreproducible findings can be built upon reproducible methods. Without denying the importance of facilitating the reproduction of methods, we dee...
[]
null
73
null
null
[ 0.00020278649753890932, -0.034563638269901276, -0.04490356519818306, 0.045780155807733536, 0.07869450747966766, 0.028323538601398468, 0.03381965681910515, 0.010219678282737732, -0.010546638630330563, -0.048394545912742615, -0.01921118050813675, -0.0027543066535145044, -0.0412166528403759, ...
Blended Conditonal Gradients
https://proceedings.mlr.press/v97/braun19a.html
[ "Gábor Braun", "Sebastian Pokutta", "Dan Tu", "Stephen Wright" ]
null
null
We present a blended conditional gradient approach for minimizing a smooth convex function over a polytope P, combining the Frank{–}Wolfe algorithm (also called conditional gradient) with gradient-based steps, different from away steps and pairwise steps, but still achieving linear convergence for strongly convex funct...
[]
null
74
null
null
[ -0.017437215894460678, -0.01722026988863945, 0.011654323898255825, 0.04157356917858124, 0.028076959773898125, 0.0583847276866436, 0.019889947026968002, -0.022070499137043953, -0.024715116247534752, -0.03401557356119156, -0.02003045752644539, 0.0043627009727060795, -0.05128712207078934, -0....
Coresets for Ordered Weighted Clustering
https://proceedings.mlr.press/v97/braverman19a.html
[ "Vladimir Braverman", "Shaofeng H.-C. Jiang", "Robert Krauthgamer", "Xuan Wu" ]
null
null
We design coresets for Ordered k-Median, a generalization of classical clustering problems such as k-Median and k-Center. Its objective function is defined via the Ordered Weighted Averaging (OWA) paradigm of Yager (1988), where data points are weighted according to a predefined weight vector, but in order of their con...
[]
null
75
1903.04351
title_snapshot
[ -0.03684530034661293, -0.021865952759981155, 0.005285118240863085, 0.03603968396782875, 0.028421232476830482, 0.050087787210941315, -0.009416583925485611, 0.006765618920326233, -0.013864325359463692, -0.056765150278806686, -0.017482982948422432, -0.061386607587337494, -0.05223140865564346, ...
Target Tracking for Contextual Bandits: Application to Demand Side Management
https://proceedings.mlr.press/v97/bregere19a.html
[ "Margaux Brégère", "Pierre Gaillard", "Yannig Goude", "Gilles Stoltz" ]
null
null
We propose a contextual-bandit approach for demand side management by offering price incentives. More precisely, a target mean consumption is set at each round and the mean consumption is modeled as a complex function of the distribution of prices sent and of some contextual variables such as the temperature, weather, ...
[]
null
76
1901.09532
title_snapshot
[ -0.022505639120936394, -0.030200239270925522, 0.013607303611934185, 0.030641013756394386, 0.028503071516752243, 0.02537475898861885, 0.011832443997263908, 0.02684047445654869, -0.024979127570986748, -0.025229882448911667, -0.011472982354462147, 0.029097281396389008, -0.0470452606678009, -0...
Active Manifolds: A non-linear analogue to Active Subspaces
https://proceedings.mlr.press/v97/bridges19a.html
[ "Robert Bridges", "Anthony Gruber", "Christopher Felder", "Miki Verma", "Chelsey Hoff" ]
null
null
We present an approach to analyze $C^1(\mathbb{R}^m)$ functions that addresses limitations present in the Active Subspaces (AS) method of Constantine et al. (2014; 2015). Under appropriate hypotheses, our Active Manifolds (AM) method identifies a 1-D curve in the domain (the active manifold) on which nearly all values ...
[]
null
77
1904.13386
title_snapshot
[ -0.04774506017565727, -0.02783438190817833, 0.04558267071843147, 0.011177820153534412, 0.0058790091425180435, 0.012456648051738739, 0.012413477525115013, -0.03935883939266205, -0.028122855350375175, -0.04671315476298332, 0.006365543231368065, -0.012995187193155289, -0.09138445556163788, 0....
Conditioning by adaptive sampling for robust design
https://proceedings.mlr.press/v97/brookes19a.html
[ "David Brookes", "Hahnbeom Park", "Jennifer Listgarten" ]
null
null
We present a method for design problems wherein the goal is to maximize or specify the value of one or more properties of interest (e.g. maximizing the fluorescence of a protein). We assume access to black box, stochastic “oracle" predictive functions, each of which maps from design space to a distribution over propert...
[]
null
78
1901.10060
title_snapshot
[ 0.004793043248355389, 0.014372116886079311, -0.022981520742177963, 0.03617256134748459, 0.07462119311094284, 0.045931167900562286, -0.0038720788434147835, -0.029825421050190926, 0.002503100549802184, -0.05681360885500908, -0.006557524669915438, 0.0007717990665696561, -0.07238662242889404, ...
Extrapolating Beyond Suboptimal Demonstrations via Inverse Reinforcement Learning from Observations
https://proceedings.mlr.press/v97/brown19a.html
[ "Daniel Brown", "Wonjoon Goo", "Prabhat Nagarajan", "Scott Niekum" ]
null
null
A critical flaw of existing inverse reinforcement learning (IRL) methods is their inability to significantly outperform the demonstrator. This is because IRL typically seeks a reward function that makes the demonstrator appear near-optimal, rather than inferring the underlying intentions of the demonstrator that may ha...
[]
null
79
1904.06387
title_snapshot
[ -0.024270879104733467, -0.021370256319642067, -0.009393760934472084, 0.0537717379629612, 0.05723968893289566, 0.01364996936172247, 0.011259366758167744, 0.0018230591667816043, -0.04524311423301697, -0.023269513621926308, -0.006756199989467859, 0.02038022130727768, -0.05424965173006058, -0....
Deep Counterfactual Regret Minimization
https://proceedings.mlr.press/v97/brown19b.html
[ "Noam Brown", "Adam Lerer", "Sam Gross", "Tuomas Sandholm" ]
null
null
Counterfactual Regret Minimization (CFR) is the leading algorithm for solving large imperfect-information games. It converges to an equilibrium by iteratively traversing the game tree. In order to deal with extremely large games, abstraction is typically applied before running CFR. The abstracted game is solved with ta...
[]
null
80
1811.00164
title_snapshot
[ -0.04876405745744705, -0.028748301789164543, 0.02046901173889637, 0.04283217713236809, 0.03946444392204285, 0.005015285685658455, -0.003615735564380884, 0.021493738517165184, -0.028705354779958725, -0.04212091490626335, 0.004697267431765795, 0.03889835253357887, -0.062439125031232834, 0.00...
Understanding the Origins of Bias in Word Embeddings
https://proceedings.mlr.press/v97/brunet19a.html
[ "Marc-Etienne Brunet", "Colleen Alkalay-Houlihan", "Ashton Anderson", "Richard Zemel" ]
null
null
Popular word embedding algorithms exhibit stereotypical biases, such as gender bias. The widespread use of these algorithms in machine learning systems can amplify stereotypes in important contexts. Although some methods have been developed to mitigate this problem, how word embedding biases arise during training is po...
[]
null
81
1810.03611
title_snapshot
[ -0.01921071857213974, -0.024881776422262192, -0.03012724593281746, 0.06846357882022858, 0.031701620668172836, 0.012480562552809715, 0.03385617956519127, 0.01990305632352829, 0.025180868804454803, -0.024887656792998314, -0.02961636893451214, 0.028413543477654457, -0.0594511553645134, -0.004...
Low Latency Privacy Preserving Inference
https://proceedings.mlr.press/v97/brutzkus19a.html
[ "Alon Brutzkus", "Ran Gilad-Bachrach", "Oren Elisha" ]
null
null
When applying machine learning to sensitive data, one has to find a balance between accuracy, information security, and computational-complexity. Recent studies combined Homomorphic Encryption with neural networks to make inferences while protecting against information leakage. However, these methods are limited by the...
[]
null
82
1812.10659
title_snapshot
[ 0.015583324246108532, -0.00510758301243186, -0.016084326431155205, 0.07249930500984192, 0.05290370061993599, 0.009873359464108944, 0.013215465471148491, -0.01830841600894928, -0.012652450241148472, -0.02684924378991127, 0.019355354830622673, -0.02488849125802517, -0.02612501010298729, 0.02...
Why do Larger Models Generalize Better? A Theoretical Perspective via the XOR Problem
https://proceedings.mlr.press/v97/brutzkus19b.html
[ "Alon Brutzkus", "Amir Globerson" ]
null
null
Empirical evidence suggests that neural networks with ReLU activations generalize better with over-parameterization. However, there is currently no theoretical analysis that explains this observation. In this work, we provide theoretical and empirical evidence that, in certain cases, overparameterized convolutional net...
[]
null
83
1810.03037
title_snapshot
[ -0.005107920616865158, -0.02662285603582859, 0.016209132969379425, 0.06007630005478859, 0.023466704413294792, 0.0256633460521698, 0.016231488436460495, 0.028589125722646713, -0.018332447856664658, -0.039222534745931625, 0.004625532776117325, -0.012875907123088837, -0.07015771418809891, 0.0...
Adversarial examples from computational constraints
https://proceedings.mlr.press/v97/bubeck19a.html
[ "Sebastien Bubeck", "Yin Tat Lee", "Eric Price", "Ilya Razenshteyn" ]
null
null
Why are classifiers in high dimension vulnerable to “adversarial” perturbations? We show that it is likely not due to information theoretic limitations, but rather it could be due to computational constraints. First we prove that, for a broad set of classification tasks, the mere existence of a robust classifier implie...
[]
null
84
1805.10204
title_snapshot
[ -0.027904225513339043, -0.015876853838562965, -0.019303997978568077, 0.04620860144495964, 0.021179908886551857, 0.008461279794573784, 0.03777044266462326, -0.02030831202864647, -0.035591185092926025, -0.025233862921595573, -0.025353288277983665, -0.015295282937586308, -0.0731470137834549, ...
Self-similar Epochs: Value in arrangement
https://proceedings.mlr.press/v97/buchnik19a.html
[ "Eliav Buchnik", "Edith Cohen", "Avinatan Hasidim", "Yossi Matias" ]
null
null
Optimization of machine learning models is commonly performed through stochastic gradient updates on randomly ordered training examples. This practice means that each fraction of an epoch comprises an independent random sample of the training data that may not preserve informative structure present in the full data. We...
[]
null
85
1803.05389
title_snapshot
[ -0.011038440279662609, -0.03455266356468201, -0.01255763042718172, 0.043339088559150696, 0.0522204227745533, 0.024802204221487045, 0.020348887890577316, -0.002199045615270734, -0.0006228333804756403, -0.045390963554382324, 0.0020331887062639, -0.020653661340475082, -0.07602641731500626, -0...
Learning Generative Models across Incomparable Spaces
https://proceedings.mlr.press/v97/bunne19a.html
[ "Charlotte Bunne", "David Alvarez-Melis", "Andreas Krause", "Stefanie Jegelka" ]
null
null
Generative Adversarial Networks have shown remarkable success in learning a distribution that faithfully recovers a reference distribution in its entirety. However, in some cases, we may want to only learn some aspects (e.g., cluster or manifold structure), while modifying others (e.g., style, orientation or dimension)...
[]
null
86
1905.05461
title_snapshot
[ -0.014598199166357517, -0.020213864743709564, -0.01724678836762905, 0.05194314941763878, 0.023420462384819984, 0.03485999256372452, 0.015313399024307728, 0.0030277641490101814, -0.003958743531256914, -0.030656160786747932, -0.010785798542201519, -0.0030968994833528996, -0.09920670837163925, ...
Rates of Convergence for Sparse Variational Gaussian Process Regression
https://proceedings.mlr.press/v97/burt19a.html
[ "David Burt", "Carl Edward Rasmussen", "Mark Van Der Wilk" ]
null
null
Excellent variational approximations to Gaussian process posteriors have been developed which avoid the $\mathcal{O}\left(N^3\right)$ scaling with dataset size $N$. They reduce the computational cost to $\mathcal{O}\left(NM^2\right)$, with $M\ll N$ the number ofinducing variables, which summarise the process. While the...
[]
null
87
1903.03571
title_snapshot
[ -0.01201074942946434, -0.0007218557293526828, 0.01406270731240511, 0.03330624848604202, 0.04347844794392586, 0.05741655454039574, 0.01882663555443287, 0.011840627528727055, -0.025954898446798325, -0.025898011401295662, -0.005336322821676731, 0.028592975810170174, -0.0648282989859581, 0.034...
What is the Effect of Importance Weighting in Deep Learning?
https://proceedings.mlr.press/v97/byrd19a.html
[ "Jonathon Byrd", "Zachary Lipton" ]
null
null
Importance-weighted risk minimization is a key ingredient in many machine learning algorithms for causal inference, domain adaptation, class imbalance, and off-policy reinforcement learning. While the effect of importance weighting is well-characterized for low-capacity misspecified models, little is known about how it...
[]
null
88
1812.03372
title_snapshot
[ -0.024407312273979187, -0.03723262622952461, -0.003702374640852213, 0.04117170721292496, 0.030306434258818626, 0.023436373099684715, 0.015449067577719688, -0.006310954689979553, -0.022844068706035614, -0.05697094649076462, -0.041943471878767014, 0.005116458982229233, -0.06898439675569534, ...
A Quantitative Analysis of the Effect of Batch Normalization on Gradient Descent
https://proceedings.mlr.press/v97/cai19a.html
[ "Yongqiang Cai", "Qianxiao Li", "Zuowei Shen" ]
null
null
Despite its empirical success and recent theoretical progress, there generally lacks a quantitative analysis of the effect of batch normalization (BN) on the convergence and stability of gradient descent. In this paper, we provide such an analysis on the simple problem of ordinary least squares (OLS), where the precise...
[]
null
89
1810.00122
title_snapshot
[ -0.05709332227706909, -0.02610219083726406, -0.008606230840086937, 0.017208008095622063, 0.038345471024513245, 0.03362354636192322, 0.037356726825237274, -0.001318480004556477, -0.04543997347354889, -0.0414256788790226, 0.003405065508559346, -0.0023351814597845078, -0.05714643374085426, -0...
Accelerated Linear Convergence of Stochastic Momentum Methods in Wasserstein Distances
https://proceedings.mlr.press/v97/can19a.html
[ "Bugra Can", "Mert Gurbuzbalaban", "Lingjiong Zhu" ]
null
null
Momentum methods such as Polyak’s heavy ball (HB) method, Nesterov’s accelerated gradient (AG) as well as accelerated projected gradient (APG) method have been commonly used in machine learning practice, but their performance is quite sensitive to noise in the gradients. We study these methods under a first-order stoch...
[]
null
90
1901.07445
title_snapshot
[ -0.04024022817611694, 0.005329324398189783, 0.013962493278086185, 0.031223300844430923, 0.013226738199591637, 0.039068084210157394, 0.02518509142100811, 0.030053067952394485, -0.029553692787885666, -0.060475099831819534, 0.0027081884909421206, -0.04606926813721657, -0.05402189865708351, -0...
Active Embedding Search via Noisy Paired Comparisons
https://proceedings.mlr.press/v97/canal19a.html
[ "Gregory Canal", "Andy Massimino", "Mark Davenport", "Christopher Rozell" ]
null
null
Suppose that we wish to estimate a user’s preference vector $w$ from paired comparisons of the form “does user $w$ prefer item $p$ or item $q$?,” where both the user and items are embedded in a low-dimensional Euclidean space with distances that reflect user and item similarities. Such observations arise in numerous se...
[]
null
91
1905.04363
title_snapshot
[ -0.005976332351565361, -0.010105673223733902, -0.006868399679660797, 0.050605665892362595, 0.02281234599649906, 0.027540674433112144, 0.013188699260354042, -0.0012186900712549686, 0.020630773156881332, -0.054075393825769424, -0.025718338787555695, 0.01064877025783062, -0.059461481869220734, ...
Dynamic Learning with Frequent New Product Launches: A Sequential Multinomial Logit Bandit Problem
https://proceedings.mlr.press/v97/cao19a.html
[ "Junyu Cao", "Wei Sun" ]
null
null
Motivated by the phenomenon that companies introduce new products to keep abreast with customers’ rapidly changing tastes, we consider a novel online learning setting where a profit-maximizing seller needs to learn customers’ preferences through offering recommendations, which may contain existing products and new prod...
[]
null
92
1904.12445
title_snapshot
[ -0.01601894572377205, -0.026855485513806343, -0.00475424574688077, 0.038304075598716736, 0.06149640306830406, 0.04478788748383522, 0.008722454309463501, 0.008009379729628563, -0.02089305780827999, -0.014135695993900299, -0.02655094675719738, 0.020308272913098335, -0.045557476580142975, -0....
Competing Against Nash Equilibria in Adversarially Changing Zero-Sum Games
https://proceedings.mlr.press/v97/cardoso19a.html
[ "Adrian Rivera Cardoso", "Jacob Abernethy", "He Wang", "Huan Xu" ]
null
null
We study the problem of repeated play in a zero-sum game in which the payoff matrix may change, in a possibly adversarial fashion, on each round; we call these Online Matrix Games. Finding the Nash Equilibrium (NE) of a two player zero-sum game is core to many problems in statistics, optimization, and economics, and fo...
[]
null
93
1907.07723
title_judge
[ -0.051430243998765945, -0.01763240247964859, 0.004828574601560831, 0.027137745171785355, 0.03547172620892525, 0.025295315310359, 0.009349450469017029, 0.02563965879380703, -0.026466626673936844, -0.04072483256459236, -0.01281718723475933, 0.020562274381518364, -0.06819038093090057, -0.0143...
Automated Model Selection with Bayesian Quadrature
https://proceedings.mlr.press/v97/chai19a.html
[ "Henry Chai", "Jean-Francois Ton", "Michael A. Osborne", "Roman Garnett" ]
null
null
We present a novel technique for tailoring Bayesian quadrature (BQ) to model selection. The state-of-the-art for comparing the evidence of multiple models relies on Monte Carlo methods, which converge slowly and are unreliable for computationally expensive models. Although previous research has shown that BQ offers sam...
[]
null
94
1902.09724
title_snapshot
[ -0.011786404997110367, -0.013640448451042175, -0.042787544429302216, 0.018569549545645714, 0.03554952144622803, 0.03969927132129669, 0.022165196016430855, -0.03693380579352379, -0.045257244259119034, -0.0542125329375267, -0.003067456418648362, 0.048163946717977524, -0.06792557239532471, -0...
Learning Action Representations for Reinforcement Learning
https://proceedings.mlr.press/v97/chandak19a.html
[ "Yash Chandak", "Georgios Theocharous", "James Kostas", "Scott Jordan", "Philip Thomas" ]
null
null
Most model-free reinforcement learning methods leverage state representations (embeddings) for generalization, but either ignore structure in the space of actions or assume the structure is provided a priori. We show how a policy can be decomposed into a component that acts in a low-dimensional space of action represen...
[]
null
95
1902.00183
title_snapshot
[ -0.016201404854655266, -0.02957540564239025, -0.015056435018777847, 0.04812026768922806, 0.045221876353025436, 0.02241319790482521, 0.008634240366518497, -0.01022863294929266, -0.007791484240442514, -0.02488579787313938, -0.012012787163257599, 0.02215588092803955, -0.09382452070713043, 0.0...
Dynamic Measurement Scheduling for Event Forecasting using Deep RL
https://proceedings.mlr.press/v97/chang19a.html
[ "Chun-Hao Chang", "Mingjie Mai", "Anna Goldenberg" ]
null
null
Imagine a patient in critical condition. What and when should be measured to forecast detrimental events, especially under the budget constraints? We answer this question by deep reinforcement learning (RL) that jointly minimizes the measurement cost and maximizes predictive gain, by scheduling strategically-timed meas...
[]
null
96
1901.09699
title_snapshot
[ -0.026993295177817345, -0.045636676251888275, 0.0006410372443497181, 0.03565984591841698, 0.052510906010866165, 0.03824548423290253, 0.04270828887820244, -0.012330124154686928, -0.02743176370859146, -0.02851243130862713, 0.008824136108160019, 0.008450735360383987, -0.045263320207595825, -0...
On Symmetric Losses for Learning from Corrupted Labels
https://proceedings.mlr.press/v97/charoenphakdee19a.html
[ "Nontawat Charoenphakdee", "Jongyeong Lee", "Masashi Sugiyama" ]
null
null
This paper aims to provide a better understanding of a symmetric loss. First, we emphasize that using a symmetric loss is advantageous in the balanced error rate (BER) minimization and area under the receiver operating characteristic curve (AUC) maximization from corrupted labels. Second, we prove general theoretical p...
[]
null
97
1901.09314
title_snapshot
[ -0.02393936552107334, -0.0013262460706755519, -0.0015298877842724323, 0.04712142050266266, 0.010991035960614681, 0.03335827589035034, 0.0024733601603657007, -0.010560719296336174, -0.025942077860236168, -0.042180728167295456, -0.007291228510439396, 0.010185467079281807, -0.050599467009305954...
Online learning with kernel losses
https://proceedings.mlr.press/v97/chatterji19a.html
[ "Niladri Chatterji", "Aldo Pacchiano", "Peter Bartlett" ]
null
null
We present a generalization of the adversarial linear bandits framework, where the underlying losses are kernel functions (with an associated reproducing kernel Hilbert space) rather than linear functions. We study a version of the exponential weights algorithm and bound its regret in this setting. Under conditions on ...
[]
null
98
1802.09732
title_snapshot
[ -0.022172970697283745, -0.017561599612236023, 0.030192555859684944, 0.06280666589736938, 0.019799474626779556, 0.03556736186146736, 0.019643565639853477, 0.00013276339450385422, -0.0027630391996353865, -0.04443414509296417, -0.04068951681256294, 0.02420528046786785, -0.05476433411240578, -...
Neural Network Attributions: A Causal Perspective
https://proceedings.mlr.press/v97/chattopadhyay19a.html
[ "Aditya Chattopadhyay", "Piyushi Manupriya", "Anirban Sarkar", "Vineeth N Balasubramanian" ]
null
null
We propose a new attribution method for neural networks developed using first principles of causality (to the best of our knowledge, the first such). The neural network architecture is viewed as a Structural Causal Model, and a methodology to compute the causal effect of each feature on the output is presented. With reas...
[]
null
99
1902.02302
title_snapshot
[ -0.01603395864367485, -0.0023921842221170664, -0.016697056591510773, 0.03463682159781456, 0.039350446313619614, 0.05724656954407692, 0.005246924236416817, 0.028155460953712463, -0.03509123623371124, -0.04846108332276344, 0.0030004619620740414, 0.009447415359318256, -0.0522565096616745, -0....
PAC Identification of Many Good Arms in Stochastic Multi-Armed Bandits
https://proceedings.mlr.press/v97/chaudhuri19a.html
[ "Arghya Roy Chaudhuri", "Shivaram Kalyanakrishnan" ]
null
null
We consider the problem of identifying any k out of the best m arms in an n-armed stochastic multi-armed bandit; framed in the PAC setting, this particular problem generalises both the problem of “best subset selection” (Kalyanakrishnan & Stone, 2010) and that of selecting “one out of the best m” arms (Roy Chaudhuri & ...
[]
null
100
1901.08386
title_snapshot
[ -0.03162824735045433, -0.010771496221423149, -0.01647985726594925, 0.05219089612364769, 0.03400689736008644, 0.030414214357733727, 0.022434178739786148, -0.011744238436222076, -0.020796194672584534, -0.04419979080557823, -0.009874433279037476, -0.010969928465783596, -0.052993081510066986, ...