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/sukhbaatar21a
Not All Memories are Created Equal: Learning to Forget by Expiring
https://proceedings.mlr.press/v139/sukhbaatar21a.html
[ "Sainbayar Sukhbaatar", "Da Ju", "Spencer Poff", "Stephen Roller", "Arthur Szlam", "Jason Weston", "Angela Fan" ]
null
null
Attention mechanisms have shown promising results in sequence modeling tasks that require long-term memory. Recent work investigated mechanisms to reduce the computational cost of preserving and storing memories. However, not all content in the past is equally important to remember. We propose Expire-Span, a method tha...
[]
null
901
2105.06548
title_snapshot
v139/summers21a
Nondeterminism and Instability in Neural Network Optimization
https://proceedings.mlr.press/v139/summers21a.html
[ "Cecilia Summers", "Michael J. Dinneen" ]
null
null
Nondeterminism in neural network optimization produces uncertainty in performance, making small improvements difficult to discern from run-to-run variability. While uncertainty can be reduced by training multiple model copies, doing so is time-consuming, costly, and harms reproducibility. In this work, we establish an ...
[]
null
902
2103.04514
title_snapshot
v139/sun21a
AutoSampling: Search for Effective Data Sampling Schedules
https://proceedings.mlr.press/v139/sun21a.html
[ "Ming Sun", "Haoxuan Dou", "Baopu Li", "Junjie Yan", "Wanli Ouyang", "Lei Cui" ]
null
null
Data sampling acts as a pivotal role in training deep learning models. However, an effective sampling schedule is difficult to learn due to its inherent high-dimension as a hyper-parameter. In this paper, we propose an AutoSampling method to automatically learn sampling schedules for model training, which consists of t...
[]
null
903
2105.13695
title_snapshot
v139/sun21b
What Makes for End-to-End Object Detection?
https://proceedings.mlr.press/v139/sun21b.html
[ "Peize Sun", "Yi Jiang", "Enze Xie", "Wenqi Shao", "Zehuan Yuan", "Changhu Wang", "Ping Luo" ]
null
null
Object detection has recently achieved a breakthrough for removing the last one non-differentiable component in the pipeline, Non-Maximum Suppression (NMS), and building up an end-to-end system. However, what makes for its one-to-one prediction has not been well understood. In this paper, we first point out that one-to...
[]
null
904
2012.05780
title_snapshot
v139/sun21c
DFAC Framework: Factorizing the Value Function via Quantile Mixture for Multi-Agent Distributional Q-Learning
https://proceedings.mlr.press/v139/sun21c.html
[ "Wei-Fang Sun", "Cheng-Kuang Lee", "Chun-Yi Lee" ]
null
null
In fully cooperative multi-agent reinforcement learning (MARL) settings, the environments are highly stochastic due to the partial observability of each agent and the continuously changing policies of the other agents. To address the above issues, we integrate distributional RL and value function factorization methods ...
[]
null
905
2102.07936
title_snapshot
v139/sun21d
Scalable Variational Gaussian Processes via Harmonic Kernel Decomposition
https://proceedings.mlr.press/v139/sun21d.html
[ "Shengyang Sun", "Jiaxin Shi", "Andrew Gordon Gordon Wilson", "Roger B Grosse" ]
null
null
We introduce a new scalable variational Gaussian process approximation which provides a high fidelity approximation while retaining general applicability. We propose the harmonic kernel decomposition (HKD), which uses Fourier series to decompose a kernel as a sum of orthogonal kernels. Our variational approximation exp...
[]
null
906
2106.05992
title_snapshot
v139/sun21e
Reasoning Over Virtual Knowledge Bases With Open Predicate Relations
https://proceedings.mlr.press/v139/sun21e.html
[ "Haitian Sun", "Patrick Verga", "Bhuwan Dhingra", "Ruslan Salakhutdinov", "William W Cohen" ]
null
null
We present the Open Predicate Query Language (OPQL); a method for constructing a virtual KB (VKB) trained entirely from text. Large Knowledge Bases (KBs) are indispensable for a wide-range of industry applications such as question answering and recommendation. Typically, KBs encode world knowledge in a structured, read...
[]
null
907
2102.07043
title_snapshot
v139/sundaram21a
PAC-Learning for Strategic Classification
https://proceedings.mlr.press/v139/sundaram21a.html
[ "Ravi Sundaram", "Anil Vullikanti", "Haifeng Xu", "Fan Yao" ]
null
null
The study of strategic or adversarial manipulation of testing data to fool a classifier has attracted much recent attention. Most previous works have focused on two extreme situations where any testing data point either is completely adversarial or always equally prefers the positive label. In this paper, we generalize...
[]
null
908
2012.03310
title_snapshot
v139/suttle21a
Reinforcement Learning for Cost-Aware Markov Decision Processes
https://proceedings.mlr.press/v139/suttle21a.html
[ "Wesley Suttle", "Kaiqing Zhang", "Zhuoran Yang", "Ji Liu", "David Kraemer" ]
null
null
Ratio maximization has applications in areas as diverse as finance, reward shaping for reinforcement learning (RL), and the development of safe artificial intelligence, yet there has been very little exploration of RL algorithms for ratio maximization. This paper addresses this deficiency by introducing two new, model-...
[]
null
909
null
null
v139/suya21a
Model-Targeted Poisoning Attacks with Provable Convergence
https://proceedings.mlr.press/v139/suya21a.html
[ "Fnu Suya", "Saeed Mahloujifar", "Anshuman Suri", "David Evans", "Yuan Tian" ]
null
null
In a poisoning attack, an adversary who controls a small fraction of the training data attempts to select that data, so a model is induced that misbehaves in a particular way. We consider poisoning attacks against convex machine learning models and propose an efficient poisoning attack designed to induce a model specif...
[]
null
910
2006.16469
title_snapshot
v139/suzuki21a
Generalization Error Bound for Hyperbolic Ordinal Embedding
https://proceedings.mlr.press/v139/suzuki21a.html
[ "Atsushi Suzuki", "Atsushi Nitanda", "Jing Wang", "Linchuan Xu", "Kenji Yamanishi", "Marc Cavazza" ]
null
null
Hyperbolic ordinal embedding (HOE) represents entities as points in hyperbolic space so that they agree as well as possible with given constraints in the form of entity $i$ is more similar to entity $j$ than to entity $k$. It has been experimentally shown that HOE can obtain representations of hierarchical data such as...
[]
null
911
2105.10475
title_snapshot
v139/swamy21a
Of Moments and Matching: A Game-Theoretic Framework for Closing the Imitation Gap
https://proceedings.mlr.press/v139/swamy21a.html
[ "Gokul Swamy", "Sanjiban Choudhury", "J. Andrew Bagnell", "Steven Wu" ]
null
null
We provide a unifying view of a large family of previous imitation learning algorithms through the lens of moment matching. At its core, our classification scheme is based on whether the learner attempts to match (1) reward or (2) action-value moments of the expert’s behavior, with each option leading to differing algo...
[]
null
912
2103.03236
title_snapshot
v139/syed21a
Parallel tempering on optimized paths
https://proceedings.mlr.press/v139/syed21a.html
[ "Saifuddin Syed", "Vittorio Romaniello", "Trevor Campbell", "Alexandre Bouchard-Cote" ]
null
null
Parallel tempering (PT) is a class of Markov chain Monte Carlo algorithms that constructs a path of distributions annealing between a tractable reference and an intractable target, and then interchanges states along the path to improve mixing in the target. The performance of PT depends on how quickly a sample from the...
[]
null
913
2102.07720
title_snapshot
v139/taghanaki21a
Robust Representation Learning via Perceptual Similarity Metrics
https://proceedings.mlr.press/v139/taghanaki21a.html
[ "Saeid A Taghanaki", "Kristy Choi", "Amir Hosein Khasahmadi", "Anirudh Goyal" ]
null
null
A fundamental challenge in artificial intelligence is learning useful representations of data that yield good performance on a downstream classification task, without overfitting to spurious input features. Extracting such task-relevant predictive information becomes particularly difficult for noisy and high-dimensiona...
[]
null
914
2106.06620
title_snapshot
v139/tahmasbi21a
DriftSurf: Stable-State / Reactive-State Learning under Concept Drift
https://proceedings.mlr.press/v139/tahmasbi21a.html
[ "Ashraf Tahmasbi", "Ellango Jothimurugesan", "Srikanta Tirthapura", "Phillip B Gibbons" ]
null
null
When learning from streaming data, a change in the data distribution, also known as concept drift, can render a previously-learned model inaccurate and require training a new model. We present an adaptive learning algorithm that extends previous drift-detection-based methods by incorporating drift detection into a broa...
[]
null
915
null
null
v139/tai21a
Sinkhorn Label Allocation: Semi-Supervised Classification via Annealed Self-Training
https://proceedings.mlr.press/v139/tai21a.html
[ "Kai Sheng Tai", "Peter D Bailis", "Gregory Valiant" ]
null
null
Self-training is a standard approach to semi-supervised learning where the learner’s own predictions on unlabeled data are used as supervision during training. In this paper, we reinterpret this label assignment process as an optimal transportation problem between examples and classes, wherein the cost of assigning an ...
[]
null
916
2102.08622
title_snapshot
v139/takemori21a
Approximation Theory Based Methods for RKHS Bandits
https://proceedings.mlr.press/v139/takemori21a.html
[ "Sho Takemori", "Masahiro Sato" ]
null
null
The RKHS bandit problem (also called kernelized multi-armed bandit problem) is an online optimization problem of non-linear functions with noisy feedback. Although the problem has been extensively studied, there are unsatisfactory results for some problems compared to the well-studied linear bandit case. Specifically, ...
[]
null
917
2010.12167
title_snapshot
v139/takezawa21a
Supervised Tree-Wasserstein Distance
https://proceedings.mlr.press/v139/takezawa21a.html
[ "Yuki Takezawa", "Ryoma Sato", "Makoto Yamada" ]
null
null
To measure the similarity of documents, the Wasserstein distance is a powerful tool, but it requires a high computational cost. Recently, for fast computation of the Wasserstein distance, methods for approximating the Wasserstein distance using a tree metric have been proposed. These tree-based methods allow fast compa...
[]
null
918
2101.11520
title_snapshot
v139/tan21a
EfficientNetV2: Smaller Models and Faster Training
https://proceedings.mlr.press/v139/tan21a.html
[ "Mingxing Tan", "Quoc Le" ]
null
null
This paper introduces EfficientNetV2, a new family of convolutional networks that have faster training speed and better parameter efficiency than previous models. To develop these models, we use a combination of training-aware neural architecture search and scaling, to jointly optimize training speed and parameter effi...
[]
null
919
2104.00298
title_snapshot
v139/tandon21a
SGA: A Robust Algorithm for Partial Recovery of Tree-Structured Graphical Models with Noisy Samples
https://proceedings.mlr.press/v139/tandon21a.html
[ "Anshoo Tandon", "Aldric Han", "Vincent Tan" ]
null
null
We consider learning Ising tree models when the observations from the nodes are corrupted by independent but non-identically distributed noise with unknown statistics. Katiyar et al. (2020) showed that although the exact tree structure cannot be recovered, one can recover a partial tree structure; that is, a structure ...
[]
null
920
2101.08917
title_snapshot
v139/tang21a
1-bit Adam: Communication Efficient Large-Scale Training with Adam’s Convergence Speed
https://proceedings.mlr.press/v139/tang21a.html
[ "Hanlin Tang", "Shaoduo Gan", "Ammar Ahmad Awan", "Samyam Rajbhandari", "Conglong Li", "Xiangru Lian", "Ji Liu", "Ce Zhang", "Yuxiong He" ]
null
null
Scalable training of large models (like BERT and GPT-3) requires careful optimization rooted in model design, architecture, and system capabilities. From a system standpoint, communication has become a major bottleneck, especially on commodity systems with standard TCP interconnects that offer limited network bandwidth...
[]
null
921
2102.02888
title_snapshot
v139/tang21b
Taylor Expansion of Discount Factors
https://proceedings.mlr.press/v139/tang21b.html
[ "Yunhao Tang", "Mark Rowland", "Remi Munos", "Michal Valko" ]
null
null
In practical reinforcement learning (RL), the discount factor used for estimating value functions often differs from that used for defining the evaluation objective. In this work, we study the effect that this discrepancy of discount factors has during learning, and discover a family of objectives that interpolate valu...
[]
null
922
2106.06170
title_snapshot
v139/tao21a
REPAINT: Knowledge Transfer in Deep Reinforcement Learning
https://proceedings.mlr.press/v139/tao21a.html
[ "Yunzhe Tao", "Sahika Genc", "Jonathan Chung", "Tao Sun", "Sunil Mallya" ]
null
null
Accelerating learning processes for complex tasks by leveraging previously learned tasks has been one of the most challenging problems in reinforcement learning, especially when the similarity between source and target tasks is low. This work proposes REPresentation And INstance Transfer (REPAINT) algorithm for knowled...
[]
null
923
2011.11827
title_snapshot
v139/tarmoun21a
Understanding the Dynamics of Gradient Flow in Overparameterized Linear models
https://proceedings.mlr.press/v139/tarmoun21a.html
[ "Salma Tarmoun", "Guilherme Franca", "Benjamin D Haeffele", "Rene Vidal" ]
null
null
We provide a detailed analysis of the dynamics ofthe gradient flow in overparameterized two-layerlinear models. A particularly interesting featureof this model is that its nonlinear dynamics can beexactly solved as a consequence of a large num-ber of conservation laws that constrain the systemto follow particular traje...
[]
null
924
null
null
v139/taskesen21a
Sequential Domain Adaptation by Synthesizing Distributionally Robust Experts
https://proceedings.mlr.press/v139/taskesen21a.html
[ "Bahar Taskesen", "Man-Chung Yue", "Jose Blanchet", "Daniel Kuhn", "Viet Anh Nguyen" ]
null
null
Least squares estimators, when trained on few target domain samples, may predict poorly. Supervised domain adaptation aims to improve the predictive accuracy by exploiting additional labeled training samples from a source distribution that is close to the target distribution. Given available data, we investigate novel ...
[]
null
925
2106.00322
title_snapshot
v139/tavares21a
A Language for Counterfactual Generative Models
https://proceedings.mlr.press/v139/tavares21a.html
[ "Zenna Tavares", "James Koppel", "Xin Zhang", "Ria Das", "Armando Solar-Lezama" ]
null
null
We present Omega, a probabilistic programming language with support for counterfactual inference. Counterfactual inference means to observe some fact in the present, and infer what would have happened had some past intervention been taken, e.g. “given that medication was not effective at dose x, what is the probability...
[]
null
926
null
null
v139/tay21a
Synthesizer: Rethinking Self-Attention for Transformer Models
https://proceedings.mlr.press/v139/tay21a.html
[ "Yi Tay", "Dara Bahri", "Donald Metzler", "Da-Cheng Juan", "Zhe Zhao", "Che Zheng" ]
null
null
The dot product self-attention is known to be central and indispensable to state-of-the-art Transformer models. But is it really required? This paper investigates the true importance and contribution of the dot product-based self-attention mechanism on the performance of Transformer models. Via extensive experiments, w...
[]
null
927
2005.00743
title_judge
v139/tay21b
OmniNet: Omnidirectional Representations from Transformers
https://proceedings.mlr.press/v139/tay21b.html
[ "Yi Tay", "Mostafa Dehghani", "Vamsi Aribandi", "Jai Gupta", "Philip M Pham", "Zhen Qin", "Dara Bahri", "Da-Cheng Juan", "Donald Metzler" ]
null
null
This paper proposes Omnidirectional Representations from Transformers (OMNINET). In OmniNet, instead of maintaining a strictly horizon-tal receptive field, each token is allowed to attend to all tokens in the entire network. This process can also be interpreted as a form of extreme or intensive attention mechanism that...
[]
null
928
2103.01075
title_snapshot
v139/teng21a
T-SCI: A Two-Stage Conformal Inference Algorithm with Guaranteed Coverage for Cox-MLP
https://proceedings.mlr.press/v139/teng21a.html
[ "Jiaye Teng", "Zeren Tan", "Yang Yuan" ]
null
null
It is challenging to deal with censored data, where we only have access to the incomplete information of survival time instead of its exact value. Fortunately, under linear predictor assumption, people can obtain guaranteed coverage for the confidence interval of survival time using methods like Cox Regression. However...
[]
null
929
2103.04556
title_snapshot
v139/terjek21a
Moreau-Yosida $f$-divergences
https://proceedings.mlr.press/v139/terjek21a.html
[ "Dávid Terjék" ]
null
null
Variational representations of $f$-divergences are central to many machine learning algorithms, with Lipschitz constrained variants recently gaining attention. Inspired by this, we define the Moreau-Yosida approximation of $f$-divergences with respect to the Wasserstein-$1$ metric. The corresponding variational formula...
[]
null
930
2102.13416
title_snapshot
v139/teterwak21a
Understanding Invariance via Feedforward Inversion of Discriminatively Trained Classifiers
https://proceedings.mlr.press/v139/teterwak21a.html
[ "Piotr Teterwak", "Chiyuan Zhang", "Dilip Krishnan", "Michael C Mozer" ]
null
null
A discriminatively trained neural net classifier can fit the training data perfectly if all information about its input other than class membership has been discarded prior to the output layer. Surprisingly, past research has discovered that some extraneous visual detail remains in the unnormalized logits. This finding...
[]
null
931
2103.07470
title_snapshot
v139/thananjeyan21a
Resource Allocation in Multi-armed Bandit Exploration: Overcoming Sublinear Scaling with Adaptive Parallelism
https://proceedings.mlr.press/v139/thananjeyan21a.html
[ "Brijen Thananjeyan", "Kirthevasan Kandasamy", "Ion Stoica", "Michael Jordan", "Ken Goldberg", "Joseph Gonzalez" ]
null
null
We study exploration in stochastic multi-armed bandits when we have access to a divisible resource that can be allocated in varying amounts to arm pulls. We focus in particular on the allocation of distributed computing resources, where we may obtain results faster by allocating more resources per pull, but might have ...
[]
null
932
2011.00330
title_snapshot
v139/thin21a
Monte Carlo Variational Auto-Encoders
https://proceedings.mlr.press/v139/thin21a.html
[ "Achille Thin", "Nikita Kotelevskii", "Arnaud Doucet", "Alain Durmus", "Eric Moulines", "Maxim Panov" ]
null
null
Variational auto-encoders (VAE) are popular deep latent variable models which are trained by maximizing an Evidence Lower Bound (ELBO). To obtain tighter ELBO and hence better variational approximations, it has been proposed to use importance sampling to get a lower variance estimate of the evidence. However, importanc...
[]
null
933
2106.15921
title_snapshot
v139/thygesen21a
Efficient Generative Modelling of Protein Structure Fragments using a Deep Markov Model
https://proceedings.mlr.press/v139/thygesen21a.html
[ "Christian B Thygesen", "Christian Skjødt Steenmans", "Ahmad Salim Al-Sibahi", "Lys Sanz Moreta", "Anders Bundgård Sørensen", "Thomas Hamelryck" ]
null
null
Fragment libraries are often used in protein structure prediction, simulation and design as a means to significantly reduce the vast conformational search space. Current state-of-the-art methods for fragment library generation do not properly account for aleatory and epistemic uncertainty, respectively due to the dynam...
[]
null
934
null
null
v139/tian21a
Understanding self-supervised learning dynamics without contrastive pairs
https://proceedings.mlr.press/v139/tian21a.html
[ "Yuandong Tian", "Xinlei Chen", "Surya Ganguli" ]
null
null
While contrastive approaches of self-supervised learning (SSL) learn representations by minimizing the distance between two augmented views of the same data point (positive pairs) and maximizing views from different data points (negative pairs), recent \emph{non-contrastive} SSL (e.g., BYOL and SimSiam) show remarkable...
[]
null
935
2102.06810
title_snapshot
v139/tian21b
Online Learning in Unknown Markov Games
https://proceedings.mlr.press/v139/tian21b.html
[ "Yi Tian", "Yuanhao Wang", "Tiancheng Yu", "Suvrit Sra" ]
null
null
We study online learning in unknown Markov games, a problem that arises in episodic multi-agent reinforcement learning where the actions of the opponents are unobservable. We show that in this challenging setting, achieving sublinear regret against the best response in hindsight is statistically hard. We then consider ...
[]
null
936
2010.15020
title_snapshot
v139/tiao21a
BORE: Bayesian Optimization by Density-Ratio Estimation
https://proceedings.mlr.press/v139/tiao21a.html
[ "Louis C Tiao", "Aaron Klein", "Matthias W Seeger", "Edwin V. Bonilla", "Cedric Archambeau", "Fabio Ramos" ]
null
null
Bayesian optimization (BO) is among the most effective and widely-used blackbox optimization methods. BO proposes solutions according to an explore-exploit trade-off criterion encoded in an acquisition function, many of which are computed from the posterior predictive of a probabilistic surrogate model. Prevalent among...
[]
null
937
2102.09009
title_snapshot
v139/tillinghast21a
Nonparametric Decomposition of Sparse Tensors
https://proceedings.mlr.press/v139/tillinghast21a.html
[ "Conor Tillinghast", "Shandian Zhe" ]
null
null
Tensor decomposition is a powerful framework for multiway data analysis. Despite the success of existing approaches, they ignore the sparse nature of the tensor data in many real-world applications, explicitly or implicitly assuming dense tensors. To address this model misspecification and to exploit the sparse tensor ...
[]
null
938
null
null
v139/tolpin21a
Probabilistic Programs with Stochastic Conditioning
https://proceedings.mlr.press/v139/tolpin21a.html
[ "David Tolpin", "Yuan Zhou", "Tom Rainforth", "Hongseok Yang" ]
null
null
We tackle the problem of conditioning probabilistic programs on distributions of observable variables. Probabilistic programs are usually conditioned on samples from the joint data distribution, which we refer to as deterministic conditioning. However, in many real-life scenarios, the observations are given as marginal...
[]
null
939
2010.00282
title_snapshot
v139/tomen21a
Deep Continuous Networks
https://proceedings.mlr.press/v139/tomen21a.html
[ "Nergis Tomen", "Silvia-Laura Pintea", "Jan Van Gemert" ]
null
null
CNNs and computational models of biological vision share some fundamental principles, which opened new avenues of research. However, fruitful cross-field research is hampered by conventional CNN architectures being based on spatially and depthwise discrete representations, which cannot accommodate certain aspects of bi...
[]
null
940
2402.01557
title_snapshot
v139/tong21a
Diffusion Earth Mover’s Distance and Distribution Embeddings
https://proceedings.mlr.press/v139/tong21a.html
[ "Alexander Y Tong", "Guillaume Huguet", "Amine Natik", "Kincaid Macdonald", "Manik Kuchroo", "Ronald Coifman", "Guy Wolf", "Smita Krishnaswamy" ]
null
null
We propose a new fast method of measuring distances between large numbers of related high dimensional datasets called the Diffusion Earth Mover’s Distance (EMD). We model the datasets as distributions supported on common data graph that is derived from the affinity matrix computed on the combined data. In such cases wh...
[]
null
941
2102.12833
title_snapshot
v139/touvron21a
Training data-efficient image transformers & distillation through attention
https://proceedings.mlr.press/v139/touvron21a.html
[ "Hugo Touvron", "Matthieu Cord", "Matthijs Douze", "Francisco Massa", "Alexandre Sablayrolles", "Herve Jegou" ]
null
null
Recently, neural networks purely based on attention were shown to address image understanding tasks such as image classification. These high-performing vision transformers are pre-trained with hundreds of millions of images using a large infrastructure, thereby limiting their adoption. In this work, we produce competit...
[]
null
942
2012.12877
title_snapshot
v139/trabucco21a
Conservative Objective Models for Effective Offline Model-Based Optimization
https://proceedings.mlr.press/v139/trabucco21a.html
[ "Brandon Trabucco", "Aviral Kumar", "Xinyang Geng", "Sergey Levine" ]
null
null
In this paper, we aim to solve data-driven model-based optimization (MBO) problems, where the goal is to find a design input that maximizes an unknown objective function provided access to only a static dataset of inputs and their corresponding objective values. Such data-driven optimization procedures are the only pra...
[]
null
943
2107.06882
title_snapshot
v139/tran21a
Sparse within Sparse Gaussian Processes using Neighbor Information
https://proceedings.mlr.press/v139/tran21a.html
[ "Gia-Lac Tran", "Dimitrios Milios", "Pietro Michiardi", "Maurizio Filippone" ]
null
null
Approximations to Gaussian processes (GPs) based on inducing variables, combined with variational inference techniques, enable state-of-the-art sparse approaches to infer GPs at scale through mini-batch based learning. In this work, we further push the limits of scalability of sparse GPs by allowing large number of ind...
[]
null
944
2011.05041
title_snapshot
v139/tran21b
SMG: A Shuffling Gradient-Based Method with Momentum
https://proceedings.mlr.press/v139/tran21b.html
[ "Trang H Tran", "Lam M Nguyen", "Quoc Tran-Dinh" ]
null
null
We combine two advanced ideas widely used in optimization for machine learning: \textit{shuffling} strategy and \textit{momentum} technique to develop a novel shuffling gradient-based method with momentum, coined \textbf{S}huffling \textbf{M}omentum \textbf{G}radient (SMG), for non-convex finite-sum optimization proble...
[]
null
945
2011.11884
title_snapshot
v139/tran-the21a
Bayesian Optimistic Optimisation with Exponentially Decaying Regret
https://proceedings.mlr.press/v139/tran-the21a.html
[ "Hung Tran-The", "Sunil Gupta", "Santu Rana", "Svetha Venkatesh" ]
null
null
Bayesian optimisation (BO) is a well known algorithm for finding the global optimum of expensive, black-box functions. The current practical BO algorithms have regret bounds ranging from $\mathcal{O}(\frac{logN}{\sqrt{N}})$ to $\mathcal O(e^{-\sqrt{N}})$, where $N$ is the number of evaluations. This paper explores the ...
[]
null
946
2105.04332
title_snapshot
v139/trauble21a
On Disentangled Representations Learned from Correlated Data
https://proceedings.mlr.press/v139/trauble21a.html
[ "Frederik Träuble", "Elliot Creager", "Niki Kilbertus", "Francesco Locatello", "Andrea Dittadi", "Anirudh Goyal", "Bernhard Schölkopf", "Stefan Bauer" ]
null
null
The focus of disentanglement approaches has been on identifying independent factors of variation in data. However, the causal variables underlying real-world observations are often not statistically independent. In this work, we bridge the gap to real-world scenarios by analyzing the behavior of the most prominent dise...
[]
null
947
2006.07886
title_snapshot
v139/treutlein21a
A New Formalism, Method and Open Issues for Zero-Shot Coordination
https://proceedings.mlr.press/v139/treutlein21a.html
[ "Johannes Treutlein", "Michael Dennis", "Caspar Oesterheld", "Jakob Foerster" ]
null
null
In many coordination problems, independently reasoning humans are able to discover mutually compatible policies. In contrast, independently trained self-play policies are often mutually incompatible. Zero-shot coordination (ZSC) has recently been proposed as a new frontier in multi-agent reinforcement learning to addre...
[]
null
948
2106.06613
title_snapshot
v139/triantafillou21a
Learning a Universal Template for Few-shot Dataset Generalization
https://proceedings.mlr.press/v139/triantafillou21a.html
[ "Eleni Triantafillou", "Hugo Larochelle", "Richard Zemel", "Vincent Dumoulin" ]
null
null
Few-shot dataset generalization is a challenging variant of the well-studied few-shot classification problem where a diverse training set of several datasets is given, for the purpose of training an adaptable model that can then learn classes from \emph{new datasets} using only a few examples. To this end, we propose t...
[]
null
949
2105.07029
title_snapshot
v139/tripuraneni21a
Provable Meta-Learning of Linear Representations
https://proceedings.mlr.press/v139/tripuraneni21a.html
[ "Nilesh Tripuraneni", "Chi Jin", "Michael Jordan" ]
null
null
Meta-learning, or learning-to-learn, seeks to design algorithms that can utilize previous experience to rapidly learn new skills or adapt to new environments. Representation learning—a key tool for performing meta-learning—learns a data representation that can transfer knowledge across multiple tasks, which is essentia...
[]
null
950
2002.11684
title_snapshot
v139/trouleau21a
Cumulants of Hawkes Processes are Robust to Observation Noise
https://proceedings.mlr.press/v139/trouleau21a.html
[ "William Trouleau", "Jalal Etesami", "Matthias Grossglauser", "Negar Kiyavash", "Patrick Thiran" ]
null
null
Multivariate Hawkes processes (MHPs) are widely used in a variety of fields to model the occurrence of causally related discrete events in continuous time. Most state-of-the-art approaches address the problem of learning MHPs from perfect traces without noise. In practice, the process through which events are collected...
[]
null
951
null
null
v139/tulsiani21a
PixelTransformer: Sample Conditioned Signal Generation
https://proceedings.mlr.press/v139/tulsiani21a.html
[ "Shubham Tulsiani", "Abhinav Gupta" ]
null
null
We propose a generative model that can infer a distribution for the underlying spatial signal conditioned on sparse samples e.g. plausible images given a few observed pixels. In contrast to sequential autoregressive generative models, our model allows conditioning on arbitrary samples and can answer distributional quer...
[]
null
952
2103.15813
title_snapshot
v139/upadhyay21a
A Framework for Private Matrix Analysis in Sliding Window Model
https://proceedings.mlr.press/v139/upadhyay21a.html
[ "Jalaj Upadhyay", "Sarvagya Upadhyay" ]
null
null
We perform a rigorous study of private matrix analysis when only the last $W$ updates to matrices are considered useful for analysis. We show the existing framework in the non-private setting is not robust to noise required for privacy. We then propose a framework robust to noise and use it to give first efficient $o(W...
[]
null
953
null
null
v139/usmanova21a
Fast Projection Onto Convex Smooth Constraints
https://proceedings.mlr.press/v139/usmanova21a.html
[ "Ilnura Usmanova", "Maryam Kamgarpour", "Andreas Krause", "Kfir Levy" ]
null
null
The Euclidean projection onto a convex set is an important problem that arises in numerous constrained optimization tasks. Unfortunately, in many cases, computing projections is computationally demanding. In this work, we focus on projection problems where the constraints are smooth and the number of constraints is sig...
[]
null
954
2109.09835
title_snapshot
v139/ustimenko21a
SGLB: Stochastic Gradient Langevin Boosting
https://proceedings.mlr.press/v139/ustimenko21a.html
[ "Aleksei Ustimenko", "Liudmila Prokhorenkova" ]
null
null
This paper introduces Stochastic Gradient Langevin Boosting (SGLB) - a powerful and efficient machine learning framework that may deal with a wide range of loss functions and has provable generalization guarantees. The method is based on a special form of the Langevin diffusion equation specifically designed for gradie...
[]
null
955
2001.07248
title_snapshot
v139/vaezipoor21a
LTL2Action: Generalizing LTL Instructions for Multi-Task RL
https://proceedings.mlr.press/v139/vaezipoor21a.html
[ "Pashootan Vaezipoor", "Andrew C Li", "Rodrigo A Toro Icarte", "Sheila A. Mcilraith" ]
null
null
We address the problem of teaching a deep reinforcement learning (RL) agent to follow instructions in multi-task environments. Instructions are expressed in a well-known formal language {–} linear temporal logic (LTL) {–} and can specify a diversity of complex, temporally extended behaviours, including conditionals and...
[]
null
956
2102.06858
title_snapshot
v139/van-gorp21a
Active Deep Probabilistic Subsampling
https://proceedings.mlr.press/v139/van-gorp21a.html
[ "Hans Van Gorp", "Iris Huijben", "Bastiaan S Veeling", "Nicola Pezzotti", "Ruud J. G. Van Sloun" ]
null
null
Subsampling a signal of interest can reduce costly data transfer, battery drain, radiation exposure and acquisition time in a wide range of problems. The recently proposed Deep Probabilistic Subsampling (DPS) method effectively integrates subsampling in an end-to-end deep learning model, but learns a static pattern for...
[]
null
957
null
null
v139/vedantam21a
CURI: A Benchmark for Productive Concept Learning Under Uncertainty
https://proceedings.mlr.press/v139/vedantam21a.html
[ "Ramakrishna Vedantam", "Arthur Szlam", "Maximillian Nickel", "Ari Morcos", "Brenden M Lake" ]
null
null
Humans can learn and reason under substantial uncertainty in a space of infinitely many compositional, productive concepts. For example, if a scene with two blue spheres qualifies as “daxy,” one can reason that the underlying concept may require scenes to have “only blue spheres” or “only spheres” or “only two objects....
[]
null
958
2010.02855
title_snapshot
v139/verma21a
Towards Domain-Agnostic Contrastive Learning
https://proceedings.mlr.press/v139/verma21a.html
[ "Vikas Verma", "Thang Luong", "Kenji Kawaguchi", "Hieu Pham", "Quoc Le" ]
null
null
Despite recent successes, most contrastive self-supervised learning methods are domain-specific, relying heavily on data augmentation techniques that require knowledge about a particular domain, such as image cropping and rotation. To overcome such limitation, we propose a domain-agnostic approach to contrastive learni...
[]
null
959
2011.04419
title_snapshot
v139/verma21b
Sparsifying Networks via Subdifferential Inclusion
https://proceedings.mlr.press/v139/verma21b.html
[ "Sagar Verma", "Jean-Christophe Pesquet" ]
null
null
Sparsifying deep neural networks is of paramount interest in many areas, especially when those networks have to be implemented on low-memory devices. In this article, we propose a new formulation of the problem of generating sparse weights for a pre-trained neural network. By leveraging the properties of standard nonli...
[]
null
960
null
null
v139/vicol21a
Unbiased Gradient Estimation in Unrolled Computation Graphs with Persistent Evolution Strategies
https://proceedings.mlr.press/v139/vicol21a.html
[ "Paul Vicol", "Luke Metz", "Jascha Sohl-Dickstein" ]
null
null
Unrolled computation graphs arise in many scenarios, including training RNNs, tuning hyperparameters through unrolled optimization, and training learned optimizers. Current approaches to optimizing parameters in such computation graphs suffer from high variance gradients, bias, slow updates, or large memory usage. We i...
[]
null
961
2112.13835
title_snapshot
v139/vincent-cuaz21a
Online Graph Dictionary Learning
https://proceedings.mlr.press/v139/vincent-cuaz21a.html
[ "Cédric Vincent-Cuaz", "Titouan Vayer", "Rémi Flamary", "Marco Corneli", "Nicolas Courty" ]
null
null
Dictionary learning is a key tool for representation learning, that explains the data as linear combination of few basic elements. Yet, this analysis is not amenable in the context of graph learning, as graphs usually belong to different metric spaces. We fill this gap by proposing a new online Graph Dictionary Learnin...
[]
null
962
2102.06555
title_snapshot
v139/vlastelica21a
Neuro-algorithmic Policies Enable Fast Combinatorial Generalization
https://proceedings.mlr.press/v139/vlastelica21a.html
[ "Marin Vlastelica", "Michal Rolinek", "Georg Martius" ]
null
null
Although model-based and model-free approaches to learning the control of systems have achieved impressive results on standard benchmarks, generalization to task variations is still lacking. Recent results suggest that generalization for standard architectures improves only after obtaining exhaustive amounts of data. W...
[]
null
963
2102.07456
title_snapshot
v139/vos21a
Efficient Training of Robust Decision Trees Against Adversarial Examples
https://proceedings.mlr.press/v139/vos21a.html
[ "Daniël Vos", "Sicco Verwer" ]
null
null
Current state-of-the-art algorithms for training robust decision trees have high runtime costs and require hours to run. We present GROOT, an efficient algorithm for training robust decision trees and random forests that runs in a matter of seconds to minutes. Where before the worst-case Gini impurity was computed iter...
[]
null
964
2012.10438
title_snapshot
v139/voynov21a
Object Segmentation Without Labels with Large-Scale Generative Models
https://proceedings.mlr.press/v139/voynov21a.html
[ "Andrey Voynov", "Stanislav Morozov", "Artem Babenko" ]
null
null
The recent rise of unsupervised and self-supervised learning has dramatically reduced the dependency on labeled data, providing high-quality representations for transfer on downstream tasks. Furthermore, recent works also employed these representations in a fully unsupervised setup for image classification, reducing th...
[]
null
965
2006.04988
title_snapshot
v139/vreugdenhil21a
Principal Component Hierarchy for Sparse Quadratic Programs
https://proceedings.mlr.press/v139/vreugdenhil21a.html
[ "Robbie Vreugdenhil", "Viet Anh Nguyen", "Armin Eftekhari", "Peyman Mohajerin Esfahani" ]
null
null
We propose a novel approximation hierarchy for cardinality-constrained, convex quadratic programs that exploits the rank-dominating eigenvectors of the quadratic matrix. Each level of approximation admits a min-max characterization whose objective function can be optimized over the binary variables analytically, while ...
[]
null
966
2105.12022
title_snapshot
v139/wadia21a
Whitening and Second Order Optimization Both Make Information in the Dataset Unusable During Training, and Can Reduce or Prevent Generalization
https://proceedings.mlr.press/v139/wadia21a.html
[ "Neha Wadia", "Daniel Duckworth", "Samuel S Schoenholz", "Ethan Dyer", "Jascha Sohl-Dickstein" ]
null
null
Machine learning is predicated on the concept of generalization: a model achieving low error on a sufficiently large training set should also perform well on novel samples from the same distribution. We show that both data whitening and second order optimization can harm or entirely prevent generalization. In general, ...
[]
null
967
2008.07545
title_snapshot
v139/wagener21a
Safe Reinforcement Learning Using Advantage-Based Intervention
https://proceedings.mlr.press/v139/wagener21a.html
[ "Nolan C Wagener", "Byron Boots", "Ching-An Cheng" ]
null
null
Many sequential decision problems involve finding a policy that maximizes total reward while obeying safety constraints. Although much recent research has focused on the development of safe reinforcement learning (RL) algorithms that produce a safe policy after training, ensuring safety during training as well remains ...
[]
null
968
2106.09110
title_snapshot
v139/wagenmaker21a
Task-Optimal Exploration in Linear Dynamical Systems
https://proceedings.mlr.press/v139/wagenmaker21a.html
[ "Andrew J Wagenmaker", "Max Simchowitz", "Kevin Jamieson" ]
null
null
Exploration in unknown environments is a fundamental problem in reinforcement learning and control. In this work, we study task-guided exploration and determine what precisely an agent must learn about their environment in order to complete a particular task. Formally, we study a broad class of decision-making problems...
[]
null
969
2102.05214
title_snapshot
v139/wan21a
Learning and Planning in Average-Reward Markov Decision Processes
https://proceedings.mlr.press/v139/wan21a.html
[ "Yi Wan", "Abhishek Naik", "Richard S Sutton" ]
null
null
We introduce learning and planning algorithms for average-reward MDPs, including 1) the first general proven-convergent off-policy model-free control algorithm without reference states, 2) the first proven-convergent off-policy model-free prediction algorithm, and 3) the first off-policy learning algorithm that converg...
[]
null
970
2006.16318
title_snapshot
v139/wan21b
Think Global and Act Local: Bayesian Optimisation over High-Dimensional Categorical and Mixed Search Spaces
https://proceedings.mlr.press/v139/wan21b.html
[ "Xingchen Wan", "Vu Nguyen", "Huong Ha", "Binxin Ru", "Cong Lu", "Michael A. Osborne" ]
null
null
High-dimensional black-box optimisation remains an important yet notoriously challenging problem. Despite the success of Bayesian optimisation methods on continuous domains, domains that are categorical, or that mix continuous and categorical variables, remain challenging. We propose a novel solution—we combine local o...
[]
null
971
2102.07188
title_snapshot
v139/wang21a
Zero-Shot Knowledge Distillation from a Decision-Based Black-Box Model
https://proceedings.mlr.press/v139/wang21a.html
[ "Zi Wang" ]
null
null
Knowledge distillation (KD) is a successful approach for deep neural network acceleration, with which a compact network (student) is trained by mimicking the softmax output of a pre-trained high-capacity network (teacher). In tradition, KD usually relies on access to the training samples and the parameters of the white...
[]
null
972
2106.03310
title_snapshot
v139/wang21b
Fairness of Exposure in Stochastic Bandits
https://proceedings.mlr.press/v139/wang21b.html
[ "Lequn Wang", "Yiwei Bai", "Wen Sun", "Thorsten Joachims" ]
null
null
Contextual bandit algorithms have become widely used for recommendation in online systems (e.g. marketplaces, music streaming, news), where they now wield substantial influence on which items get shown to users. This raises questions of fairness to the items — and to the sellers, artists, and writers that benefit from ...
[]
null
973
2103.02735
title_snapshot
v139/wang21c
A Proxy Variable View of Shared Confounding
https://proceedings.mlr.press/v139/wang21c.html
[ "Yixin Wang", "David Blei" ]
null
null
Causal inference from observational data can be biased by unobserved confounders. Confounders{—}the variables that affect both the treatments and the outcome{—}induce spurious non-causal correlations between the two. Without additional conditions, unobserved confounders generally make causal quantities hard to identify...
[]
null
974
null
null
v139/wang21d
Fast Algorithms for Stackelberg Prediction Game with Least Squares Loss
https://proceedings.mlr.press/v139/wang21d.html
[ "Jiali Wang", "He Chen", "Rujun Jiang", "Xudong Li", "Zihao Li" ]
null
null
The Stackelberg prediction game (SPG) has been extensively used to model the interactions between the learner and data provider in the training process of various machine learning algorithms. Particularly, SPGs played prominent roles in cybersecurity applications, such as intrusion detection, banking fraud detection, s...
[]
null
975
2105.05531
title_snapshot
v139/wang21e
Accelerate CNNs from Three Dimensions: A Comprehensive Pruning Framework
https://proceedings.mlr.press/v139/wang21e.html
[ "Wenxiao Wang", "Minghao Chen", "Shuai Zhao", "Long Chen", "Jinming Hu", "Haifeng Liu", "Deng Cai", "Xiaofei He", "Wei Liu" ]
null
null
Most neural network pruning methods, such as filter-level and layer-level prunings, prune the network model along one dimension (depth, width, or resolution) solely to meet a computational budget. However, such a pruning policy often leads to excessive reduction of that dimension, thus inducing a huge accuracy loss. To...
[]
null
976
2010.04879
title_snapshot
v139/wang21f
Explainable Automated Graph Representation Learning with Hyperparameter Importance
https://proceedings.mlr.press/v139/wang21f.html
[ "Xin Wang", "Shuyi Fan", "Kun Kuang", "Wenwu Zhu" ]
null
null
Current graph representation (GR) algorithms require huge demand of human experts in hyperparameter tuning, which significantly limits their practical applications, leading to an urge for automated graph representation without human intervention. Although automated machine learning (AutoML) serves as a good candidate f...
[]
null
977
null
null
v139/wang21g
Self-Tuning for Data-Efficient Deep Learning
https://proceedings.mlr.press/v139/wang21g.html
[ "Ximei Wang", "Jinghan Gao", "Mingsheng Long", "Jianmin Wang" ]
null
null
Deep learning has made revolutionary advances to diverse applications in the presence of large-scale labeled datasets. However, it is prohibitively time-costly and labor-expensive to collect sufficient labeled data in most realistic scenarios. To mitigate the requirement for labeled data, semi-supervised learning (SSL)...
[]
null
978
2102.12903
title_snapshot
v139/wang21h
Label Distribution Learning Machine
https://proceedings.mlr.press/v139/wang21h.html
[ "Jing Wang", "Xin Geng" ]
null
null
Although Label Distribution Learning (LDL) has witnessed extensive classification applications, it faces the challenge of objective mismatch – the objective of LDL mismatches that of classification, which has seldom been noticed in existing studies. Our goal is to solve the objective mismatch and improve the classifica...
[]
null
979
null
null
v139/wang21i
AlphaNet: Improved Training of Supernets with Alpha-Divergence
https://proceedings.mlr.press/v139/wang21i.html
[ "Dilin Wang", "Chengyue Gong", "Meng Li", "Qiang Liu", "Vikas Chandra" ]
null
null
Weight-sharing neural architecture search (NAS) is an effective technique for automating efficient neural architecture design. Weight-sharing NAS builds a supernet that assembles all the architectures as its sub-networks and jointly trains the supernet with the sub-networks. The success of weight-sharing NAS heavily re...
[]
null
980
2102.07954
title_snapshot
v139/wang21j
Global Convergence of Policy Gradient for Linear-Quadratic Mean-Field Control/Game in Continuous Time
https://proceedings.mlr.press/v139/wang21j.html
[ "Weichen Wang", "Jiequn Han", "Zhuoran Yang", "Zhaoran Wang" ]
null
null
Recent years have witnessed the success of multi-agent reinforcement learning, which has motivated new research directions for mean-field control (MFC) and mean-field game (MFG), as the multi-agent system can be well approximated by a mean-field problem when the number of agents grows to be very large. In this paper, w...
[]
null
981
2008.06845
title_snapshot
v139/wang21k
SG-PALM: a Fast Physically Interpretable Tensor Graphical Model
https://proceedings.mlr.press/v139/wang21k.html
[ "Yu Wang", "Alfred Hero" ]
null
null
We propose a new graphical model inference procedure, called SG-PALM, for learning conditional dependency structure of high-dimensional tensor-variate data. Unlike most other tensor graphical models the proposed model is interpretable and computationally scalable to high dimension. Physical interpretability follows fro...
[]
null
982
2105.12271
title_snapshot
v139/wang21l
Deep Generative Learning via Schrödinger Bridge
https://proceedings.mlr.press/v139/wang21l.html
[ "Gefei Wang", "Yuling Jiao", "Qian Xu", "Yang Wang", "Can Yang" ]
null
null
We propose to learn a generative model via entropy interpolation with a Schr{ö}dinger Bridge. The generative learning task can be formulated as interpolating between a reference distribution and a target distribution based on the Kullback-Leibler divergence. At the population level, this entropy interpolation is charac...
[]
null
983
2106.10410
title_snapshot
v139/wang21m
Robust Inference for High-Dimensional Linear Models via Residual Randomization
https://proceedings.mlr.press/v139/wang21m.html
[ "Y. Samuel Wang", "Si Kai Lee", "Panos Toulis", "Mladen Kolar" ]
null
null
We propose a residual randomization procedure designed for robust inference using Lasso estimates in the high-dimensional setting. Compared to earlier work that focuses on sub-Gaussian errors, the proposed procedure is designed to work robustly in settings that also include heavy-tailed covariates and errors. Moreover,...
[]
null
984
2106.07717
title_snapshot
v139/wang21n
A Modular Analysis of Provable Acceleration via Polyak’s Momentum: Training a Wide ReLU Network and a Deep Linear Network
https://proceedings.mlr.press/v139/wang21n.html
[ "Jun-Kun Wang", "Chi-Heng Lin", "Jacob D Abernethy" ]
null
null
Incorporating a so-called “momentum” dynamic in gradient descent methods is widely used in neural net training as it has been broadly observed that, at least empirically, it often leads to significantly faster convergence. At the same time, there are very few theoretical guarantees in the literature to explain this app...
[]
null
985
2010.01618
title_snapshot
v139/wang21o
Optimal Non-Convex Exact Recovery in Stochastic Block Model via Projected Power Method
https://proceedings.mlr.press/v139/wang21o.html
[ "Peng Wang", "Huikang Liu", "Zirui Zhou", "Anthony Man-Cho So" ]
null
null
In this paper, we study the problem of exact community recovery in the symmetric stochastic block model, where a graph of $n$ vertices is randomly generated by partitioning the vertices into $K \ge 2$ equal-sized communities and then connecting each pair of vertices with probability that depends on their community memb...
[]
null
986
2106.05644
title_snapshot
v139/wang21p
ConvexVST: A Convex Optimization Approach to Variance-stabilizing Transformation
https://proceedings.mlr.press/v139/wang21p.html
[ "Mengfan Wang", "Boyu Lyu", "Guoqiang Yu" ]
null
null
The variance-stabilizing transformation (VST) problem is to transform heteroscedastic data to homoscedastic data so that they are more tractable for subsequent analysis. However, most of the existing approaches focus on finding an analytical solution for a certain parametric distribution, which severely limits the appl...
[]
null
987
null
null
v139/wang21q
The Implicit Bias for Adaptive Optimization Algorithms on Homogeneous Neural Networks
https://proceedings.mlr.press/v139/wang21q.html
[ "Bohan Wang", "Qi Meng", "Wei Chen", "Tie-Yan Liu" ]
null
null
Despite their overwhelming capacity to overfit, deep neural networks trained by specific optimization algorithms tend to generalize relatively well to unseen data. Recently, researchers explained it by investigating the implicit bias of optimization algorithms. A remarkable progress is the work (Lyu & Li, 2019), which ...
[]
null
988
2012.06244
title_snapshot
v139/wang21r
Robust Learning for Data Poisoning Attacks
https://proceedings.mlr.press/v139/wang21r.html
[ "Yunjuan Wang", "Poorya Mianjy", "Raman Arora" ]
null
null
We investigate the robustness of stochastic approximation approaches against data poisoning attacks. We focus on two-layer neural networks with ReLU activation and show that under a specific notion of separability in the RKHS induced by the infinite-width network, training (finite-width) networks with stochastic gradie...
[]
null
989
null
null
v139/wang21s
SketchEmbedNet: Learning Novel Concepts by Imitating Drawings
https://proceedings.mlr.press/v139/wang21s.html
[ "Alexander Wang", "Mengye Ren", "Richard Zemel" ]
null
null
Sketch drawings capture the salient information of visual concepts. Previous work has shown that neural networks are capable of producing sketches of natural objects drawn from a small number of classes. While earlier approaches focus on generation quality or retrieval, we explore properties of image representations le...
[]
null
990
2009.04806
title_snapshot
v139/wang21t
Directional Bias Amplification
https://proceedings.mlr.press/v139/wang21t.html
[ "Angelina Wang", "Olga Russakovsky" ]
null
null
Mitigating bias in machine learning systems requires refining our understanding of bias propagation pathways: from societal structures to large-scale data to trained models to impact on society. In this work, we focus on one aspect of the problem, namely bias amplification: the tendency of models to amplify the biases ...
[]
null
991
2102.12594
title_snapshot
v139/wang21u
An exact solver for the Weston-Watkins SVM subproblem
https://proceedings.mlr.press/v139/wang21u.html
[ "Yutong Wang", "Clayton Scott" ]
null
null
Recent empirical evidence suggests that the Weston-Watkins support vector machine is among the best performing multiclass extensions of the binary SVM. Current state-of-the-art solvers repeatedly solve a particular subproblem approximately using an iterative strategy. In this work, we propose an algorithm that solves t...
[]
null
992
2102.05640
title_snapshot
v139/wang21v
SCC: an efficient deep reinforcement learning agent mastering the game of StarCraft II
https://proceedings.mlr.press/v139/wang21v.html
[ "Xiangjun Wang", "Junxiao Song", "Penghui Qi", "Peng Peng", "Zhenkun Tang", "Wei Zhang", "Weimin Li", "Xiongjun Pi", "Jujie He", "Chao Gao", "Haitao Long", "Quan Yuan" ]
null
null
AlphaStar, the AI that reaches GrandMaster level in StarCraft II, is a remarkable milestone demonstrating what deep reinforcement learning can achieve in complex Real-Time Strategy (RTS) games. However, the complexities of the game, algorithms and systems, and especially the tremendous amount of computation needed are ...
[]
null
993
2012.13169
title_snapshot
v139/wang21w
Quantum algorithms for reinforcement learning with a generative model
https://proceedings.mlr.press/v139/wang21w.html
[ "Daochen Wang", "Aarthi Sundaram", "Robin Kothari", "Ashish Kapoor", "Martin Roetteler" ]
null
null
Reinforcement learning studies how an agent should interact with an environment to maximize its cumulative reward. A standard way to study this question abstractly is to ask how many samples an agent needs from the environment to learn an optimal policy for a $\gamma$-discounted Markov decision process (MDP). For such ...
[]
null
994
2112.08451
title_snapshot
v139/wang21x
Matrix Completion with Model-free Weighting
https://proceedings.mlr.press/v139/wang21x.html
[ "Jiayi Wang", "Raymond K. W. Wong", "Xiaojun Mao", "Kwun Chuen Gary Chan" ]
null
null
In this paper, we propose a novel method for matrix completion under general non-uniform missing structures. By controlling an upper bound of a novel balancing error, we construct weights that can actively adjust for the non-uniformity in the empirical risk without explicitly modeling the observation probabilities, and...
[]
null
995
2106.05850
title_snapshot
v139/wang21y
UniSpeech: Unified Speech Representation Learning with Labeled and Unlabeled Data
https://proceedings.mlr.press/v139/wang21y.html
[ "Chengyi Wang", "Yu Wu", "Yao Qian", "Kenichi Kumatani", "Shujie Liu", "Furu Wei", "Michael Zeng", "Xuedong Huang" ]
null
null
In this paper, we propose a unified pre-training approach called UniSpeech to learn speech representations with both labeled and unlabeled data, in which supervised phonetic CTC learning and phonetically-aware contrastive self-supervised learning are conducted in a multi-task learning manner. The resultant representati...
[]
null
996
2101.07597
title_snapshot
v139/wang21z
Instabilities of Offline RL with Pre-Trained Neural Representation
https://proceedings.mlr.press/v139/wang21z.html
[ "Ruosong Wang", "Yifan Wu", "Ruslan Salakhutdinov", "Sham Kakade" ]
null
null
In offline reinforcement learning (RL), we seek to utilize offline data to evaluate (or learn) policies in scenarios where the data are collected from a distribution that substantially differs from that of the target policy to be evaluated. Recent theoretical advances have shown that such sample-efficient offline RL is...
[]
null
997
2103.04947
title_snapshot
v139/wang21aa
Learning to Weight Imperfect Demonstrations
https://proceedings.mlr.press/v139/wang21aa.html
[ "Yunke Wang", "Chang Xu", "Bo Du", "Honglak Lee" ]
null
null
This paper investigates how to weight imperfect expert demonstrations for generative adversarial imitation learning (GAIL). The agent is expected to perform behaviors demonstrated by experts. But in many applications, experts could also make mistakes and their demonstrations would mislead or slow the learning process o...
[]
null
998
null
null
v139/wang21ab
Evolving Attention with Residual Convolutions
https://proceedings.mlr.press/v139/wang21ab.html
[ "Yujing Wang", "Yaming Yang", "Jiangang Bai", "Mingliang Zhang", "Jing Bai", "Jing Yu", "Ce Zhang", "Gao Huang", "Yunhai Tong" ]
null
null
Transformer is a ubiquitous model for natural language processing and has attracted wide attentions in computer vision. The attention maps are indispensable for a transformer model to encode the dependencies among input tokens. However, they are learned independently in each layer and sometimes fail to capture precise ...
[]
null
999
2102.12895
title_snapshot
v139/wang21ac
Guarantees for Tuning the Step Size using a Learning-to-Learn Approach
https://proceedings.mlr.press/v139/wang21ac.html
[ "Xiang Wang", "Shuai Yuan", "Chenwei Wu", "Rong Ge" ]
null
null
Choosing the right parameters for optimization algorithms is often the key to their success in practice. Solving this problem using a learning-to-learn approach—using meta-gradient descent on a meta-objective based on the trajectory that the optimizer generates—was recently shown to be effective. However, the meta-opti...
[]
null
1,000
2006.16495
title_snapshot