paper_id
stringlengths
9
24
title
stringlengths
10
192
paper_url
stringlengths
44
59
authors
listlengths
1
23
type
stringclasses
0 values
primary_area
stringclasses
0 values
abstract
large_stringlengths
82
4.49k
keywords
listlengths
0
0
TL;DR
large_stringclasses
0 values
submission_number
int64
1
1.18k
arxiv_id
stringlengths
10
10
arxiv_id_source
stringclasses
2 values
v139/feng21g
Understanding Noise Injection in GANs
https://proceedings.mlr.press/v139/feng21g.html
[ "Ruili Feng", "Deli Zhao", "Zheng-Jun Zha" ]
null
null
Noise injection is an effective way of circumventing overfitting and enhancing generalization in machine learning, the rationale of which has been validated in deep learning as well. Recently, noise injection exhibits surprising effectiveness when generating high-fidelity images in Generative Adversarial Networks (GANs...
[]
null
301
null
null
v139/fey21a
GNNAutoScale: Scalable and Expressive Graph Neural Networks via Historical Embeddings
https://proceedings.mlr.press/v139/fey21a.html
[ "Matthias Fey", "Jan E. Lenssen", "Frank Weichert", "Jure Leskovec" ]
null
null
We present GNNAutoScale (GAS), a framework for scaling arbitrary message-passing GNNs to large graphs. GAS prunes entire sub-trees of the computation graph by utilizing historical embeddings from prior training iterations, leading to constant GPU memory consumption in respect to input node size without dropping any dat...
[]
null
302
2106.05609
title_snapshot
v139/filos21a
PsiPhi-Learning: Reinforcement Learning with Demonstrations using Successor Features and Inverse Temporal Difference Learning
https://proceedings.mlr.press/v139/filos21a.html
[ "Angelos Filos", "Clare Lyle", "Yarin Gal", "Sergey Levine", "Natasha Jaques", "Gregory Farquhar" ]
null
null
We study reinforcement learning (RL) with no-reward demonstrations, a setting in which an RL agent has access to additional data from the interaction of other agents with the same environment. However, it has no access to the rewards or goals of these agents, and their objectives and levels of expertise may vary widely...
[]
null
303
2102.12560
title_snapshot
v139/finzi21a
A Practical Method for Constructing Equivariant Multilayer Perceptrons for Arbitrary Matrix Groups
https://proceedings.mlr.press/v139/finzi21a.html
[ "Marc Finzi", "Max Welling", "Andrew Gordon Wilson" ]
null
null
Symmetries and equivariance are fundamental to the generalization of neural networks on domains such as images, graphs, and point clouds. Existing work has primarily focused on a small number of groups, such as the translation, rotation, and permutation groups. In this work we provide a completely general algorithm for...
[]
null
304
2104.09459
title_snapshot
v139/fisch21a
Few-Shot Conformal Prediction with Auxiliary Tasks
https://proceedings.mlr.press/v139/fisch21a.html
[ "Adam Fisch", "Tal Schuster", "Tommi Jaakkola", "Dr.Regina Barzilay" ]
null
null
We develop a novel approach to conformal prediction when the target task has limited data available for training. Conformal prediction identifies a small set of promising output candidates in place of a single prediction, with guarantees that the set contains the correct answer with high probability. When training data...
[]
null
305
2102.08898
title_snapshot
v139/fischer21a
Scalable Certified Segmentation via Randomized Smoothing
https://proceedings.mlr.press/v139/fischer21a.html
[ "Marc Fischer", "Maximilian Baader", "Martin Vechev" ]
null
null
We present a new certification method for image and point cloud segmentation based on randomized smoothing. The method leverages a novel scalable algorithm for prediction and certification that correctly accounts for multiple testing, necessary for ensuring statistical guarantees. The key to our approach is reliance on...
[]
null
306
2107.00228
title_snapshot
v139/fischer21b
What’s in the Box? Exploring the Inner Life of Neural Networks with Robust Rules
https://proceedings.mlr.press/v139/fischer21b.html
[ "Jonas Fischer", "Anna Olah", "Jilles Vreeken" ]
null
null
We propose a novel method for exploring how neurons within neural networks interact. In particular, we consider activation values of a network for given data, and propose to mine noise-robust rules of the form X {\rightarrow} Y , where X and Y are sets of neurons in different layers. We identify the best set of rules b...
[]
null
307
null
null
v139/flaspohler21a
Online Learning with Optimism and Delay
https://proceedings.mlr.press/v139/flaspohler21a.html
[ "Genevieve E Flaspohler", "Francesco Orabona", "Judah Cohen", "Soukayna Mouatadid", "Miruna Oprescu", "Paulo Orenstein", "Lester Mackey" ]
null
null
Inspired by the demands of real-time climate and weather forecasting, we develop optimistic online learning algorithms that require no parameter tuning and have optimal regret guarantees under delayed feedback. Our algorithms—DORM, DORM+, and AdaHedgeD—arise from a novel reduction of delayed online learning to optimist...
[]
null
308
2106.06885
title_snapshot
v139/fontaine21a
Online A-Optimal Design and Active Linear Regression
https://proceedings.mlr.press/v139/fontaine21a.html
[ "Xavier Fontaine", "Pierre Perrault", "Michal Valko", "Vianney Perchet" ]
null
null
We consider in this paper the problem of optimal experiment design where a decision maker can choose which points to sample to obtain an estimate $\hat{\beta}$ of the hidden parameter $\beta^{\star}$ of an underlying linear model. The key challenge of this work lies in the heteroscedasticity assumption that we make, me...
[]
null
309
1906.08509
title_snapshot
v139/foster21a
Deep Adaptive Design: Amortizing Sequential Bayesian Experimental Design
https://proceedings.mlr.press/v139/foster21a.html
[ "Adam Foster", "Desi R Ivanova", "Ilyas Malik", "Tom Rainforth" ]
null
null
We introduce Deep Adaptive Design (DAD), a method for amortizing the cost of adaptive Bayesian experimental design that allows experiments to be run in real-time. Traditional sequential Bayesian optimal experimental design approaches require substantial computation at each stage of the experiment. This makes them unsui...
[]
null
310
2103.02438
title_snapshot
v139/fotakis21a
Efficient Online Learning for Dynamic k-Clustering
https://proceedings.mlr.press/v139/fotakis21a.html
[ "Dimitris Fotakis", "Georgios Piliouras", "Stratis Skoulakis" ]
null
null
In this work, we study dynamic clustering problems from the perspective of online learning. We consider an online learning problem, called \textit{Dynamic $k$-Clustering}, in which $k$ centers are maintained in a metric space over time (centers may change positions) such as a dynamically changing set of $r$ clients is ...
[]
null
311
2106.04336
title_snapshot
v139/fraboni21a
Clustered Sampling: Low-Variance and Improved Representativity for Clients Selection in Federated Learning
https://proceedings.mlr.press/v139/fraboni21a.html
[ "Yann Fraboni", "Richard Vidal", "Laetitia Kameni", "Marco Lorenzi" ]
null
null
This work addresses the problem of optimizing communications between server and clients in federated learning (FL). Current sampling approaches in FL are either biased, or non optimal in terms of server-clients communications and training stability. To overcome this issue, we introduce clustered sampling for clients se...
[]
null
312
2105.05883
title_snapshot
v139/frei21a
Agnostic Learning of Halfspaces with Gradient Descent via Soft Margins
https://proceedings.mlr.press/v139/frei21a.html
[ "Spencer Frei", "Yuan Cao", "Quanquan Gu" ]
null
null
We analyze the properties of gradient descent on convex surrogates for the zero-one loss for the agnostic learning of halfspaces. We show that when a quantity we refer to as the \textit{soft margin} is well-behaved—a condition satisfied by log-concave isotropic distributions among others—minimizers of convex surrogates...
[]
null
313
2010.00539
title_snapshot
v139/frei21b
Provable Generalization of SGD-trained Neural Networks of Any Width in the Presence of Adversarial Label Noise
https://proceedings.mlr.press/v139/frei21b.html
[ "Spencer Frei", "Yuan Cao", "Quanquan Gu" ]
null
null
We consider a one-hidden-layer leaky ReLU network of arbitrary width trained by stochastic gradient descent (SGD) following an arbitrary initialization. We prove that SGD produces neural networks that have classification accuracy competitive with that of the best halfspace over the distribution for a broad class of dis...
[]
null
314
2101.01152
title_snapshot
v139/freidling21a
Post-selection inference with HSIC-Lasso
https://proceedings.mlr.press/v139/freidling21a.html
[ "Tobias Freidling", "Benjamin Poignard", "Héctor Climente-González", "Makoto Yamada" ]
null
null
Detecting influential features in non-linear and/or high-dimensional data is a challenging and increasingly important task in machine learning. Variable selection methods have thus been gaining much attention as well as post-selection inference. Indeed, the selected features can be significantly flawed when the selecti...
[]
null
315
2010.15659
title_snapshot
v139/frerix21a
Variational Data Assimilation with a Learned Inverse Observation Operator
https://proceedings.mlr.press/v139/frerix21a.html
[ "Thomas Frerix", "Dmitrii Kochkov", "Jamie Smith", "Daniel Cremers", "Michael Brenner", "Stephan Hoyer" ]
null
null
Variational data assimilation optimizes for an initial state of a dynamical system such that its evolution fits observational data. The physical model can subsequently be evolved into the future to make predictions. This principle is a cornerstone of large scale forecasting applications such as numerical weather predic...
[]
null
316
2102.11192
title_snapshot
v139/frohlich21a
Bayesian Quadrature on Riemannian Data Manifolds
https://proceedings.mlr.press/v139/frohlich21a.html
[ "Christian Fröhlich", "Alexandra Gessner", "Philipp Hennig", "Bernhard Schölkopf", "Georgios Arvanitidis" ]
null
null
Riemannian manifolds provide a principled way to model nonlinear geometric structure inherent in data. A Riemannian metric on said manifolds determines geometry-aware shortest paths and provides the means to define statistical models accordingly. However, these operations are typically computationally demanding. To eas...
[]
null
317
2102.06645
title_snapshot
v139/fu21a
Learn-to-Share: A Hardware-friendly Transfer Learning Framework Exploiting Computation and Parameter Sharing
https://proceedings.mlr.press/v139/fu21a.html
[ "Cheng Fu", "Hanxian Huang", "Xinyun Chen", "Yuandong Tian", "Jishen Zhao" ]
null
null
Task-specific fine-tuning on pre-trained transformers has achieved performance breakthroughs in multiple NLP tasks. Yet, as both computation and parameter size grows linearly with the number of sub-tasks, it is increasingly difficult to adopt such methods to the real world due to unrealistic memory and computation over...
[]
null
318
null
null
v139/fu21b
Learning Task Informed Abstractions
https://proceedings.mlr.press/v139/fu21b.html
[ "Xiang Fu", "Ge Yang", "Pulkit Agrawal", "Tommi Jaakkola" ]
null
null
Current model-based reinforcement learning methods struggle when operating from complex visual scenes due to their inability to prioritize task-relevant features. To mitigate this problem, we propose learning Task Informed Abstractions (TIA) that explicitly separates reward-correlated visual features from distractors. ...
[]
null
319
2106.15612
title_snapshot
v139/fu21c
Double-Win Quant: Aggressively Winning Robustness of Quantized Deep Neural Networks via Random Precision Training and Inference
https://proceedings.mlr.press/v139/fu21c.html
[ "Yonggan Fu", "Qixuan Yu", "Meng Li", "Vikas Chandra", "Yingyan Lin" ]
null
null
Quantization is promising in enabling powerful yet complex deep neural networks (DNNs) to be deployed into resource constrained platforms. However, quantized DNNs are vulnerable to adversarial attacks unless being equipped with sophisticated techniques, leading to a dilemma of struggling between DNNs’ efficiency and ro...
[]
null
320
null
null
v139/fu21d
Auto-NBA: Efficient and Effective Search Over the Joint Space of Networks, Bitwidths, and Accelerators
https://proceedings.mlr.press/v139/fu21d.html
[ "Yonggan Fu", "Yongan Zhang", "Yang Zhang", "David Cox", "Yingyan Lin" ]
null
null
While maximizing deep neural networks’ (DNNs’) acceleration efficiency requires a joint search/design of three different yet highly coupled aspects, including the networks, bitwidths, and accelerators, the challenges associated with such a joint search have not yet been fully understood and addressed. The key challenge...
[]
null
321
2106.06575
title_snapshot
v139/fujimoto21a
A Deep Reinforcement Learning Approach to Marginalized Importance Sampling with the Successor Representation
https://proceedings.mlr.press/v139/fujimoto21a.html
[ "Scott Fujimoto", "David Meger", "Doina Precup" ]
null
null
Marginalized importance sampling (MIS), which measures the density ratio between the state-action occupancy of a target policy and that of a sampling distribution, is a promising approach for off-policy evaluation. However, current state-of-the-art MIS methods rely on complex optimization tricks and succeed mostly on s...
[]
null
322
2106.06854
title_snapshot
v139/fumero21a
Learning disentangled representations via product manifold projection
https://proceedings.mlr.press/v139/fumero21a.html
[ "Marco Fumero", "Luca Cosmo", "Simone Melzi", "Emanuele Rodola" ]
null
null
We propose a novel approach to disentangle the generative factors of variation underlying a given set of observations. Our method builds upon the idea that the (unknown) low-dimensional manifold underlying the data space can be explicitly modeled as a product of submanifolds. This definition of disentanglement gives ri...
[]
null
323
2103.01638
title_snapshot
v139/furuta21a
Policy Information Capacity: Information-Theoretic Measure for Task Complexity in Deep Reinforcement Learning
https://proceedings.mlr.press/v139/furuta21a.html
[ "Hiroki Furuta", "Tatsuya Matsushima", "Tadashi Kozuno", "Yutaka Matsuo", "Sergey Levine", "Ofir Nachum", "Shixiang Shane Gu" ]
null
null
Progress in deep reinforcement learning (RL) research is largely enabled by benchmark task environments. However, analyzing the nature of those environments is often overlooked. In particular, we still do not have agreeable ways to measure the difficulty or solvability of a task, given that each has fundamentally diffe...
[]
null
324
2103.12726
title_snapshot
v139/gao21a
An Information-Geometric Distance on the Space of Tasks
https://proceedings.mlr.press/v139/gao21a.html
[ "Yansong Gao", "Pratik Chaudhari" ]
null
null
This paper prescribes a distance between learning tasks modeled as joint distributions on data and labels. Using tools in information geometry, the distance is defined to be the length of the shortest weight trajectory on a Riemannian manifold as a classifier is fitted on an interpolated task. The interpolated task evo...
[]
null
325
2011.00613
title_snapshot
v139/gao21b
Maximum Mean Discrepancy Test is Aware of Adversarial Attacks
https://proceedings.mlr.press/v139/gao21b.html
[ "Ruize Gao", "Feng Liu", "Jingfeng Zhang", "Bo Han", "Tongliang Liu", "Gang Niu", "Masashi Sugiyama" ]
null
null
The maximum mean discrepancy (MMD) test could in principle detect any distributional discrepancy between two datasets. However, it has been shown that the MMD test is unaware of adversarial attacks–the MMD test failed to detect the discrepancy between natural data and adversarial data. Given this phenomenon, we raise a...
[]
null
326
2010.11415
title_snapshot
v139/gao21c
Unsupervised Co-part Segmentation through Assembly
https://proceedings.mlr.press/v139/gao21c.html
[ "Qingzhe Gao", "Bin Wang", "Libin Liu", "Baoquan Chen" ]
null
null
Co-part segmentation is an important problem in computer vision for its rich applications. We propose an unsupervised learning approach for co-part segmentation from images. For the training stage, we leverage motion information embedded in videos and explicitly extract latent representations to segment meaningful obje...
[]
null
327
2106.05897
title_snapshot
v139/gao21d
Discriminative Complementary-Label Learning with Weighted Loss
https://proceedings.mlr.press/v139/gao21d.html
[ "Yi Gao", "Min-Ling Zhang" ]
null
null
Complementary-label learning (CLL) deals with the weak supervision scenario where each training instance is associated with one \emph{complementary} label, which specifies the class label that the instance does \emph{not} belong to. Given the training instance ${\bm x}$, existing CLL approaches aim at modeling the \emp...
[]
null
328
null
null
v139/garg21a
RATT: Leveraging Unlabeled Data to Guarantee Generalization
https://proceedings.mlr.press/v139/garg21a.html
[ "Saurabh Garg", "Sivaraman Balakrishnan", "Zico Kolter", "Zachary Lipton" ]
null
null
To assess generalization, machine learning scientists typically either (i) bound the generalization gap and then (after training) plug in the empirical risk to obtain a bound on the true risk; or (ii) validate empirically on holdout data. However, (i) typically yields vacuous guarantees for overparameterized models; an...
[]
null
329
2105.00303
title_snapshot
v139/garg21b
On Proximal Policy Optimization’s Heavy-tailed Gradients
https://proceedings.mlr.press/v139/garg21b.html
[ "Saurabh Garg", "Joshua Zhanson", "Emilio Parisotto", "Adarsh Prasad", "Zico Kolter", "Zachary Lipton", "Sivaraman Balakrishnan", "Ruslan Salakhutdinov", "Pradeep Ravikumar" ]
null
null
Modern policy gradient algorithms such as Proximal Policy Optimization (PPO) rely on an arsenal of heuristics, including loss clipping and gradient clipping, to ensure successful learning. These heuristics are reminiscent of techniques from robust statistics, commonly used for estimation in outlier-rich ("heavy-tailed"...
[]
null
330
2102.10264
title_snapshot
v139/garreau21a
What does LIME really see in images?
https://proceedings.mlr.press/v139/garreau21a.html
[ "Damien Garreau", "Dina Mardaoui" ]
null
null
The performance of modern algorithms on certain computer vision tasks such as object recognition is now close to that of humans. This success was achieved at the price of complicated architectures depending on millions of parameters and it has become quite challenging to understand how particular predictions are made. ...
[]
null
331
2102.06307
title_snapshot
v139/gauthier21a
Parametric Graph for Unimodal Ranking Bandit
https://proceedings.mlr.press/v139/gauthier21a.html
[ "Camille-Sovanneary Gauthier", "Romaric Gaudel", "Elisa Fromont", "Boammani Aser Lompo" ]
null
null
We tackle the online ranking problem of assigning $L$ items to $K$ positions on a web page in order to maximize the number of user clicks. We propose an original algorithm, easy to implement and with strong theoretical guarantees to tackle this problem in the Position-Based Model (PBM) setting, well suited for applicat...
[]
null
332
null
null
v139/geerts21a
Let’s Agree to Degree: Comparing Graph Convolutional Networks in the Message-Passing Framework
https://proceedings.mlr.press/v139/geerts21a.html
[ "Floris Geerts", "Filip Mazowiecki", "Guillermo Perez" ]
null
null
In this paper we cast neural networks defined on graphs as message-passing neural networks (MPNNs) to study the distinguishing power of different classes of such models. We are interested in when certain architectures are able to tell vertices apart based on the feature labels given as input with the graph. We consider...
[]
null
333
2004.02593
title_snapshot
v139/geffner21a
On the difficulty of unbiased alpha divergence minimization
https://proceedings.mlr.press/v139/geffner21a.html
[ "Tomas Geffner", "Justin Domke" ]
null
null
Several approximate inference algorithms have been proposed to minimize an alpha-divergence between an approximating distribution and a target distribution. Many of these algorithms introduce bias, the magnitude of which becomes problematic in high dimensions. Other algorithms are unbiased. These often seem to suffer f...
[]
null
334
2010.09541
title_snapshot
v139/gentzel21a
How and Why to Use Experimental Data to Evaluate Methods for Observational Causal Inference
https://proceedings.mlr.press/v139/gentzel21a.html
[ "Amanda M Gentzel", "Purva Pruthi", "David Jensen" ]
null
null
Methods that infer causal dependence from observational data are central to many areas of science, including medicine, economics, and the social sciences. A variety of theoretical properties of these methods have been proven, but empirical evaluation remains a challenge, largely due to the lack of observational data se...
[]
null
335
2010.03051
title_snapshot
v139/ghalme21a
Strategic Classification in the Dark
https://proceedings.mlr.press/v139/ghalme21a.html
[ "Ganesh Ghalme", "Vineet Nair", "Itay Eilat", "Inbal Talgam-Cohen", "Nir Rosenfeld" ]
null
null
Strategic classification studies the interaction between a classification rule and the strategic agents it governs. Agents respond by manipulating their features, under the assumption that the classifier is known. However, in many real-life scenarios of high-stake classification (e.g., credit scoring), the classifier i...
[]
null
336
2102.11592
title_snapshot
v139/ghasemipour21a
EMaQ: Expected-Max Q-Learning Operator for Simple Yet Effective Offline and Online RL
https://proceedings.mlr.press/v139/ghasemipour21a.html
[ "Seyed Kamyar Seyed Ghasemipour", "Dale Schuurmans", "Shixiang Shane Gu" ]
null
null
Off-policy reinforcement learning (RL) holds the promise of sample-efficient learning of decision-making policies by leveraging past experience. However, in the offline RL setting – where a fixed collection of interactions are provided and no further interactions are allowed – it has been shown that standard off-policy...
[]
null
337
2007.11091
title_snapshot
v139/ghazi21a
Differentially Private Aggregation in the Shuffle Model: Almost Central Accuracy in Almost a Single Message
https://proceedings.mlr.press/v139/ghazi21a.html
[ "Badih Ghazi", "Ravi Kumar", "Pasin Manurangsi", "Rasmus Pagh", "Amer Sinha" ]
null
null
The shuffle model of differential privacy has attracted attention in the literature due to it being a middle ground between the well-studied central and local models. In this work, we study the problem of summing (aggregating) real numbers or integers, a basic primitive in numerous machine learning tasks, in the shuffl...
[]
null
338
2109.13158
title_snapshot
v139/ghuge21a
The Power of Adaptivity for Stochastic Submodular Cover
https://proceedings.mlr.press/v139/ghuge21a.html
[ "Rohan Ghuge", "Anupam Gupta", "Viswanath Nagarajan" ]
null
null
In the stochastic submodular cover problem, the goal is to select a subset of stochastic items of minimum expected cost to cover a submodular function. Solutions in this setting correspond to a sequential decision process that selects items one by one “adaptively” (depending on prior observations). While such adaptive ...
[]
null
339
2106.16115
title_snapshot
v139/gillenwater21a
Differentially Private Quantiles
https://proceedings.mlr.press/v139/gillenwater21a.html
[ "Jennifer Gillenwater", "Matthew Joseph", "Alex Kulesza" ]
null
null
Quantiles are often used for summarizing and understanding data. If that data is sensitive, it may be necessary to compute quantiles in a way that is differentially private, providing theoretical guarantees that the result does not reveal private information. However, when multiple quantiles are needed, existing differ...
[]
null
340
2102.08244
title_snapshot
v139/gluch21a
Query Complexity of Adversarial Attacks
https://proceedings.mlr.press/v139/gluch21a.html
[ "Grzegorz Gluch", "Rüdiger Urbanke" ]
null
null
There are two main attack models considered in the adversarial robustness literature: black-box and white-box. We consider these threat models as two ends of a fine-grained spectrum, indexed by the number of queries the adversary can ask. Using this point of view we investigate how many queries the adversary needs to m...
[]
null
341
2010.01039
title_snapshot
v139/gogianu21a
Spectral Normalisation for Deep Reinforcement Learning: An Optimisation Perspective
https://proceedings.mlr.press/v139/gogianu21a.html
[ "Florin Gogianu", "Tudor Berariu", "Mihaela C Rosca", "Claudia Clopath", "Lucian Busoniu", "Razvan Pascanu" ]
null
null
Most of the recent deep reinforcement learning advances take an RL-centric perspective and focus on refinements of the training objective. We diverge from this view and show we can recover the performance of these developments not by changing the objective, but by regularising the value-function estimator. Constraining...
[]
null
342
2105.05246
title_snapshot
v139/golany21a
12-Lead ECG Reconstruction via Koopman Operators
https://proceedings.mlr.press/v139/golany21a.html
[ "Tomer Golany", "Kira Radinsky", "Daniel Freedman", "Saar Minha" ]
null
null
32% of all global deaths in the world are caused by cardiovascular diseases. Early detection, especially for patients with ischemia or cardiac arrhythmia, is crucial. To reduce the time between symptoms onset and treatment, wearable ECG sensors were developed to allow for the recording of the full 12-lead ECG signal at...
[]
null
343
null
null
v139/gondal21a
Function Contrastive Learning of Transferable Meta-Representations
https://proceedings.mlr.press/v139/gondal21a.html
[ "Muhammad Waleed Gondal", "Shruti Joshi", "Nasim Rahaman", "Stefan Bauer", "Manuel Wuthrich", "Bernhard Schölkopf" ]
null
null
Meta-learning algorithms adapt quickly to new tasks that are drawn from the same task distribution as the training tasks. The mechanism leading to fast adaptation is the conditioning of a downstream predictive model on the inferred representation of the task’s underlying data generative process, or \emph{function}. Thi...
[]
null
344
2010.07093
title_snapshot
v139/gong21a
Active Slices for Sliced Stein Discrepancy
https://proceedings.mlr.press/v139/gong21a.html
[ "Wenbo Gong", "Kaibo Zhang", "Yingzhen Li", "Jose Miguel Hernandez-Lobato" ]
null
null
Sliced Stein discrepancy (SSD) and its kernelized variants have demonstrated promising successes in goodness-of-fit tests and model learning in high dimensions. Despite the theoretical elegance, their empirical performance depends crucially on the search of the optimal slicing directions to discriminate between two dis...
[]
null
345
2102.03159
title_snapshot
v139/gorantla21a
On the Problem of Underranking in Group-Fair Ranking
https://proceedings.mlr.press/v139/gorantla21a.html
[ "Sruthi Gorantla", "Amit Deshpande", "Anand Louis" ]
null
null
Bias in ranking systems, especially among the top ranks, can worsen social and economic inequalities, polarize opinions, and reinforce stereotypes. On the other hand, a bias correction for minority groups can cause more harm if perceived as favoring group-fair outcomes over meritocracy. Most group-fair ranking algorith...
[]
null
346
2010.06986
title_snapshot
v139/gorbunov21a
MARINA: Faster Non-Convex Distributed Learning with Compression
https://proceedings.mlr.press/v139/gorbunov21a.html
[ "Eduard Gorbunov", "Konstantin P. Burlachenko", "Zhize Li", "Peter Richtarik" ]
null
null
We develop and analyze MARINA: a new communication efficient method for non-convex distributed learning over heterogeneous datasets. MARINA employs a novel communication compression strategy based on the compression of gradient differences that is reminiscent of but different from the strategy employed in the DIANA met...
[]
null
347
2102.07845
title_snapshot
v139/gosgens21a
Systematic Analysis of Cluster Similarity Indices: How to Validate Validation Measures
https://proceedings.mlr.press/v139/gosgens21a.html
[ "Martijn M Gösgens", "Alexey Tikhonov", "Liudmila Prokhorenkova" ]
null
null
Many cluster similarity indices are used to evaluate clustering algorithms, and choosing the best one for a particular task remains an open problem. We demonstrate that this problem is crucial: there are many disagreements among the indices, these disagreements do affect which algorithms are preferred in applications, ...
[]
null
348
1911.04773
title_snapshot
v139/goyal21a
Revisiting Point Cloud Shape Classification with a Simple and Effective Baseline
https://proceedings.mlr.press/v139/goyal21a.html
[ "Ankit Goyal", "Hei Law", "Bowei Liu", "Alejandro Newell", "Jia Deng" ]
null
null
Processing point cloud data is an important component of many real-world systems. As such, a wide variety of point-based approaches have been proposed, reporting steady benchmark improvements over time. We study the key ingredients of this progress and uncover two critical results. First, we find that auxiliary factors...
[]
null
349
2106.05304
title_snapshot
v139/graf21a
Dissecting Supervised Contrastive Learning
https://proceedings.mlr.press/v139/graf21a.html
[ "Florian Graf", "Christoph Hofer", "Marc Niethammer", "Roland Kwitt" ]
null
null
Minimizing cross-entropy over the softmax scores of a linear map composed with a high-capacity encoder is arguably the most popular choice for training neural networks on supervised learning tasks. However, recent works show that one can directly optimize the encoder instead, to obtain equally (or even more) discrimina...
[]
null
350
2102.08817
title_snapshot
v139/grathwohl21a
Oops I Took A Gradient: Scalable Sampling for Discrete Distributions
https://proceedings.mlr.press/v139/grathwohl21a.html
[ "Will Grathwohl", "Kevin Swersky", "Milad Hashemi", "David Duvenaud", "Chris Maddison" ]
null
null
We propose a general and scalable approximate sampling strategy for probabilistic models with discrete variables. Our approach uses gradients of the likelihood function with respect to its discrete inputs to propose updates in a Metropolis-Hastings sampler. We show empirically that this approach outperforms generic sam...
[]
null
351
2102.04509
title_snapshot
v139/greenberg21a
Detecting Rewards Deterioration in Episodic Reinforcement Learning
https://proceedings.mlr.press/v139/greenberg21a.html
[ "Ido Greenberg", "Shie Mannor" ]
null
null
In many RL applications, once training ends, it is vital to detect any deterioration in the agent performance as soon as possible. Furthermore, it often has to be done without modifying the policy and under minimal assumptions regarding the environment. In this paper, we address this problem by focusing directly on the...
[]
null
352
2010.11660
title_snapshot
v139/gu21a
Crystallization Learning with the Delaunay Triangulation
https://proceedings.mlr.press/v139/gu21a.html
[ "Jiaqi Gu", "Guosheng Yin" ]
null
null
Based on the Delaunay triangulation, we propose the crystallization learning to estimate the conditional expectation function in the framework of nonparametric regression. By conducting the crystallization search for the Delaunay simplices closest to the target point in a hierarchical way, the crystallization learning ...
[]
null
353
null
null
v139/guan21a
AutoAttend: Automated Attention Representation Search
https://proceedings.mlr.press/v139/guan21a.html
[ "Chaoyu Guan", "Xin Wang", "Wenwu Zhu" ]
null
null
Self-attention mechanisms have been widely adopted in many machine learning areas, including Natural Language Processing (NLP) and Graph Representation Learning (GRL), etc. However, existing works heavily rely on hand-crafted design to obtain customized attention mechanisms. In this paper, we automate Key, Query and Va...
[]
null
354
null
null
v139/gultchin21a
Operationalizing Complex Causes: A Pragmatic View of Mediation
https://proceedings.mlr.press/v139/gultchin21a.html
[ "Limor Gultchin", "David Watson", "Matt Kusner", "Ricardo Silva" ]
null
null
We examine the problem of causal response estimation for complex objects (e.g., text, images, genomics). In this setting, classical \emph{atomic} interventions are often not available (e.g., changes to characters, pixels, DNA base-pairs). Instead, we only have access to indirect or \emph{crude} interventions (e.g., enr...
[]
null
355
2106.05074
title_snapshot
v139/guminov21a
On a Combination of Alternating Minimization and Nesterov’s Momentum
https://proceedings.mlr.press/v139/guminov21a.html
[ "Sergey Guminov", "Pavel Dvurechensky", "Nazarii Tupitsa", "Alexander Gasnikov" ]
null
null
Alternating minimization (AM) procedures are practically efficient in many applications for solving convex and non-convex optimization problems. On the other hand, Nesterov’s accelerated gradient is theoretically optimal first-order method for convex optimization. In this paper we combine AM and Nesterov’s acceleration...
[]
null
356
1906.03622
title_snapshot
v139/guo21a
Decentralized Single-Timescale Actor-Critic on Zero-Sum Two-Player Stochastic Games
https://proceedings.mlr.press/v139/guo21a.html
[ "Hongyi Guo", "Zuyue Fu", "Zhuoran Yang", "Zhaoran Wang" ]
null
null
We study the global convergence and global optimality of the actor-critic algorithm applied for the zero-sum two-player stochastic games in a decentralized manner. We focus on the single-timescale setting where the critic is updated by applying the Bellman operator only once and the actor is updated by policy gradient ...
[]
null
357
null
null
v139/guo21b
Adversarial Policy Learning in Two-player Competitive Games
https://proceedings.mlr.press/v139/guo21b.html
[ "Wenbo Guo", "Xian Wu", "Sui Huang", "Xinyu Xing" ]
null
null
In a two-player deep reinforcement learning task, recent work shows an attacker could learn an adversarial policy that triggers a target agent to perform poorly and even react in an undesired way. However, its efficacy heavily relies upon the zero-sum assumption made in the two-player game. In this work, we propose a n...
[]
null
358
null
null
v139/guo21c
Soft then Hard: Rethinking the Quantization in Neural Image Compression
https://proceedings.mlr.press/v139/guo21c.html
[ "Zongyu Guo", "Zhizheng Zhang", "Runsen Feng", "Zhibo Chen" ]
null
null
Quantization is one of the core components in lossy image compression. For neural image compression, end-to-end optimization requires differentiable approximations of quantization, which can generally be grouped into three categories: additive uniform noise, straight-through estimator and soft-to-hard annealing. Traini...
[]
null
359
2104.05168
title_snapshot
v139/gupta21a
UneVEn: Universal Value Exploration for Multi-Agent Reinforcement Learning
https://proceedings.mlr.press/v139/gupta21a.html
[ "Tarun Gupta", "Anuj Mahajan", "Bei Peng", "Wendelin Boehmer", "Shimon Whiteson" ]
null
null
VDN and QMIX are two popular value-based algorithms for cooperative MARL that learn a centralized action value function as a monotonic mixing of per-agent utilities. While this enables easy decentralization of the learned policy, the restricted joint action value function can prevent them from solving tasks that requir...
[]
null
360
2010.02974
title_snapshot
v139/gupta21b
Distribution-Free Calibration Guarantees for Histogram Binning without Sample Splitting
https://proceedings.mlr.press/v139/gupta21b.html
[ "Chirag Gupta", "Aaditya Ramdas" ]
null
null
We prove calibration guarantees for the popular histogram binning (also called uniform-mass binning) method of Zadrozny and Elkan (2001). Histogram binning has displayed strong practical performance, but theoretical guarantees have only been shown for sample split versions that avoid ’double dipping’ the data. We demon...
[]
null
361
2105.04656
title_snapshot
v139/gupta21c
Correcting Exposure Bias for Link Recommendation
https://proceedings.mlr.press/v139/gupta21c.html
[ "Shantanu Gupta", "Hao Wang", "Zachary Lipton", "Yuyang Wang" ]
null
null
Link prediction methods are frequently applied in recommender systems, e.g., to suggest citations for academic papers or friends in social networks. However, exposure bias can arise when users are systematically underexposed to certain relevant items. For example, in citation networks, authors might be more likely to e...
[]
null
362
2106.07041
title_snapshot
v139/gurbuzbalaban21a
The Heavy-Tail Phenomenon in SGD
https://proceedings.mlr.press/v139/gurbuzbalaban21a.html
[ "Mert Gurbuzbalaban", "Umut Simsekli", "Lingjiong Zhu" ]
null
null
In recent years, various notions of capacity and complexity have been proposed for characterizing the generalization properties of stochastic gradient descent (SGD) in deep learning. Some of the popular notions that correlate well with the performance on unseen data are (i) the ‘flatness’ of the local minimum found by ...
[]
null
363
2006.04740
title_snapshot
v139/gurel21a
Knowledge Enhanced Machine Learning Pipeline against Diverse Adversarial Attacks
https://proceedings.mlr.press/v139/gurel21a.html
[ "Nezihe Merve Gürel", "Xiangyu Qi", "Luka Rimanic", "Ce Zhang", "Bo Li" ]
null
null
Despite the great successes achieved by deep neural networks (DNNs), recent studies show that they are vulnerable against adversarial examples, which aim to mislead DNNs by adding small adversarial perturbations. Several defenses have been proposed against such attacks, while many of them have been adaptively attacked....
[]
null
364
2106.06235
title_snapshot
v139/gyorgy21a
Adapting to Delays and Data in Adversarial Multi-Armed Bandits
https://proceedings.mlr.press/v139/gyorgy21a.html
[ "Andras Gyorgy", "Pooria Joulani" ]
null
null
We consider the adversarial multi-armed bandit problem under delayed feedback. We analyze variants of the Exp3 algorithm that tune their step size using only information (about the losses and delays) available at the time of the decisions, and obtain regret guarantees that adapt to the observed (rather than the worst-c...
[]
null
365
2010.06022
title_snapshot
v139/hafez-kolahi21a
Rate-Distortion Analysis of Minimum Excess Risk in Bayesian Learning
https://proceedings.mlr.press/v139/hafez-kolahi21a.html
[ "Hassan Hafez-Kolahi", "Behrad Moniri", "Shohreh Kasaei", "Mahdieh Soleymani Baghshah" ]
null
null
In parametric Bayesian learning, a prior is assumed on the parameter $W$ which determines the distribution of samples. In this setting, Minimum Excess Risk (MER) is defined as the difference between the minimum expected loss achievable when learning from data and the minimum expected loss that could be achieved if $W$ ...
[]
null
366
2105.04180
title_snapshot
v139/hallak21a
Regret Minimization in Stochastic Non-Convex Learning via a Proximal-Gradient Approach
https://proceedings.mlr.press/v139/hallak21a.html
[ "Nadav Hallak", "Panayotis Mertikopoulos", "Volkan Cevher" ]
null
null
This paper develops a methodology for regret minimization with stochastic first-order oracle feedback in online, constrained, non-smooth, non-convex problems. In this setting, the minimization of external regret is beyond reach for first-order methods, and there are no gradient-based algorithmic frameworks capable of p...
[]
null
367
2010.06250
title_snapshot
v139/han21a
Diversity Actor-Critic: Sample-Aware Entropy Regularization for Sample-Efficient Exploration
https://proceedings.mlr.press/v139/han21a.html
[ "Seungyul Han", "Youngchul Sung" ]
null
null
In this paper, sample-aware policy entropy regularization is proposed to enhance the conventional policy entropy regularization for better exploration. Exploiting the sample distribution obtainable from the replay buffer, the proposed sample-aware entropy regularization maximizes the entropy of the weighted sum of the ...
[]
null
368
2006.01419
title_snapshot
v139/han21b
Adversarial Combinatorial Bandits with General Non-linear Reward Functions
https://proceedings.mlr.press/v139/han21b.html
[ "Yanjun Han", "Yining Wang", "Xi Chen" ]
null
null
In this paper we study the adversarial combinatorial bandit with a known non-linear reward function, extending existing work on adversarial linear combinatorial bandit. {The adversarial combinatorial bandit with general non-linear reward is an important open problem in bandit literature, and it is still unclear whether...
[]
null
369
2101.01301
title_snapshot
v139/hang21a
A Collective Learning Framework to Boost GNN Expressiveness for Node Classification
https://proceedings.mlr.press/v139/hang21a.html
[ "Mengyue Hang", "Jennifer Neville", "Bruno Ribeiro" ]
null
null
Collective Inference (CI) is a procedure designed to boost weak relational classifiers, specially for node classification tasks. Graph Neural Networks (GNNs) are strong classifiers that have been used with great success. Unfortunately, most existing practical GNNs are not most-expressive (universal). Thus, it is an ope...
[]
null
370
2003.12169
title_judge
v139/hanjie21a
Grounding Language to Entities and Dynamics for Generalization in Reinforcement Learning
https://proceedings.mlr.press/v139/hanjie21a.html
[ "Austin W. Hanjie", "Victor Y Zhong", "Karthik Narasimhan" ]
null
null
We investigate the use of natural language to drive the generalization of control policies and introduce the new multi-task environment Messenger with free-form text manuals describing the environment dynamics. Unlike previous work, Messenger does not assume prior knowledge connecting text and state observations {—} th...
[]
null
371
2101.07393
title_snapshot
v139/hao21a
Sparse Feature Selection Makes Batch Reinforcement Learning More Sample Efficient
https://proceedings.mlr.press/v139/hao21a.html
[ "Botao Hao", "Yaqi Duan", "Tor Lattimore", "Csaba Szepesvari", "Mengdi Wang" ]
null
null
This paper provides a statistical analysis of high-dimensional batch reinforcement learning (RL) using sparse linear function approximation. When there is a large number of candidate features, our result sheds light on the fact that sparsity-aware methods can make batch RL more sample efficient. We first consider the o...
[]
null
372
2011.04019
title_snapshot
v139/hao21b
Bootstrapping Fitted Q-Evaluation for Off-Policy Inference
https://proceedings.mlr.press/v139/hao21b.html
[ "Botao Hao", "Xiang Ji", "Yaqi Duan", "Hao Lu", "Csaba Szepesvari", "Mengdi Wang" ]
null
null
Bootstrapping provides a flexible and effective approach for assessing the quality of batch reinforcement learning, yet its theoretical properties are poorly understood. In this paper, we study the use of bootstrapping in off-policy evaluation (OPE), and in particular, we focus on the fitted Q-evaluation (FQE) that is ...
[]
null
373
2102.03607
title_snapshot
v139/hao21c
Compressed Maximum Likelihood
https://proceedings.mlr.press/v139/hao21c.html
[ "Yi Hao", "Alon Orlitsky" ]
null
null
Maximum likelihood (ML) is one of the most fundamental and general statistical estimation techniques. Inspired by recent advances in estimating distribution functionals, we propose $\textit{compressed maximum likelihood}$ (CML) that applies ML to the compressed samples. We then show that CML is sample-efficient for sev...
[]
null
374
null
null
v139/hartford21a
Valid Causal Inference with (Some) Invalid Instruments
https://proceedings.mlr.press/v139/hartford21a.html
[ "Jason S Hartford", "Victor Veitch", "Dhanya Sridhar", "Kevin Leyton-Brown" ]
null
null
Instrumental variable methods provide a powerful approach to estimating causal effects in the presence of unobserved confounding. But a key challenge when applying them is the reliance on untestable "exclusion" assumptions that rule out any relationship between the instrument variable and the response that is not media...
[]
null
375
2006.11386
title_snapshot
v139/hashimoto21a
Model Performance Scaling with Multiple Data Sources
https://proceedings.mlr.press/v139/hashimoto21a.html
[ "Tatsunori Hashimoto" ]
null
null
Real-world machine learning systems are often trained using a mix of data sources with varying cost and quality. Understanding how the size and composition of a training dataset affect model performance is critical for advancing our understanding of generalization, as well as designing more effective data collection po...
[]
null
376
null
null
v139/havtorn21a
Hierarchical VAEs Know What They Don’t Know
https://proceedings.mlr.press/v139/havtorn21a.html
[ "Jakob D. Havtorn", "Jes Frellsen", "Søren Hauberg", "Lars Maaløe" ]
null
null
Deep generative models have been demonstrated as state-of-the-art density estimators. Yet, recent work has found that they often assign a higher likelihood to data from outside the training distribution. This seemingly paradoxical behavior has caused concerns over the quality of the attained density estimates. In the c...
[]
null
377
2102.08248
title_snapshot
v139/hayase21a
SPECTRE: defending against backdoor attacks using robust statistics
https://proceedings.mlr.press/v139/hayase21a.html
[ "Jonathan Hayase", "Weihao Kong", "Raghav Somani", "Sewoong Oh" ]
null
null
Modern machine learning increasingly requires training on a large collection of data from multiple sources, not all of which can be trusted. A particularly frightening scenario is when a small fraction of corrupted data changes the behavior of the trained model when triggered by an attacker-specified watermark. Such a ...
[]
null
378
2104.11315
title_snapshot
v139/hazan21a
Boosting for Online Convex Optimization
https://proceedings.mlr.press/v139/hazan21a.html
[ "Elad Hazan", "Karan Singh" ]
null
null
We consider the decision-making framework of online convex optimization with a very large number of experts. This setting is ubiquitous in contextual and reinforcement learning problems, where the size of the policy class renders enumeration and search within the policy class infeasible. Instead, we consider generalizi...
[]
null
379
2102.09305
title_snapshot
v139/he21a
PipeTransformer: Automated Elastic Pipelining for Distributed Training of Large-scale Models
https://proceedings.mlr.press/v139/he21a.html
[ "Chaoyang He", "Shen Li", "Mahdi Soltanolkotabi", "Salman Avestimehr" ]
null
null
The size of Transformer models is growing at an unprecedented rate. It has taken less than one year to reach trillion-level parameters since the release of GPT-3 (175B). Training such models requires both substantial engineering efforts and enormous computing resources, which are luxuries most research teams cannot aff...
[]
null
380
2102.03161
title_judge
v139/he21b
SoundDet: Polyphonic Moving Sound Event Detection and Localization from Raw Waveform
https://proceedings.mlr.press/v139/he21b.html
[ "Yuhang He", "Niki Trigoni", "Andrew Markham" ]
null
null
We present a new framework SoundDet, which is an end-to-end trainable and light-weight framework, for polyphonic moving sound event detection and localization. Prior methods typically approach this problem by preprocessing raw waveform into time-frequency representations, which is more amenable to process with well-est...
[]
null
381
2106.06969
title_snapshot
v139/he21c
Logarithmic Regret for Reinforcement Learning with Linear Function Approximation
https://proceedings.mlr.press/v139/he21c.html
[ "Jiafan He", "Dongruo Zhou", "Quanquan Gu" ]
null
null
Reinforcement learning (RL) with linear function approximation has received increasing attention recently. However, existing work has focused on obtaining $\sqrt{T}$-type regret bound, where $T$ is the number of interactions with the MDP. In this paper, we show that logarithmic regret is attainable under two recently p...
[]
null
382
2011.11566
title_snapshot
v139/heidari21a
Finding Relevant Information via a Discrete Fourier Expansion
https://proceedings.mlr.press/v139/heidari21a.html
[ "Mohsen Heidari", "Jithin Sreedharan", "Gil I Shamir", "Wojciech Szpankowski" ]
null
null
A fundamental obstacle in learning information from data is the presence of nonlinear redundancies and dependencies in it. To address this, we propose a Fourier-based approach to extract relevant information in the supervised setting. We first develop a novel Fourier expansion for functions of correlated binary random ...
[]
null
383
null
null
v139/heliou21a
Zeroth-Order Non-Convex Learning via Hierarchical Dual Averaging
https://proceedings.mlr.press/v139/heliou21a.html
[ "Amélie Héliou", "Matthieu Martin", "Panayotis Mertikopoulos", "Thibaud Rahier" ]
null
null
We propose a hierarchical version of dual averaging for zeroth-order online non-convex optimization {–} i.e., learning processes where, at each stage, the optimizer is facing an unknown non-convex loss function and only receives the incurred loss as feedback. The proposed class of policies relies on the construction of...
[]
null
384
2109.05829
title_snapshot
v139/henderson21a
Improving Molecular Graph Neural Network Explainability with Orthonormalization and Induced Sparsity
https://proceedings.mlr.press/v139/henderson21a.html
[ "Ryan Henderson", "Djork-Arné Clevert", "Floriane Montanari" ]
null
null
Rationalizing which parts of a molecule drive the predictions of a molecular graph convolutional neural network (GCNN) can be difficult. To help, we propose two simple regularization techniques to apply during the training of GCNNs: Batch Representation Orthonormalization (BRO) and Gini regularization. BRO, inspired by...
[]
null
385
2105.04854
title_snapshot
v139/hessel21a
Muesli: Combining Improvements in Policy Optimization
https://proceedings.mlr.press/v139/hessel21a.html
[ "Matteo Hessel", "Ivo Danihelka", "Fabio Viola", "Arthur Guez", "Simon Schmitt", "Laurent Sifre", "Theophane Weber", "David Silver", "Hado Van Hasselt" ]
null
null
We propose a novel policy update that combines regularized policy optimization with model learning as an auxiliary loss. The update (henceforth Muesli) matches MuZero’s state-of-the-art performance on Atari. Notably, Muesli does so without using deep search: it acts directly with a policy network and has computation sp...
[]
null
386
2104.06159
title_snapshot
v139/hilgard21a
Learning Representations by Humans, for Humans
https://proceedings.mlr.press/v139/hilgard21a.html
[ "Sophie Hilgard", "Nir Rosenfeld", "Mahzarin R Banaji", "Jack Cao", "David Parkes" ]
null
null
When machine predictors can achieve higher performance than the human decision-makers they support, improving the performance of human decision-makers is often conflated with improving machine accuracy. Here we propose a framework to directly support human decision-making, in which the role of machines is to reframe pr...
[]
null
387
1905.12686
title_snapshot
v139/hiranandani21a
Optimizing Black-box Metrics with Iterative Example Weighting
https://proceedings.mlr.press/v139/hiranandani21a.html
[ "Gaurush Hiranandani", "Jatin Mathur", "Harikrishna Narasimhan", "Mahdi Milani Fard", "Sanmi Koyejo" ]
null
null
We consider learning to optimize a classification metric defined by a black-box function of the confusion matrix. Such black-box learning settings are ubiquitous, for example, when the learner only has query access to the metric of interest, or in noisy-label and domain adaptation applications where the learner must ev...
[]
null
388
2102.09492
title_snapshot
v139/hirsch21a
Trees with Attention for Set Prediction Tasks
https://proceedings.mlr.press/v139/hirsch21a.html
[ "Roy Hirsch", "Ran Gilad-Bachrach" ]
null
null
In many machine learning applications, each record represents a set of items. For example, when making predictions from medical records, the medications prescribed to a patient are a set whose size is not fixed and whose order is arbitrary. However, most machine learning algorithms are not designed to handle set struct...
[]
null
389
null
null
v139/hodgkinson21a
Multiplicative Noise and Heavy Tails in Stochastic Optimization
https://proceedings.mlr.press/v139/hodgkinson21a.html
[ "Liam Hodgkinson", "Michael Mahoney" ]
null
null
Although stochastic optimization is central to modern machine learning, the precise mechanisms underlying its success, and in particular, the precise role of the stochasticity, still remain unclear. Modeling stochastic optimization algorithms as discrete random recurrence relations, we show that multiplicative noise, a...
[]
null
390
2006.06293
title_snapshot
v139/hoedt21a
MC-LSTM: Mass-Conserving LSTM
https://proceedings.mlr.press/v139/hoedt21a.html
[ "Pieter-Jan Hoedt", "Frederik Kratzert", "Daniel Klotz", "Christina Halmich", "Markus Holzleitner", "Grey S Nearing", "Sepp Hochreiter", "Guenter Klambauer" ]
null
null
The success of Convolutional Neural Networks (CNNs) in computer vision is mainly driven by their strong inductive bias, which is strong enough to allow CNNs to solve vision-related tasks with random weights, meaning without learning. Similarly, Long Short-Term Memory (LSTM) has a strong inductive bias towards storing i...
[]
null
391
2101.05186
title_snapshot
v139/hoiem21a
Learning Curves for Analysis of Deep Networks
https://proceedings.mlr.press/v139/hoiem21a.html
[ "Derek Hoiem", "Tanmay Gupta", "Zhizhong Li", "Michal Shlapentokh-Rothman" ]
null
null
Learning curves model a classifier’s test error as a function of the number of training samples. Prior works show that learning curves can be used to select model parameters and extrapolate performance. We investigate how to use learning curves to evaluate design choices, such as pretraining, architecture, and data aug...
[]
null
392
2010.11029
title_snapshot
v139/holderrieth21a
Equivariant Learning of Stochastic Fields: Gaussian Processes and Steerable Conditional Neural Processes
https://proceedings.mlr.press/v139/holderrieth21a.html
[ "Peter Holderrieth", "Michael J Hutchinson", "Yee Whye Teh" ]
null
null
Motivated by objects such as electric fields or fluid streams, we study the problem of learning stochastic fields, i.e. stochastic processes whose samples are fields like those occurring in physics and engineering. Considering general transformations such as rotations and reflections, we show that spatial invariance of...
[]
null
393
2011.12916
title_snapshot
v139/hong21a
Latent Programmer: Discrete Latent Codes for Program Synthesis
https://proceedings.mlr.press/v139/hong21a.html
[ "Joey Hong", "David Dohan", "Rishabh Singh", "Charles Sutton", "Manzil Zaheer" ]
null
null
A key problem in program synthesis is searching over the large space of possible programs. Human programmers might decide the high-level structure of the desired program before thinking about the details; motivated by this intuition, we consider two-level search for program synthesis, in which the synthesizer first gen...
[]
null
394
2012.00377
title_snapshot
v139/hong21b
Chebyshev Polynomial Codes: Task Entanglement-based Coding for Distributed Matrix Multiplication
https://proceedings.mlr.press/v139/hong21b.html
[ "Sangwoo Hong", "Heecheol Yang", "Youngseok Yoon", "Taehyun Cho", "Jungwoo Lee" ]
null
null
Distributed computing has been a prominent solution to efficiently process massive datasets in parallel. However, the existence of stragglers is one of the major concerns that slows down the overall speed of distributed computing. To deal with this problem, we consider a distributed matrix multiplication scenario where...
[]
null
395
null
null
v139/hosseini21a
Federated Learning of User Verification Models Without Sharing Embeddings
https://proceedings.mlr.press/v139/hosseini21a.html
[ "Hossein Hosseini", "Hyunsin Park", "Sungrack Yun", "Christos Louizos", "Joseph Soriaga", "Max Welling" ]
null
null
We consider the problem of training User Verification (UV) models in federated setup, where each user has access to the data of only one class and user embeddings cannot be shared with the server or other users. To address this problem, we propose Federated User Verification (FedUV), a framework in which users jointly ...
[]
null
396
2104.08776
title_snapshot
v139/hsieh21a
The Limits of Min-Max Optimization Algorithms: Convergence to Spurious Non-Critical Sets
https://proceedings.mlr.press/v139/hsieh21a.html
[ "Ya-Ping Hsieh", "Panayotis Mertikopoulos", "Volkan Cevher" ]
null
null
Compared to minimization, the min-max optimization in machine learning applications is considerably more convoluted because of the existence of cycles and similar phenomena. Such oscillatory behaviors are well-understood in the convex-concave regime, and many algorithms are known to overcome them. In this paper, we go ...
[]
null
397
2006.09065
title_snapshot
v139/hu21a
Near-Optimal Representation Learning for Linear Bandits and Linear RL
https://proceedings.mlr.press/v139/hu21a.html
[ "Jiachen Hu", "Xiaoyu Chen", "Chi Jin", "Lihong Li", "Liwei Wang" ]
null
null
This paper studies representation learning for multi-task linear bandits and multi-task episodic RL with linear value function approximation. We first consider the setting where we play $M$ linear bandits with dimension $d$ concurrently, and these bandits share a common $k$-dimensional linear representation so that $k\...
[]
null
398
2102.04132
title_snapshot
v139/hu21b
On the Random Conjugate Kernel and Neural Tangent Kernel
https://proceedings.mlr.press/v139/hu21b.html
[ "Zhengmian Hu", "Heng Huang" ]
null
null
We investigate the distributions of Conjugate Kernel (CK) and Neural Tangent Kernel (NTK) for ReLU networks with random initialization. We derive the precise distributions and moments of the diagonal elements of these kernels. For a feedforward network, these values converge in law to a log-normal distribution when the...
[]
null
399
null
null
v139/hu21c
Off-Belief Learning
https://proceedings.mlr.press/v139/hu21c.html
[ "Hengyuan Hu", "Adam Lerer", "Brandon Cui", "Luis Pineda", "Noam Brown", "Jakob Foerster" ]
null
null
The standard problem setting in Dec-POMDPs is self-play, where the goal is to find a set of policies that play optimally together. Policies learned through self-play may adopt arbitrary conventions and implicitly rely on multi-step reasoning based on fragile assumptions about other agents’ actions and thus fail when pa...
[]
null
400
2103.04000
title_snapshot