paper_id
stringlengths
9
13
title
stringlengths
15
153
paper_url
stringlengths
41
45
authors
listlengths
1
21
type
stringclasses
3 values
primary_area
stringclasses
0 values
abstract
large_stringlengths
310
2.43k
keywords
listlengths
1
32
TL;DR
large_stringlengths
23
250
submission_number
int64
1
11.7k
arxiv_id
stringlengths
10
10
arxiv_id_source
stringclasses
2 values
nYz2_BbZnYk
Representing Long-Range Context for Graph Neural Networks with Global Attention
https://openreview.net/forum?id=nYz2_BbZnYk
[ "Zhanghao Wu", "Paras Jain", "Matthew A. Wright", "Azalia Mirhoseini", "Joseph E. Gonzalez", "Ion Stoica" ]
Poster
null
Graph neural networks are powerful architectures for structured datasets. However, current methods struggle to represent long-range dependencies. Scaling the depth or width of GNNs is insufficient to broaden receptive fields as larger GNNs encounter optimization instabilities such as vanishing gradients and representat...
[ "graph neural networks", "transformers", "long-range context", "OpenGraphBenchmark" ]
We generalize the Transformer structure to improve long-range context for graph neural networks to achieve a new SOTA on the OpenGraphBenchmark.
11,704
2201.08821
title_snapshot
GPwmbxtG9Ow
Bootstrapping the Error of Oja's Algorithm
https://openreview.net/forum?id=GPwmbxtG9Ow
[ "Robert Lunde", "Purnamrita Sarkar", "Rachel Ward" ]
Spotlight
null
We consider the problem of quantifying uncertainty for the estimation error of the leading eigenvector from Oja's algorithm for streaming principal component analysis, where the data are generated IID from some unknown distribution. By combining classical tools from the U-statistics literature with recent results on h...
[ "Gaussian approximation", "bootstrap", "Streaming PCA" ]
We establish a high-dimensional central limit theorem and online bootstrap procedure for inferring the error of Oja's algorithm, which is a widely used method in Streaming PCA
11,689
2106.14857
title_snapshot
h7FqQ6hCK18
Linear Convergence in Federated Learning: Tackling Client Heterogeneity and Sparse Gradients
https://openreview.net/forum?id=h7FqQ6hCK18
[ "Aritra Mitra", "Rayana Jaafar", "George J. Pappas", "Hamed Hassani" ]
Poster
null
We consider a standard federated learning (FL) setup where a group of clients periodically coordinate with a central server to train a statistical model. We develop a general algorithmic framework called FedLin to tackle some of the key challenges intrinsic to FL, namely objective heterogeneity, systems heterogeneity, ...
[ "Federated Learning", "Distributed Optimization", "Compression", "Heterogeneity", "Linear Convergence" ]
We develop a novel federated learning algorithm that guarantees linear convergence under client heterogeneity and aggressive gradient sparsification, and provide a tight linear convergence rate analysis.
11,681
2102.07053
title_snapshot
73FeFxePGc
Adaptable Agent Populations via a Generative Model of Policies
https://openreview.net/forum?id=73FeFxePGc
[ "Kenneth Derek", "Phillip Isola" ]
Poster
null
In the natural world, life has found innumerable ways to survive and often thrive. Between and even within species, each individual is in some manner unique, and this diversity lends adaptability and robustness to life. In this work, we aim to learn a space of diverse and high-reward policies in a given environment. To...
[ "reinforcement learning", "quality diversity", "multiagent", "generative models", "latent policy" ]
We learn a multi-modal policy space in a reinforcement learning setting that creates diverse and adaptable agent populations.
11,670
2107.07506
title_snapshot
w6U6g5Bvug
The Utility of Explainable AI in Ad Hoc Human-Machine Teaming
https://openreview.net/forum?id=w6U6g5Bvug
[ "Rohan R Paleja", "Muyleng Ghuy", "Nadun Ranawaka Arachchige", "Reed Jensen", "Matthew Gombolay" ]
Poster
null
Recent advances in machine learning have led to growing interest in Explainable AI (xAI) to enable humans to gain insight into the decision-making of machine learning models. Despite this recent interest, the utility of xAI techniques has not yet been characterized in human-machine teaming. Importantly, xAI offers the ...
[ "Explainable AI", "Human-Machine Teaming", "Situational Awareness", "Ad Hoc Teaming" ]
We present two human-subject studies quantifying the benefits of deploying Explainable AI techniques within a human-machine teaming scenario, finding that the benefits of xAI are not universal.
11,642
2209.03943
title_snapshot
e9_UPqMNfi
Control Variates for Slate Off-Policy Evaluation
https://openreview.net/forum?id=e9_UPqMNfi
[ "Nikos Vlassis", "Ashok Chandrashekar", "Fernando Amat", "Nathan Kallus" ]
Poster
null
We study the problem of off-policy evaluation from batched contextual bandit data with multidimensional actions, often termed slates. The problem is common to recommender systems and user-interface optimization, and it is particularly challenging because of the combinatorially-sized action space. Swaminathan et al. (20...
[ "off-policy evaluation", "combinatorial actions", "slate bandits", "control variates" ]
Using control variates, we develop improved estimators for off-policy evaluation of contextual slate bandits
11,629
2106.07914
title_snapshot
oIhzg4GJeOf
Learning Semantic Representations to Verify Hardware Designs
https://openreview.net/forum?id=oIhzg4GJeOf
[ "Shobha Vasudevan", "Wenjie Jiang", "David Bieber", "Rishabh Singh", "HAMID SHOJAEI", "C. Richard Ho", "Charles Sutton" ]
Poster
null
Verification is a serious bottleneck in the industrial hardware design cycle, routinely requiring person-years of effort. Practical verification relies on a "best effort" process that simulates the design on test inputs. This suggests a new research question: Can this simulation data be exploited to learn a continuous ...
[ "Hardware Design", "Verification", "Graph Convolutional Networks", "Test generation" ]
Deep networks can learn semantic abstractions of hardware designs, analogous to software --- this brings down time from overnight to few seconds and scales to industrial designs.
11,628
null
null
rD6ulZFTbf
Successor Feature Landmarks for Long-Horizon Goal-Conditioned Reinforcement Learning
https://openreview.net/forum?id=rD6ulZFTbf
[ "Christopher Hoang", "Sungryull Sohn", "Jongwook Choi", "Wilka Torrico Carvalho", "Honglak Lee" ]
Poster
null
Operating in the real-world often requires agents to learn about a complex environment and apply this understanding to achieve a breadth of goals. This problem, known as goal-conditioned reinforcement learning (GCRL), becomes especially challenging for long-horizon goals. Current methods have tackled this problem by au...
[ "successor features", "goal-conditioned RL", "graph-based planning" ]
Graph-based planning framework exploiting successor features to achieve long-horizon goal-conditioned RL
11,626
2111.09858
title_snapshot
Uwh-v1HSw-x
Training Neural Networks with Fixed Sparse Masks
https://openreview.net/forum?id=Uwh-v1HSw-x
[ "Yi-Lin Sung", "Varun Nair", "Colin Raffel" ]
Poster
null
During typical gradient-based training of deep neural networks, all of the model's parameters are updated at each iteration. Recent work has shown that it is possible to update only a small subset of the model's parameters during training, which can alleviate storage and communication requirements. In this paper, we sh...
[ "Fisher information", "sparse updates", "parameter-efficient transfer learning", "distributed training", "efficient checkpointing" ]
We introduce a method for pre-computing a fixed mask that selects a subset of network parameters to update, alleviating storage and communication costs.
11,625
2111.09839
title_snapshot
ZBeCVICs1Ua
KALE Flow: A Relaxed KL Gradient Flow for Probabilities with Disjoint Support
https://openreview.net/forum?id=ZBeCVICs1Ua
[ "Pierre Glaser", "Michael Arbel", "Arthur Gretton" ]
Poster
null
We study the gradient flow for a relaxed approximation to the Kullback-Leibler (KL) divergence between a moving source and a fixed target distribution. This approximation, termed the KALE (KL approximate lower-bound estimator), solves a regularized version of the Fenchel dual problem defining the KL over a restricted c...
[ "Generative Models", "Kernel Methods", "Optimal Transportation", "Probability Divergences" ]
We introduce the KALE, a kernel based probability divergence that interpolates between the KL and the MMD, and study its wasserstein gradient flow.
11,611
2106.08929
title_snapshot
9B0JMeySlZM
Answering Complex Causal Queries With the Maximum Causal Set Effect
https://openreview.net/forum?id=9B0JMeySlZM
[ "Zachary Markovich" ]
Poster
null
The standard tools of causal inference have been developed to answer simple causal queries which can be easily formalized as a small number of statistical estimands in the context of a particular structural causal model (SCM); however, scientific theories often make diffuse predictions about a large number of causal va...
[ "Causal Inference", "frequentist inference", "hypothesis testing" ]
Develops a framework for testing complex causal theories that make predictions about the joint influence of a large number of causal variables
11,608
null
null
8RnRLP4SHe0
Regulating algorithmic filtering on social media
https://openreview.net/forum?id=8RnRLP4SHe0
[ "Sarah Cen", "Devavrat Shah" ]
Spotlight
null
By filtering the content that users see, social media platforms have the ability to influence users' perceptions and decisions, from their dining choices to their voting preferences. This influence has drawn scrutiny, with many calling for regulations on filtering algorithms, but designing and enforcing regulations rem...
[ "social media", "regulation", "audit", "filtering algorithm", "performance cost", "content diversity", "counterfactual", "hypothesis testing", "minimum-variance unbiased estimator" ]
We propose an auditing procedure for enforcing social media regulations, provide theoretical guarantees on the audit, study whether there is a performance-regulation tradeoff, and find that content diversity plays a key role.
11,603
2006.09647
title_snapshot
dvyUaK4neD0
The Skellam Mechanism for Differentially Private Federated Learning
https://openreview.net/forum?id=dvyUaK4neD0
[ "Naman Agarwal", "Peter Kairouz", "Ziyu Liu" ]
Poster
null
We introduce the multi-dimensional Skellam mechanism, a discrete differential privacy mechanism based on the difference of two independent Poisson random variables. To quantify its privacy guarantees, we analyze the privacy loss distribution via a numerical evaluation and provide a sharp bound on the Rényi divergence b...
[ "Differential Privacy", "Distributed Learning", "Federated Learning", "Secure Aggregation" ]
We introduce and analyze the multi-dimensional Skellam mechanism, an easy-to-implement discrete differential privacy mechanism for centralized and federated learning applications with secure aggregation.
11,593
2110.04995
title_snapshot
SFLSOd_hv-4
Robust Predictable Control
https://openreview.net/forum?id=SFLSOd_hv-4
[ "Benjamin Eysenbach", "Ruslan Salakhutdinov", "Sergey Levine" ]
Spotlight
null
Many of the challenges facing today's reinforcement learning (RL) algorithms, such as robustness, generalization, transfer, and computational efficiency are closely related to compression. Prior work has convincingly argued why minimizing information is useful in the supervised learning setting, but standard RL algori...
[ "reinforcement learning", "information bottleneck" ]
We propose a method for learning robust and predictable policies in RL using ideas from compression.
11,587
2109.03214
title_snapshot
dHc1p5eoecb
Optimal prediction of Markov chains with and without spectral gap
https://openreview.net/forum?id=dHc1p5eoecb
[ "Yanjun Han", "Soham Jana", "Yihong Wu" ]
Poster
null
We study the following learning problem with dependent data: Given a trajectory of length $n$ from a stationary Markov chain with $k$ states, the goal is to predict the distribution of the next state. For $3 \leq k \leq O(\sqrt{n})$, the optimal prediction risk in the Kullback-Leibler divergence is shown to be $\Theta(...
[ "Markov chains", "prediction", "redundancy", "spectral gap", "mixing time", "Kullback Leibler risk" ]
We study a prediction problem on Markov chains with finite state space and obtain optimal minimax rates.
11,568
2106.13947
title_snapshot
J2YvvXDp7H
Sample-Efficient Reinforcement Learning for Linearly-Parameterized MDPs with a Generative Model
https://openreview.net/forum?id=J2YvvXDp7H
[ "Bingyan Wang", "Yuling Yan", "Jianqing Fan" ]
Poster
null
The curse of dimensionality is a widely known issue in reinforcement learning (RL). In the tabular setting where the state space $\mathcal{S}$ and the action space $\mathcal{A}$ are both finite, to obtain a near optimal policy with sampling access to a generative model, the minimax optimal sample complexity scales line...
[ "model-based reinforcement learning", "vanilla Q-learning", "linear transition model", "sample complexity", "leave-one-out analysis" ]
null
11,533
2105.14016
title_snapshot
jgze2dDL9y8
Reinforcement Learning with State Observation Costs in Action-Contingent Noiselessly Observable Markov Decision Processes
https://openreview.net/forum?id=jgze2dDL9y8
[ "HyunJi Nam", "Scott L Fleming", "Emma Brunskill" ]
Poster
null
Many real-world problems that require making optimal sequences of decisions under uncertainty involve costs when the agent wishes to obtain information about its environment. We design and analyze algorithms for reinforcement learning (RL) in Action-Contingent Noiselessly Observable MDPs (ACNO-MDPs), a special class of...
[ "Reinforcement Learning", "Observation Cost", "Markov Decision Process", "MDP", "Partially Observable Markov Decision Process", "POMDP", "Probably Approximately Correct", "PAC", "Healthcare", "Health care" ]
We provide theory and algorithms for efficient reinforcement learning in a special class of POMDPs exhibiting full observability (albeit at a cost) contingent on an agent's actions
11,527
null
null
K5YKjaMjbja
Neural Algorithmic Reasoners are Implicit Planners
https://openreview.net/forum?id=K5YKjaMjbja
[ "Andreea Deac", "Petar Veličković", "Ognjen Milinković", "Pierre-Luc Bacon", "Jian Tang", "Mladen Nikolic" ]
Spotlight
null
Implicit planning has emerged as an elegant technique for combining learned models of the world with end-to-end model-free reinforcement learning. We study the class of implicit planners inspired by value iteration, an algorithm that is guaranteed to yield perfect policies in fully-specified tabular environments. We fi...
[ "graph neural networks", "value iteration", "implicit planning", "algorithmic bottleneck" ]
We study value iteration-based implicit planning methods, discover an algorithmic bottleneck which leaves them vulnerable in low-data scenarios. By performing value iteration-style planing in the latent space, we successfully break this bottleneck.
11,515
2110.05442
title_snapshot
1LLemKrsgQp
Learning in two-player zero-sum partially observable Markov games with perfect recall
https://openreview.net/forum?id=1LLemKrsgQp
[ "Tadashi Kozuno", "Pierre MENARD", "Remi Munos", "Michal Valko" ]
Poster
null
We study the problem of learning a Nash equilibrium (NE) in an extensive game with imperfect information (EGII) through self-play. Precisely, we focus on two-player, zero-sum, episodic, tabular EGII under the \textit{perfect-recall} assumption where the only feedback is realizations of the game (bandit feedback). In pa...
[ "online learning", "reinforcement learning", "Nash equilibrium", "Markov games" ]
Self-play online learning for two-player zero-sum, tabular, episodic, partially observable Markov games
11,512
2106.06279
title_judge
RUQ1zwZR8_
Differentially Private Learning with Adaptive Clipping
https://openreview.net/forum?id=RUQ1zwZR8_
[ "Galen Andrew", "Om Thakkar", "Hugh Brendan McMahan", "Swaroop Ramaswamy" ]
Poster
null
Existing approaches for training neural networks with user-level differential privacy (e.g., DP Federated Averaging) in federated learning (FL) settings involve bounding the contribution of each user's model update by {\em clipping} it to some constant value. However there is no good {\em a priori} setting of the clipp...
[ "Differential privacy", "federated learning", "privacy", "security", "federated" ]
In federated averaging with differential privacy, adapt the clipping norm to a (privately estimated) quantile of the update norm distribution, eliminating critical but difficult-to-estimate clipping norm hyperparameter.
11,498
1905.03871
title_snapshot
urueR03mkng
Leveraging the Inductive Bias of Large Language Models for Abstract Textual Reasoning
https://openreview.net/forum?id=urueR03mkng
[ "Christopher Michael Rytting", "David Wingate" ]
Poster
null
Large natural language models (LMs) (such as GPT-3 or T5) demonstrate impressive abilities across a range of general NLP tasks. Here, we show that the knowledge embedded in such models provides a useful inductive bias, not just on traditional NLP tasks, but also in the nontraditional task of training a symbolic reasoni...
[ "natural language processing", "reasoning", "transfer learning", "generalization" ]
We characterize the ability of connectionist pre-trained language models to generalize on symbolic reasoning tasks.
11,495
2110.02370
title_snapshot
O4TE57kehc1
Neural Circuit Synthesis from Specification Patterns
https://openreview.net/forum?id=O4TE57kehc1
[ "Frederik Schmitt", "Christopher Hahn", "Markus Norman Rabe", "Bernd Finkbeiner" ]
Poster
null
We train hierarchical Transformers on the task of synthesizing hardware circuits directly out of high-level logical specifications in linear-time temporal logic (LTL). The LTL synthesis problem is a well-known algorithmic challenge with a long history and an annual competition is organized to track the improvement of al...
[ "Transformer", "Temporal Logic", "Synthesis", "Circuits" ]
We train hierarchical Transformers on the task of synthesizing hardware circuits directly out of high-level logical specifications in linear-time temporal logic.
11,489
2107.11864
title_snapshot
YygA0yppTR
A Winning Hand: Compressing Deep Networks Can Improve Out-of-Distribution Robustness
https://openreview.net/forum?id=YygA0yppTR
[ "James Diffenderfer", "Brian R Bartoldson", "Shreya Chaganti", "Jize Zhang", "Bhavya Kailkhura" ]
Poster
null
Successful adoption of deep learning (DL) in the wild requires models to be: (1) compact, (2) accurate, and (3) robust to distributional shifts. Unfortunately, efforts towards simultaneously meeting these requirements have mostly been unsuccessful. This raises an important question: Is the inability to create Compact, ...
[ "robustness", "compression", "pruning", "binarization", "lottery-ticket hypothesis" ]
null
11,474
2106.09129
title_snapshot
iorEu783qJ5
Particle Cloud Generation with Message Passing Generative Adversarial Networks
https://openreview.net/forum?id=iorEu783qJ5
[ "Raghav Kansal", "Javier Duarte", "Hao Su", "Breno Orzari", "Thiago R F P Tomei", "Maurizio Pierini", "Mary Touranakou", "Jean-roch Vlimant", "Dimitrios Gunopulos" ]
Poster
null
In high energy physics (HEP), jets are collections of correlated particles produced ubiquitously in particle collisions such as those at the CERN Large Hadron Collider (LHC). Machine learning (ML)-based generative models, such as generative adversarial networks (GANs), have the potential to significantly accelerate LHC...
[ "generative models", "gans", "point clouds", "physics" ]
We publish a new point-cloud-based high energy physics dataset, and use it to develop a message-passing GAN (MPGAN) approach to simulate particle collisions.
11,473
2106.11535
title_snapshot
nNfj0pVn4Q
Interpolation can hurt robust generalization even when there is no noise
https://openreview.net/forum?id=nNfj0pVn4Q
[ "Konstantin Donhauser", "Alexandru Tifrea", "Michael Aerni", "Reinhard Heckel", "Fanny Yang" ]
Poster
null
Numerous recent works show that overparameterization implicitly reduces variance for min-norm interpolators and max-margin classifiers. These findings suggest that ridge regularization has vanishing benefits in high dimensions. We challenge this narrative by showing that, even in the absence of noise, avoiding interpo...
[ "regularization", "high dimensional statistics", "learning theory", "robustness" ]
We reveal unexpected benefits of regularization even in the overparameterized regime by proving that for both linear regression and classification, avoiding interpolation significantly improves generalization.
11,445
2108.02883
title_snapshot
cCQAzuT5q4
Online Selective Classification with Limited Feedback
https://openreview.net/forum?id=cCQAzuT5q4
[ "Aditya Gangrade", "Anil Kag", "Ashok Cutkosky", "Venkatesh Saligrama" ]
Spotlight
null
Motivated by applications to resource-limited and safety-critical domains, we study selective classification in the online learning model, wherein a predictor may abstain from classifying an instance. For example, this may model an adaptive decision to invoke more resources on this instance. Two salient aspects of the ...
[ "Selective Classification", "Online Learning" ]
null
11,413
2110.14243
title_snapshot
bXehDYUjjXi
A Variational Perspective on Diffusion-Based Generative Models and Score Matching
https://openreview.net/forum?id=bXehDYUjjXi
[ "Chin-Wei Huang", "Jae Hyun Lim", "Aaron Courville" ]
Spotlight
null
Discrete-time diffusion-based generative models and score matching methods have shown promising results in modeling high-dimensional image data. Recently, Song et al. (2021) show that diffusion processes that transform data into noise can be reversed via learning the score function, i.e. the gradient of the log-density...
[ "Diffusion model", "Neural SDE", "Neural ODE", "normalizing flows", "score matching", "hierarchical VAE", "generative modelling", "maximum likelihood", "variational lower bound" ]
We derived an ELBO for continuous-time diffusion models using stochastic calculus, and made connection to continuous-time normalizing flows, hierarchical VAE, and score-based generative models.
11,401
2106.02808
title_snapshot
sthiz9zeXGG
DRONE: Data-aware Low-rank Compression for Large NLP Models
https://openreview.net/forum?id=sthiz9zeXGG
[ "Patrick CHen", "Hsiang-Fu Yu", "Inderjit S Dhillon", "Cho-Jui Hsieh" ]
Poster
null
The representations learned by large-scale NLP models such as BERT have been widely used in various tasks. However, the increasing model size of the pre-trained models also brings efficiency challenges, including inference speed and model size when deploying models on mobile devices. Specifically, most operations in BE...
[ "Acceleration", "low-rank" ]
A generalized Low-rank method which leverages data distribution.
11,400
null
null
FyaSaEbNm1W
Demystifying and Generalizing BinaryConnect
https://openreview.net/forum?id=FyaSaEbNm1W
[ "Tim Dockhorn", "Yaoliang Yu", "Eyyub Sari", "Mahdi Zolnouri", "Vahid Partovi Nia" ]
Poster
null
BinaryConnect (BC) and its many variations have become the de facto standard for neural network quantization. However, our understanding of the inner workings of BC is still quite limited. We attempt to close this gap in four different aspects: (a) we show that existing quantization algorithms, including post-training ...
[ "neural network quantization", "binary connect", "generalized conditional gradient", "proximal connect" ]
We generalize BinaryConnect and prove convergence for the resulting algorithm.
11,393
2110.13220
title_snapshot
4XOrn_Y-dqp
The Benefits of Implicit Regularization from SGD in Least Squares Problems
https://openreview.net/forum?id=4XOrn_Y-dqp
[ "Difan Zou", "Jingfeng Wu", "Vladimir Braverman", "Quanquan Gu", "Dean Foster", "Sham M. Kakade" ]
Poster
null
Stochastic gradient descent (SGD) exhibits strong algorithmic regularization effects in practice, which has been hypothesized to play an important role in the generalization of modern machine learning approaches. In this work, we seek to understand these issues in the simpler setting of linear regression (including bot...
[ "stochastic gradient descent", "implicit regularization", "ridge regression", "overparameterization" ]
We compare the implicit regularization of SGD to the explicit regularization in ridge regression and show that SGD can be competitive for a broad class of problems.
11,392
2108.04552
title_snapshot
_kwj6V53ZqB
Grounding Representation Similarity Through Statistical Testing
https://openreview.net/forum?id=_kwj6V53ZqB
[ "Frances Ding", "Jean-Stanislas Denain", "Jacob Steinhardt" ]
Poster
null
To understand neural network behavior, recent works quantitatively compare different networks' learned representations using canonical correlation analysis (CCA), centered kernel alignment (CKA), and other dissimilarity measures. Unfortunately, these widely used measures often disagree on fundamental observations, such...
[ "representation similarity", "dissimilarity", "metric", "CKA", "CCA", "Orthogonal Procrustes", "benchmark", "representation learning", "probing", "deep networks" ]
We quantitatively benchmark dissimilarity measures for representations and find that Orthogonal Procrustes performs best.
11,390
2108.01661
title_judge
zdC5eXljMPy
Weighted model estimation for offline model-based reinforcement learning
https://openreview.net/forum?id=zdC5eXljMPy
[ "Toru Hishinuma", "Kei Senda" ]
Poster
null
This paper discusses model estimation in offline model-based reinforcement learning (MBRL), which is important for subsequent policy improvement using an estimated model. From the viewpoint of covariate shift, a natural idea is model estimation weighted by the ratio of the state-action distributions of offline data and...
[ "model-based reinforcement learning" ]
We present a weighted model estimation method for MBRL and show the validity.
11,385
null
null
RzYrn625bu8
VATT: Transformers for Multimodal Self-Supervised Learning from Raw Video, Audio and Text
https://openreview.net/forum?id=RzYrn625bu8
[ "Hassan Akbari", "Liangzhe Yuan", "Rui Qian", "Wei-Hong Chuang", "Shih-Fu Chang", "Yin Cui", "Boqing Gong" ]
Poster
null
We present a framework for learning multimodal representations from unlabeled data using convolution-free Transformer architectures. Specifically, our Video-Audio-Text Transformer (VATT) takes raw signals as inputs and extracts multimodal representations that are rich enough to benefit a variety of downstream tasks. We...
[ "Self-supervised Learning", "Multimodal Understanding", "Transformer", "Contrastive Learning", "Video Recognition", "Audio Recognition" ]
A pure Transformer-based pipeline for learning semantic representations from raw video, audio, and text without supervision
11,382
2104.11178
title_snapshot
LBhruMnhgIB
Unintended Selection: Persistent Qualification Rate Disparities and Interventions
https://openreview.net/forum?id=LBhruMnhgIB
[ "Reilly Raab", "Yang Liu" ]
Spotlight
null
Realistically---and equitably---modeling the dynamics of group-level disparities in machine learning remains an open problem. In particular, we desire models that do not suppose inherent differences between artificial groups of people---but rather endogenize disparities by appeal to unequal initial conditions of insula...
[ "Fairness", "Evolution", "Qualification Rate Disparity", "Intervention", "Replicator Dynamics", "Feedback Control" ]
Careless deployment of machine learning classifiers can induce social changes that maintain existing disparities between structurally equivalent groups.
11,361
2111.01201
title_snapshot
_n59kgzSFef
Circa: Stochastic ReLUs for Private Deep Learning
https://openreview.net/forum?id=_n59kgzSFef
[ "Zahra Ghodsi", "Nandan Kumar Jha", "Brandon Reagen", "Siddharth Garg" ]
Poster
null
The simultaneous rise of machine learning as a service and concerns over user privacy have increasingly motivated the need for private inference (PI). While recent work demonstrates PI is possible using cryptographic primitives, the computational overheads render it impractical. State-of-art deep networks are inadequat...
[ "privacy preserving machine learning", "private inference", "stochastic ReLU" ]
null
11,347
2106.08475
title_snapshot
uY-XMIbyXec
Shift-Robust GNNs: Overcoming the Limitations of Localized Graph Training data
https://openreview.net/forum?id=uY-XMIbyXec
[ "Qi Zhu", "Natalia Ponomareva", "Jiawei Han", "Bryan Perozzi" ]
Poster
null
There has been a recent surge of interest in designing Graph Neural Networks (GNNs) for semi-supervised learning tasks. Unfortunately this work has assumed that the nodes labeled for use in training were selected uniformly at random (i.e. are an IID sample). However in many real world scenarios gathering labels for gra...
[ "graph semi-supervised learning", "distributional shifts", "transfer learning" ]
In semi-supervised graph learning, gathering labels uniformly at random can be a great challenge. We present Shift-Robust GNN to account for distributional differences between biased training data and the graph's true inference distribution.
11,339
2108.01099
title_snapshot
HnLDt9v6Q-j
Two steps to risk sensitivity
https://openreview.net/forum?id=HnLDt9v6Q-j
[ "Christopher Gagne", "Peter Dayan" ]
Spotlight
null
Distributional reinforcement learning (RL) – in which agents learn about all the possible long-term consequences of their actions, and not just the expected value – is of great recent interest. One of the most important affordances of a distributional view is facilitating a modern, measured, approach to risk when outco...
[ "Risk measures", "Conditional value-at-risk", "Two-step task", "Time-consistency", "Distributional reinforcement learning", "Anxiety" ]
We use a conditional value-at-risk (CVaR) measure to show that many human subjects are highly risk averse in the popular two-step task; and use simulations in a novel domain to suggest how to distinguish time-consistent/inconsistent CVaRiants.
11,333
2111.06803
title_snapshot
0fPgXqP1Mq
A unified framework for bandit multiple testing
https://openreview.net/forum?id=0fPgXqP1Mq
[ "Ziyu Xu", "Ruodu Wang", "Aaditya Ramdas" ]
Poster
null
In bandit multiple hypothesis testing, each arm corresponds to a different null hypothesis that we wish to test, and the goal is to design adaptive algorithms that correctly identify large set of interesting arms (true discoveries), while only mistakenly identifying a few uninteresting ones (false discoveries). One com...
[ "bandits", "multiple testing" ]
A unified framework that ensures false discovery rate (FDR) control with minimal assumptions for the multiple testing problem in bandit settings using recent advances in e-variables and e-processes (alternatives to p-values and p-processes).
11,308
2107.07322
title_snapshot
2j_cut38wv
Backdoor Attack with Imperceptible Input and Latent Modification
https://openreview.net/forum?id=2j_cut38wv
[ "Khoa Doan", "Yingjie Lao", "Ping Li" ]
Poster
null
Recent studies have shown that deep neural networks (DNN) are vulnerable to various adversarial attacks. In particular, an adversary can inject a stealthy backdoor into a model such that the compromised model will behave normally without the presence of the trigger. Techniques for generating backdoor images that are vi...
[ "backdoor attacks", "generative models", "wasserstein distance" ]
Backdoor Attack with Imperceptible Input and Latent Modification
11,297
null
null
X2Cxixkcpx
Structured Reordering for Modeling Latent Alignments in Sequence Transduction
https://openreview.net/forum?id=X2Cxixkcpx
[ "bailin wang", "Mirella Lapata", "Ivan Titov" ]
Poster
null
Despite success in many domains, neural models struggle in settings where train and test examples are drawn from different distributions. In particular, in contrast to humans, conventional sequence-to-sequence (seq2seq) models fail to generalize systematically, i.e., interpret sentences representing novel combinations ...
[ "language understanding", "sequence models", "latent variable models", "compositional generalization", "systematic generalization" ]
A general seq2seq model with latent discrete alignments via separable permutations.
11,280
2106.03257
title_snapshot
bJz3cFePTna
Mixture Proportion Estimation and PU Learning:A Modern Approach
https://openreview.net/forum?id=bJz3cFePTna
[ "Saurabh Garg", "Yifan Wu", "Alex Smola", "Sivaraman Balakrishnan", "Zachary Chase Lipton" ]
Spotlight
null
Given only positive examples and unlabeled examples (from both positive and negative classes), we might hope nevertheless to estimate an accurate positive-versus-negative classifier. Formally, this task is broken down into two subtasks: (i) Mixture Proportion Estimation (MPE)---determining the fraction of positive exam...
[ "PU learning", "Mixture Proportion Estimation" ]
Given only Positive (P) and Unlabeled (U) data, containing both P and Negative (N) samples, we propose new approaches to estimate fraction of P in U and learn P vs N classifier.
11,279
2111.00980
title_snapshot
rYhBGWYm6AU
Intriguing Properties of Contrastive Losses
https://openreview.net/forum?id=rYhBGWYm6AU
[ "Ting Chen", "Calvin Luo", "Lala Li" ]
Poster
null
We study three intriguing properties of contrastive learning. First, we generalize the standard contrastive loss to a broader family of losses, and we find that various instantiations of the generalized loss perform similarly under the presence of a multi-layer non-linear projection head. Second, we study if instance-b...
[ "contrastive learning", "contrastive loss", "feature suppression", "self-supervised learning", "computer vision" ]
We study three intriguing properties of contrastive loss. In particular, we highlight that feature suppression poses an open challenge to contrastive learning research.
11,269
2011.02803
title_snapshot
GOnkx08Gm6
A/B Testing for Recommender Systems in a Two-sided Marketplace
https://openreview.net/forum?id=GOnkx08Gm6
[ "Preetam Nandy", "Divya Venugopalan", "Chun Lo", "Shaunak Chatterjee" ]
Poster
null
Two-sided marketplaces are standard business models of many online platforms (e.g., Amazon, Facebook, LinkedIn), wherein the platforms have consumers, buyers or content viewers on one side and producers, sellers or content-creators on the other. Consumer side measurement of the impact of a treatment variant can be done...
[ "A/B Testing", "Two-sided Marketplace", "Treatment Effect Estimation", "Recommender Systems" ]
We propose a novel A/B testing framework for seller or producer side measurements in a two-sided marketplace.
11,235
2106.00762
title_snapshot
VUJlv99HgAZ
Surrogate Regret Bounds for Polyhedral Losses
https://openreview.net/forum?id=VUJlv99HgAZ
[ "Rafael Frongillo", "Bo Waggoner" ]
Poster
null
Surrogate risk minimization is an ubiquitous paradigm in supervised machine learning, wherein a target problem is solved by minimizing a surrogate loss on a dataset. Surrogate regret bounds, also called excess risk bounds, are a common tool to prove generalization rates for surrogate risk minimization. While surrogat...
[ "surrogate regret bounds", "excess risk bounds", "polyhedral losses", "calibration", "property elicitation" ]
Any polyhedral surrogate loss achieves a linear surrogate regret bound, while "non-polyhedral" losses do not.
11,230
2110.14031
title_snapshot
Kug2s3rHiG3
Complexity Lower Bounds for Nonconvex-Strongly-Concave Min-Max Optimization
https://openreview.net/forum?id=Kug2s3rHiG3
[ "Haochuan Li", "Yi Tian", "Jingzhao Zhang", "Ali Jadbabaie" ]
Poster
null
We provide a first-order oracle complexity lower bound for finding stationary points of min-max optimization problems where the objective function is smooth, nonconvex in the minimization variable, and strongly concave in the maximization variable. We establish a lower bound of $\Omega\left(\sqrt{\kappa}\epsilon^{-2}\r...
[ "min-max optimization", "lower bound", "oracle complexity" ]
We provide a nearly optimal first-order oracle complexity lower bound for nonconvex-strongly-concave min-max optimization problems.
11,221
2104.08708
title_snapshot
_6DawVPqyl
Hard-Attention for Scalable Image Classification
https://openreview.net/forum?id=_6DawVPqyl
[ "Athanasios Papadopoulos", "Pawel Korus", "Nasir Memon" ]
Poster
null
Can we leverage high-resolution information without the unsustainable quadratic complexity to input scale? We propose Traversal Network (TNet), a novel multi-scale hard-attention architecture, which traverses image scale-space in a top-down fashion, visiting only the most informative image regions along the way. TNet o...
[ "classification", "hard-attention", "multi-scale", "scalability", "high-resolution", "interpretability" ]
We propose TNet, a novel multi-scale hard-attention architecture, in order to leverage high-resolution information without the unsustainable quadratic complexity to input scale.
11,193
2102.10212
title_snapshot
GEKTIKvslP
Fair Exploration via Axiomatic Bargaining
https://openreview.net/forum?id=GEKTIKvslP
[ "Jackie Baek", "Vivek Farias" ]
Spotlight
null
Motivated by the consideration of fairly sharing the cost of exploration between multiple groups in learning problems, we develop the Nash bargaining solution in the context of multi-armed bandits. Specifically, the 'grouped' bandit associated with any multi-armed bandit problem associates, with each time step, a singl...
[ "bandits", "fairness", "exploration", "Nash bargaining" ]
We study how to fairly allocate the burden of exploration for multi-armed bandits with groups using the Nash bargaining framework.
11,188
2106.02553
title_snapshot
9BvDIW6_qxZ
Is Bang-Bang Control All You Need? Solving Continuous Control with Bernoulli Policies
https://openreview.net/forum?id=9BvDIW6_qxZ
[ "Tim Seyde", "Igor Gilitschenski", "Wilko Schwarting", "Bartolomeo Stellato", "Martin Riedmiller", "Markus Wulfmeier", "Daniela Rus" ]
Poster
null
Reinforcement learning (RL) for continuous control typically employs distributions whose support covers the entire action space. In this work, we investigate the colloquially known phenomenon that trained agents often prefer actions at the boundaries of that space. We draw theoretical connections to the emergence of ba...
[ "reinforcement learning", "optimal control", "robotics" ]
Bernoulli policies can achieve state-of-the-art performance on several continuous control benchmarks and we take a closer look at why and how.
11,181
2111.02552
title_snapshot
Xa9Ba6NsJ6
On the Value of Interaction and Function Approximation in Imitation Learning
https://openreview.net/forum?id=Xa9Ba6NsJ6
[ "Nived Rajaraman", "Yanjun Han", "Lin Yang", "Jingbo Liu", "Jiantao Jiao", "Kannan Ramchandran" ]
Poster
null
We study the statistical guarantees for the Imitation Learning (IL) problem in episodic MDPs. Rajaraman et al. (2020) show an information theoretic lower bound that in the worst case, a learner which can even actively query the expert policy suffers from a suboptimality growing quadratically in the length of the horizo...
[ "Reinforcement Learning", "Imitation Learning", "Minimax rates" ]
Provable and optimal statistical rates for Imitation Learning with (i) an interactive expert, and (ii) under linear function approximation
11,180
null
null
hGmrNwR8qQP
The Causal-Neural Connection: Expressiveness, Learnability, and Inference
https://openreview.net/forum?id=hGmrNwR8qQP
[ "Kevin Muyuan Xia", "Kai-Zhan Lee", "Yoshua Bengio", "Elias Bareinboim" ]
Poster
null
One of the central elements of any causal inference is an object called structural causal model (SCM), which represents a collection of mechanisms and exogenous sources of random variation of the system under investigation (Pearl, 2000). An important property of many kinds of neural networks is universal approximabilit...
[ "causal inference", "deep learning", "neural models", "causal identification", "causal estimation" ]
We introduce the neural causal model (NCM), a type of structural causal model (SCM) composed of neural networks, which can solve the problems of causal effect identification and estimation given a causal diagram as an inductive bias.
11,179
2107.00793
title_snapshot
x1Lp2bOlVIo
Diffusion Normalizing Flow
https://openreview.net/forum?id=x1Lp2bOlVIo
[ "Qinsheng Zhang", "Yongxin Chen" ]
Poster
null
We present a novel generative modeling method called diffusion normalizing flow based on stochastic differential equations (SDEs). The algorithm consists of two neural SDEs: a forward SDE that gradually adds noise to the data to transform the data into Gaussian random noise, and a backward SDE that gradually removes th...
[ "normalizing flow", "diffusion probabilistic models", "density estimation", "generative models" ]
Diffusion models with learnable drift
11,175
2110.07579
title_snapshot
Fj6kQJbHwM9
Manifold Topology Divergence: a Framework for Comparing Data Manifolds.
https://openreview.net/forum?id=Fj6kQJbHwM9
[ "Serguei Barannikov", "Ilya Trofimov", "Grigorii Sotnikov", "Ekaterina Trimbach", "Alexander Korotin", "Alexander Filippov", "Evgeny Burnaev" ]
Poster
null
We propose a framework for comparing data manifolds, aimed, in particular, towards the evaluation of deep generative models. We describe a novel tool, Cross-Barcode(P,Q), that, given a pair of distributions in a high-dimensional space, tracks multiscale topology spacial discrepancies between manifolds on which the dist...
[ "data manifolds", "point clouds", "persistent homology", "topology", "generative models", "generative adversarial networks", "mode-dropping", "3D-shapes", "time-series" ]
We introduce a topology-based domain agnostic methodology for comparing data manifolds.
11,171
2106.04024
title_snapshot
nWz-Si-uTzt
Automated Discovery of Adaptive Attacks on Adversarial Defenses
https://openreview.net/forum?id=nWz-Si-uTzt
[ "Chengyuan Yao", "Pavol Bielik", "PETAR TSANKOV", "Martin Vechev" ]
Poster
null
Reliable evaluation of adversarial defenses is a challenging task, currently limited to an expert who manually crafts attacks that exploit the defense’s inner workings, or to approaches based on ensemble of fixed attacks, none of which may be effective for the specific defense at hand. Our key observation is that adapt...
[ "Deep Learning", "Adversarial Attacks", "Adversarial Defences", "Robustness" ]
null
11,166
2102.11860
title_snapshot
HCOdL3dWab
Inverse Problems Leveraging Pre-trained Contrastive Representations
https://openreview.net/forum?id=HCOdL3dWab
[ "Sriram Ravula", "Georgios Smyrnis", "Matt Jordan", "Alex Dimakis" ]
Poster
null
We study a new family of inverse problems for recovering representations of corrupted data. We assume access to a pre-trained representation learning network R(x) that operates on clean images, like CLIP. The problem is to recover the representation of an image R(x), if we are only given a corrupted version A(x), for s...
[ "Inverse Problems", "Representation Learning", "Contrastive Learning", "Robustness" ]
We obtain representations for highly corrupted images by using a supervised inversion method, which leverages contrastive learning.
11,153
2110.07439
title_snapshot
SwfsoPuGYku
Robust Implicit Networks via Non-Euclidean Contractions
https://openreview.net/forum?id=SwfsoPuGYku
[ "Saber Jafarpour", "Alexander Davydov", "Anton Proskurnikov", "Francesco Bullo" ]
Poster
null
Implicit neural networks, a.k.a., deep equilibrium networks, are a class of implicit-depth learning models where function evaluation is performed by solving a fixed point equation. They generalize classic feedforward models and are equivalent to infinite-depth weight-tied feedforward networks. While implicit models sho...
[ "Implicit deep learning", "Contraction theory", "Robust neural networks." ]
We design a framework based on non-Euclidean contraction theory to study the well-posedness and stability of implicit neural networks.
11,149
2106.03194
title_snapshot
KJ5h-yfUHa
Attention Bottlenecks for Multimodal Fusion
https://openreview.net/forum?id=KJ5h-yfUHa
[ "Arsha Nagrani", "Shan Yang", "Anurag Arnab", "Aren Jansen", "Cordelia Schmid", "Chen Sun" ]
Poster
null
Humans perceive the world by concurrently processing and fusing high-dimensional inputs from multiple modalities such as vision and audio. Machine perception models, in stark contrast, are typically modality-specific and optimised for unimodal benchmarks. A common approach for building multimodal models is to simply c...
[ "multimodal", "fusion", "attention", "audiovisual", "transformers", "video" ]
We propose a new multimodal fusion model for video that exchanges cross-modal information between modalities via a small number of 'attention bottleneck' latents, achieving state of the art results for video classification.
11,137
2107.00135
title_snapshot
omDF-uQ_OZ
Identification of the Generalized Condorcet Winner in Multi-dueling Bandits
https://openreview.net/forum?id=omDF-uQ_OZ
[ "Björn Haddenhorst", "Viktor Bengs", "Eyke Hüllermeier" ]
Poster
null
The reliable identification of the “best” arm while keeping the sample complexity as low as possible is a common task in the field of multi-armed bandits. In the multi-dueling variant of multi-armed bandits, where feedback is provided in the form of a winning arm among as set of k chosen ones, a reasonable notion of be...
[ "Best arm identification", "Condorcet Winner", "Dvoretzky–Kiefer–Wolfowitz inequality", "Multi-armed Bandits", "Preference Learning" ]
We prove (up to logarithmic terms) optimal upper and lower bounds on the sample complexity for finding with high confidence the generalized Condorcet winner in the preference-based multi-armed bandits setting
11,132
null
null
EHUsTBGIP17
Explicit loss asymptotics in the gradient descent training of neural networks
https://openreview.net/forum?id=EHUsTBGIP17
[ "Maksim Velikanov", "Dmitry Yarotsky" ]
Poster
null
Current theoretical results on optimization trajectories of neural networks trained by gradient descent typically have the form of rigorous but potentially loose bounds on the loss values. In the present work we take a different approach and show that the learning trajectory of a wide network in a lazy training regime ...
[ "Deep learning theory", "neural networks", "gradient descent", "Neural Tangent Kernel", "spectral theory" ]
We derive explicit loss asymptotics for gradient descent training of neural networks
11,130
null
null
Mobm1AGs64v
Fast Training Method for Stochastic Compositional Optimization Problems
https://openreview.net/forum?id=Mobm1AGs64v
[ "Hongchang Gao", "Heng Huang" ]
Poster
null
The stochastic compositional optimization problem covers a wide range of machine learning models, such as sparse additive models and model-agnostic meta-learning. Thus, it is necessary to develop efficient methods for its optimization. Existing methods for the stochastic compositional optimization problem only focus o...
[ "Decentralized optimization", "stochastic compositional optimization problem", "model-agnostic meta-learning" ]
Decentralized training methods for stochastic compositional optimization problems
11,123
null
null
SjxC07jABZ4
Bubblewrap: Online tiling and real-time flow prediction on neural manifolds
https://openreview.net/forum?id=SjxC07jABZ4
[ "Anne Draelos", "Pranjal Gupta", "Na Young Jun", "Chaichontat Sriworarat", "John Pearson" ]
Poster
null
While most classic studies of function in experimental neuroscience have focused on the coding properties of individual neurons, recent developments in recording technologies have resulted in an increasing emphasis on the dynamics of neural populations. This has given rise to a wide variety of models for analyzing popu...
[ "neuroscience", "neural populations", "online", "real-time" ]
An online learning approach based on modeling probability flow along a tiled manifold accurately predicts population dynamics in neuroscience experiments.
11,117
2108.13941
title_snapshot
VOjwYOfGZcL
Adversarially robust learning for security-constrained optimal power flow
https://openreview.net/forum?id=VOjwYOfGZcL
[ "Priya L. Donti", "Aayushya Agarwal", "Neeraj Vijay Bedmutha", "Larry Pileggi", "J Zico Kolter" ]
Poster
null
In recent years, the ML community has seen surges of interest in both adversarially robust learning and implicit layers, but connections between these two areas have seldom been explored. In this work, we combine innovations from these areas to tackle the problem of N-k security-constrained optimal power flow (SCOPF). ...
[ "security-constrained optimal power flow", "adversarial robustness", "implicit layers", "bi-level optimization" ]
We address the problem of N-k security-constrained optimal power flow (SCOPF) via methods from adversarially robust training.
11,105
2111.06961
title_snapshot
-8QSntMuqBV
The Many Faces of Adversarial Risk
https://openreview.net/forum?id=-8QSntMuqBV
[ "Muni Sreenivas Pydi", "Varun Jog" ]
Poster
null
Adversarial risk quantifies the performance of classifiers on adversarially perturbed data. Numerous definitions of adversarial risk---not all mathematically rigorous and differing subtly in the details---have appeared in the literature. In this paper, we revisit these definitions, make them rigorous, and critically ex...
[ "Adversarial risk", "robustness", "optimal transport", "game theory", "Nash equilibrium", "2-alternating capacities", "Strassen's theorem", "measurable selection theorem", "Wasserstein distance", "robust statistics" ]
We study adversarial risk from the viewpoints of optimal transport, 2-alternating capacities and game theory, extending some recent results and revealing new connections.
11,102
2201.08956
title_snapshot
uXc42E9ZPFs
Language models enable zero-shot prediction of the effects of mutations on protein function
https://openreview.net/forum?id=uXc42E9ZPFs
[ "Joshua Meier", "Roshan Rao", "Robert Verkuil", "Jason Liu", "Tom Sercu", "Alexander Rives" ]
Poster
null
Modeling the effect of sequence variation on function is a fundamental problem for understanding and designing proteins. Since evolution encodes information about function into patterns in protein sequences, unsupervised models of variant effects can be learned from sequence data. The approach to date has been to fit a...
[ "Proteins", "language modeling", "generative biology", "zero-shot learning", "unsupervised learning", "variant prediction" ]
Using zero-shot inference, language models capture the effect of mutations on protein function, performing at state-of-the-art.
11,072
null
null
gbtDcLzwKUb
BCD Nets: Scalable Variational Approaches for Bayesian Causal Discovery
https://openreview.net/forum?id=gbtDcLzwKUb
[ "Chris Cundy", "Aditya Grover", "Stefano Ermon" ]
Poster
null
A structural equation model (SEM) is an effective framework to reason over causal relationships represented via a directed acyclic graph (DAG). Recent advances have enabled effective maximum-likelihood point estimation of DAGs from observational data. However, a point estimate may not accurately capture the uncertaint...
[ "Bayesian", "DAG", "Variational Inference", "Causal Discovery" ]
We design a variational method to learn distributions over linear structural equation parameters from data, unlike previous point estimate methods.
11,064
2112.02761
title_snapshot
RgH0gGH9B64
Joint inference and input optimization in equilibrium networks
https://openreview.net/forum?id=RgH0gGH9B64
[ "Swaminathan Gurumurthy", "Shaojie Bai", "Zachary Manchester", "J Zico Kolter" ]
Poster
null
Many tasks in deep learning involve optimizing over the inputs to a network to minimize or maximize some objective; examples include optimization over latent spaces in a generative model to match a target image, or adversarially perturbing an input to worsen classifier performance. Performing such optimization, howeve...
[ "Deep Equilibrium Models", "Bi-level Optimization", "Generative Models", "Inverse Problems", "Adversarial Training" ]
Jointly optimizing over the input and output for implicit models with applications to problems like generative modelling, inverse problems, adversarial examples and gradient-based meta-learning
11,024
2111.13236
title_snapshot
-AV3AKwgiG
REMIPS: Physically Consistent 3D Reconstruction of Multiple Interacting People under Weak Supervision
https://openreview.net/forum?id=-AV3AKwgiG
[ "Mihai Fieraru", "Mihai Zanfir", "Teodor Alexandru Szente", "Eduard Gabriel Bazavan", "Vlad Olaru", "Cristian Sminchisescu" ]
Poster
null
The three-dimensional reconstruction of multiple interacting humans given a monocular image is crucial for the general task of scene understanding, as capturing the subtleties of interaction is often the very reason for taking a picture. Current 3D human reconstruction methods either treat each person independently, ig...
[ "human", "pose", "shape", "reconstruction", "3d", "people", "transformer", "mesh" ]
null
11,010
null
null
zzdf0CirJM4
Batch Active Learning at Scale
https://openreview.net/forum?id=zzdf0CirJM4
[ "Gui Citovsky", "Giulia DeSalvo", "Claudio Gentile", "Lazaros Karydas", "Anand Rajagopalan", "Afshin Rostamizadeh", "Sanjiv Kumar" ]
Poster
null
The ability to train complex and highly effective models often requires an abundance of training data, which can easily become a bottleneck in cost, time, and computational resources. Batch active learning, which adaptively issues batched queries to a labeling oracle, is a common approach for addressing this problem. T...
[ "active learning", "large scale", "batch active learning" ]
We develop a batch active learning approach that is effective for very large batch sizes (100K-1M).
11,001
2107.14263
title_snapshot
jYzSTzvDP3p
Neural Pseudo-Label Optimism for the Bank Loan Problem
https://openreview.net/forum?id=jYzSTzvDP3p
[ "Aldo Pacchiano", "Shaun Singh", "Edward Chou", "Alexander C. Berg", "Jakob Nicolaus Foerster" ]
Poster
null
We study a class of classification problems best exemplified by the \emph{bank loan} problem, where a lender decides whether or not to issue a loan. The lender only observes whether a customer will repay a loan if the loan is issued to begin with, and thus modeled decisions affect what data is available to the lender f...
[ "neural bandits", "function approximation", "classification", "bank loan problem" ]
We propose a tractable algorithm for neural contextual bandits in the setting of online classification.
11,000
2112.02185
title_snapshot
jV5m8NAWb0E
How Tight Can PAC-Bayes be in the Small Data Regime?
https://openreview.net/forum?id=jV5m8NAWb0E
[ "Andrew Y. K. Foong", "Wessel Bruinsma", "David R. Burt", "Richard E Turner" ]
Poster
null
In this paper, we investigate the question: _Given a small number of datapoints, for example $N = 30$, how tight can PAC-Bayes and test set bounds be made?_ For such small datasets, test set bounds adversely affect generalisation performance by withholding data from the training procedure. In this setting, PAC-Bayes bo...
[ "PAC-Bayes", "learning theory", "generalization" ]
We investigate how tight a standard proof of PAC-Bayes theorems can be made.
10,996
2106.03542
title_snapshot
PFBHMlpaWY
General Nonlinearities in SO(2)-Equivariant CNNs
https://openreview.net/forum?id=PFBHMlpaWY
[ "Daniel Franzen", "Michael Wand" ]
Poster
null
Invariance under symmetry is an important problem in machine learning. Our paper looks specifically at equivariant neural networks where transformations of inputs yield homomorphic transformations of outputs. Here, steerable CNNs have emerged as the standard solution. An inherent problem of steerable representations is...
[ "deep learning", "equivariance", "steerable CNNs", "group convolution", "harmonic distortion analysis" ]
We improve the ability of using (more) general nonlinerities in SO(2)-equivariant steerable networks.
10,994
null
null
yaksQCYcRs
Neural Program Generation Modulo Static Analysis
https://openreview.net/forum?id=yaksQCYcRs
[ "Rohan Mukherjee", "Yeming Wen", "Dipak Chaudhari", "Thomas Reps", "Swarat Chaudhuri", "Chris Jermaine" ]
Spotlight
null
State-of-the-art neural models of source code tend to be evaluated on the generation of individual expressions and lines of code, and commonly fail on long-horizon tasks such as the generation of entire method bodies. We propose to address this deficiency using weak supervision from a static program analyzer. Our neuro...
[ "Program generation", "neurosymbolic learning", "attribute grammars", "program synthesis" ]
We proposed to tackle the long-horizon code generation challenge using weak supervision from a static program analyzer.
10,989
2111.01633
title_snapshot
CeByDMy0YTL
Environment Generation for Zero-Shot Compositional Reinforcement Learning
https://openreview.net/forum?id=CeByDMy0YTL
[ "Izzeddin Gur", "Natasha Jaques", "Yingjie Miao", "Jongwook Choi", "Manoj Tiwari", "Honglak Lee", "Aleksandra Faust" ]
Poster
null
Many real-world problems are compositional – solving them requires completing interdependent sub-tasks, either in series or in parallel, that can be represented as a dependency graph. Deep reinforcement learning (RL) agents often struggle to learn such complex tasks due to the long time horizons and sparse rewards. To ...
[ "Adversarial Environment Generation", "Compositional Reinforcement Learning", "Minimax Regret Adversary", "Auto Curriculum" ]
We propose an adversarial environment generation method for zero-shot compositional generalization in reinforcement learning.
10,980
2201.08896
title_snapshot
Pkzvd9ONEPr
Combining Human Predictions with Model Probabilities via Confusion Matrices and Calibration
https://openreview.net/forum?id=Pkzvd9ONEPr
[ "Gavin Kerrigan", "Padhraic Smyth", "Mark Steyvers" ]
Poster
null
An increasingly common use case for machine learning models is augmenting the abilities of human decision makers. For classification tasks where neither the human nor model are perfectly accurate, a key step in obtaining high performance is combining their individual predictions in a manner that leverages their relativ...
[ "human-machine", "human-AI", "human-in-the-loop", "calibration", "classification" ]
We combine the class-level output of a human with the probabilistic output of a classifier in order to achieve low misclassification rates.
10,972
2109.14591
title_snapshot
oZg-aOyHL-h
Efficiently Learning One Hidden Layer ReLU Networks From Queries
https://openreview.net/forum?id=oZg-aOyHL-h
[ "Sitan Chen", "Adam Klivans", "Raghu Meka" ]
Poster
null
While the problem of PAC learning neural networks from samples has received considerable attention in recent years, in certain settings like model extraction attacks, it is reasonable to imagine having more than just the ability to observe random labeled examples. Motivated by this, we consider the following problem: g...
[ "PAC learning", "polynomial-time algorithms", "neural networks", "query learning", "model extraction" ]
We give the first provable, polynomial-time algorithm for learning two-layer neural networks from queries.
10,967
2111.04727
title_judge
Nb03vOtUfz
K-level Reasoning for Zero-Shot Coordination in Hanabi
https://openreview.net/forum?id=Nb03vOtUfz
[ "Brandon Cui", "Hengyuan Hu", "Luis Pineda", "Jakob Nicolaus Foerster" ]
Poster
null
The standard problem setting in cooperative multi-agent settings is \emph{self-play} (SP), where the goal is to train a \emph{team} of agents that works well together. However, optimal SP policies commonly contain arbitrary conventions (``handshakes'') and are not compatible with other, independently trained ...
[ "Multi-Agent Reinforcement Learning", "Reinforcement Learning", "Zero-Shot Coordination", "Deep Reinforcement Learning" ]
With a simple engineering optimization, jointly training all levels of a K-Level Reasoning Hierarchy, we are able to stabilize and improve Zero-Shot Coordination results in Hanabi.
10,963
2207.07166
title_snapshot
oyHWvdvkZDv
Efficient Truncated Linear Regression with Unknown Noise Variance
https://openreview.net/forum?id=oyHWvdvkZDv
[ "Constantinos Costis Daskalakis", "Patroklos Stefanou", "Rui Yao", "Emmanouil Zampetakis" ]
Poster
null
Truncated linear regression is a classical challenge in Statistics, wherein a label, $y = w^T x + \varepsilon$, and its corresponding feature vector, $x \in \mathbb{R}^k$, are only observed if the label falls in some subset $S \subseteq \mathbb{R}$; otherwise the existence of the pair $(x, y)$ is hidden from observatio...
[ "regression", "theory of computation", "truncated statistics", "linear regression", "truncation bias", "stochastic gradient descent", "asymptotic normality" ]
null
10,957
2208.12042
title_snapshot
qL_juuU4P3Y
Partition and Code: learning how to compress graphs
https://openreview.net/forum?id=qL_juuU4P3Y
[ "Giorgos Bouritsas", "Andreas Loukas", "Nikolaos Karalias", "Michael M. Bronstein" ]
Poster
null
Can we use machine learning to compress graph data? The absence of ordering in graphs poses a significant challenge to conventional compression algorithms, limiting their attainable gains as well as their ability to discover relevant patterns. On the other hand, most graph compression approaches rely on domain-dependen...
[ "lossless graph compression", "neural compression", "graph neural networks" ]
We introduce a flexible, end-to-end machine learning framework for lossless graph compression based on graph partitioning, dictionary learning and entropy coding
10,921
2107.01952
title_snapshot
XL9DWRG7mJn
Rethinking gradient sparsification as total error minimization
https://openreview.net/forum?id=XL9DWRG7mJn
[ "Atal Narayan Sahu", "Aritra Dutta", "Ahmed M. Abdelmoniem", "Trambak Banerjee", "Marco Canini", "Panos Kalnis" ]
Spotlight
null
Gradient compression is a widely-established remedy to tackle the communication bottleneck in distributed training of large deep neural networks (DNNs). Under the error-feedback framework, Top-$k$ sparsification, sometimes with $k$ as little as 0.1% of the gradient size, enables training to the same model quality as th...
[ "Gradient compression", "Distributed optimization" ]
null
10,919
2108.00951
title_snapshot
gjBz22V93a
Fast Multi-Resolution Transformer Fine-tuning for Extreme Multi-label Text Classification
https://openreview.net/forum?id=gjBz22V93a
[ "Jiong Zhang", "Wei-Cheng Chang", "Hsiang-Fu Yu", "Inderjit S Dhillon" ]
Poster
null
Extreme multi-label text classification~(XMC) seeks to find relevant labels from an extreme large label collection for a given text input. Many real-world applications can be formulated as XMC problems, such as recommendation systems, document tagging and semantic search. Recently, transformer based XMC methods, such a...
[ "transformers", "extreme multi-label text classification" ]
We propose XR-Transformer, a novel extreme multi-label text classification (XMC) model that fine-tunes transformer on multi-resolution tasks and establish SOTA result on XMC benchmarks with significant less training time than competing methods.
10,898
2110.00685
title_snapshot
lI2To0NGe3Q
Regularized Frank-Wolfe for Dense CRFs: Generalizing Mean Field and Beyond
https://openreview.net/forum?id=lI2To0NGe3Q
[ "D. Khuê Lê-Huu", "Karteek Alahari" ]
Poster
null
We introduce regularized Frank-Wolfe, a general and effective algorithm for inference and learning of dense conditional random fields (CRFs). The algorithm optimizes a nonconvex continuous relaxation of the CRF inference problem using vanilla Frank-Wolfe with approximate updates, which are equivalent to minimizing a re...
[ "mean field", "frank-wolfe", "conditional gradient", "crfs", "mrfs", "conditional random fields", "markov random fields", "map inference", "semantic segementation" ]
This paper proposes a new class of CRF inference algorithms called Regularized Frank-Wolfe that includes existing algorithms, such as mean field or concave-convex procedure, as special cases.
10,894
2110.14759
title_snapshot
xmx5rE9QP7R
You Are the Best Reviewer of Your Own Papers: An Owner-Assisted Scoring Mechanism
https://openreview.net/forum?id=xmx5rE9QP7R
[ "Weijie J Su" ]
Poster
null
I consider the setting where reviewers offer very noisy scores for a number of items for the selection of high-quality ones (e.g., peer review of large conference proceedings) whereas the owner of these items knows the true underlying scores but prefers not to provide this information. To address this withholding of in...
[ "Peer review", "mechanism design", "ranking", "isotonic regression", "utility", "convex optimization" ]
null
10,875
2110.14802
title_snapshot
DbxKZvfOIhu
Beyond BatchNorm: Towards a Unified Understanding of Normalization in Deep Learning
https://openreview.net/forum?id=DbxKZvfOIhu
[ "Ekdeep Singh Lubana", "Robert P. Dick", "Hidenori Tanaka" ]
Poster
null
Inspired by BatchNorm, there has been an explosion of normalization layers in deep learning. Recent works have identified a multitude of beneficial properties in BatchNorm to explain its success. However, given the pursuit of alternative normalization layers, these properties need to be generalized so that any given la...
[ "Normalization layers", "BatchNorm", "Unified framework" ]
We identify key properties in randomly initialized networks that accurately determine success and failure modes of different normalization layers
10,862
2106.05956
title_snapshot
crnXK0jC2F
Three Operator Splitting with Subgradients, Stochastic Gradients, and Adaptive Learning Rates
https://openreview.net/forum?id=crnXK0jC2F
[ "Alp Yurtsever", "Alex Gu", "Suvrit Sra" ]
Poster
null
Three Operator Splitting (TOS) (Davis & Yin, 2017) can minimize the sum of multiple convex functions effectively when an efficient gradient oracle or proximal operator is available for each term. This requirement often fails in machine learning applications: (i) instead of full gradients only stochastic gradients may b...
[ "three operator splitting", "nonsmooth optimization", "stochastic optimization", "adaptive step-size" ]
null
10,861
2110.03274
title_snapshot
fThfMoV7Ri
Bandit Phase Retrieval
https://openreview.net/forum?id=fThfMoV7Ri
[ "Tor Lattimore", "Botao Hao" ]
Poster
null
We study a bandit version of phase retrieval where the learner chooses actions $(A_t)_{t=1}^n$ in the $d$-dimensional unit ball and the expected reward is $\langle A_t, \theta_\star \rangle^2$ with $\theta_\star \in \mathbb R^d$ an unknown parameter vector. We prove an upper bound on the minimax cumulative regret in th...
[ "Bandit phase retrieval", "minimax regret" ]
We study a bandit version of phase retrieval and prove the minimax cumulative regret as well as simple regret.
10,853
2106.01660
title_snapshot
L5vbEVIePyb
Flexible Option Learning
https://openreview.net/forum?id=L5vbEVIePyb
[ "Martin Klissarov", "Doina Precup" ]
Spotlight
null
Temporal abstraction in reinforcement learning (RL), offers the promise of improving generalization and knowledge transfer in complex environments, by propagating information more efficiently over time. Although option learning was initially formulated in a way that allows updating many options simultaneously, using of...
[ "temporal abstraction", "options", "hierarchical reinforcement learning", "deep reinforcement learning", "reinforcement learning" ]
We propose updates that allow for learning all relevant options simulteneously without introducing additional estimators. We verify that our approach can improve sample efficiency and can open the door to more flexibility when learning options.
10,841
2112.03097
title_snapshot
uVPZCMVtsSG
Weisfeiler and Lehman Go Cellular: CW Networks
https://openreview.net/forum?id=uVPZCMVtsSG
[ "Cristian Bodnar", "Fabrizio Frasca", "Nina Otter", "Yu Guang Wang", "Pietro Liò", "Guido Montufar", "Michael M. Bronstein" ]
Poster
null
Graph Neural Networks (GNNs) are limited in their expressive power, struggle with long-range interactions and lack a principled way to model higher-order structures. These problems can be attributed to the strong coupling between the computational graph and the input graph structure. The recently proposed Message Passi...
[ "graph neural networks", "graph representation learning", "simplicial complexes", "cell complexes", "cw complexes", "simplicial neural networks" ]
We propose a message passing scheme on CW complexes, study its properties and expressive power and apply it to molecular graphs
10,838
2106.12575
title_snapshot
4wVlNqBJXg
Optimal Best-Arm Identification Methods for Tail-Risk Measures
https://openreview.net/forum?id=4wVlNqBJXg
[ "Shubhada Agrawal", "Wouter M Koolen", "Sandeep Kumar Juneja" ]
Poster
null
Conditional value-at-risk (CVaR) and value-at-risk (VaR) are popular tail-risk measures in finance and insurance industries as well as in highly reliable, safety-critical uncertain environments where often the underlying probability distributions are heavy-tailed. We use the multi-armed bandit best-arm identification f...
[ "Multi-armed bandits", "pure exploration", "CVaR", "VaR", "tail-risk", "heavy-tailed distributions", "best-arm identification" ]
We consider the best-arm identification problem in the multi-armed bandit framework where an arm with the smallest tail-risk measure is identified.
10,834
2008.07606
title_snapshot
h1bPe7spQkr
Implicit Generative Copulas
https://openreview.net/forum?id=h1bPe7spQkr
[ "Tim Janke", "Mohamed Ghanmi", "Florian Steinke" ]
Poster
null
Copulas are a powerful tool for modeling multivariate distributions as they allow to separately estimate the univariate marginal distributions and the joint dependency structure. However, known parametric copulas offer limited flexibility especially in high dimensions, while commonly used non-parametric methods suffer ...
[ "copulas", "implicit generative models", "generative neural networks" ]
we use generative neural networks to sample from high-dimensional copulas under guarenteed marginal uniformity
10,817
2109.14567
title_snapshot
lS_rOGT9lfG
Synthetic Design: An Optimization Approach to Experimental Design with Synthetic Controls
https://openreview.net/forum?id=lS_rOGT9lfG
[ "Nick Doudchenko", "Khashayar Khosravi", "Jean Pouget-Abadie", "Sebastien Lahaie", "Miles Lubin", "Vahab Mirrokni", "Jann Spiess", "Guido Imbens" ]
Poster
null
We investigate the optimal design of experimental studies that have pre-treatment outcome data available. The average treatment effect is estimated as the difference between the weighted average outcomes of the treated and control units. A number of commonly used approaches fit this formulation, including the differen...
[ "experimental design", "synthetic control", "causal inference" ]
Synthetic Design: An Optimization Approach to Experimental Design with Synthetic Controls
10,805
2112.00278
title_snapshot
MqCzSKCQ1QB
Adversarial Robustness without Adversarial Training: A Teacher-Guided Curriculum Learning Approach
https://openreview.net/forum?id=MqCzSKCQ1QB
[ "Anindya Sarkar", "Anirban Sarkar", "Sowrya Gali", "Vineeth N. Balasubramanian" ]
Poster
null
Current SOTA adversarially robust models are mostly based on adversarial training (AT) and differ only by some regularizers either at inner maximization or outer minimization steps. Being repetitive in nature during the inner maximization step, they take a huge time to train. We propose a non-iterative method that enfo...
[ "Adversarial Robustness", "Inner Maximization", "Outer Minimization", "Attribution Map", "Curriculum Learning" ]
We propose a non-iterative robust training technique which takes 10-20% time over existing adversarial training (AT) based models and outperforms strong baselines for both adversarial as well as natural accuracies.
10,803
2111.00295
title_judge
vIDBSGl3vzl
Safe Reinforcement Learning by Imagining the Near Future
https://openreview.net/forum?id=vIDBSGl3vzl
[ "Garrett Thomas", "Yuping Luo", "Tengyu Ma" ]
Poster
null
Safe reinforcement learning is a promising path toward applying reinforcement learning algorithms to real-world problems, where suboptimal behaviors may lead to actual negative consequences. In this work, we focus on the setting where unsafe states can be avoided by planning ahead a short time into the future. In this ...
[ "safe reinforcement learning", "model-based reinforcement learning" ]
We devise a short-horizon model-based reinforcement learning algorithm to avoid safety violations by penalizing unsafe trajectories.
10,794
2202.07789
title_snapshot
CONAi0Bh26d
When in Doubt: Neural Non-Parametric Uncertainty Quantification for Epidemic Forecasting
https://openreview.net/forum?id=CONAi0Bh26d
[ "Harshavardhan Kamarthi", "Lingkai Kong", "Alexander Rodriguez", "Chao Zhang", "B. Aditya Prakash" ]
Poster
null
Accurate and trustworthy epidemic forecasting is an important problem for public health planning and disease mitigation. Most existing epidemic forecasting models disregard uncertainty quantification, resulting in mis-calibrated predictions. Recent works in deep neural models for uncertainty-aware time-series forecasti...
[ "Epidemic Forecasting", "Uncertainty Quantification", "Time-Series forecasting", "Deep Probabilistic Models" ]
A novel non-parametric deep generative model for accurate and calibrated uncertainty quantification in real-time epidemic forecasting.
10,789
2106.03904
title_snapshot
oAog3W9w6R
Understanding the Effect of Stochasticity in Policy Optimization
https://openreview.net/forum?id=oAog3W9w6R
[ "Jincheng Mei", "Bo Dai", "Chenjun Xiao", "Csaba Szepesvari", "Dale Schuurmans" ]
Poster
null
We study the effect of stochasticity in on-policy policy optimization, and make the following four contributions. \emph{First}, we show that the preferability of optimization methods depends critically on whether stochastic versus exact gradients are used. In particular, unlike the true gradient setting, geometric info...
[ "reinforcement learning", "policy optimization", "policy gradient", "global convergence" ]
Fast on-policy stochastic policy optimization comes with positive failure probability, and the key factor is "committal rate" rather than variance.
10,780
2110.15572
title_snapshot
bYi_2708mKK
Retiring Adult: New Datasets for Fair Machine Learning
https://openreview.net/forum?id=bYi_2708mKK
[ "Frances Ding", "Moritz Hardt", "John Miller", "Ludwig Schmidt" ]
Oral
null
Although the fairness community has recognized the importance of data, researchers in the area primarily rely on UCI Adult when it comes to tabular data. Derived from a 1994 US Census survey, this dataset has appeared in hundreds of research papers where it served as the basis for the development and comparison of many...
[ "UCI Adult", "datasets", "benchmarks", "Census data", "fairness", "archaeology", "fair machine learning" ]
null
10,778
2108.04884
title_snapshot
gaftyBQ4Lu
Convex-Concave Min-Max Stackelberg Games
https://openreview.net/forum?id=gaftyBQ4Lu
[ "Denizalp Goktas", "Amy Greenwald" ]
Poster
null
Min-max optimization problems (i.e., min-max games) have been attracting a great deal of attention because of their applicability to a wide range of machine learning problems. Although significant progress has been made recently, the literature to date has focused on games with independent strategy sets; little is know...
[ "min-max", "first order methods", "Stackelberg games", "zero-sum games", "Fisher market", "equilibrium", "competitive equilibrium" ]
We introduce polynomial time methods for convex-concave min-max Stackelberg games (with dependent strategy sets) and show their application to the computation of competitive equilibria.
10,756
2110.05192
title_snapshot
WBuLBaoEKNK
Revealing and Protecting Labels in Distributed Training
https://openreview.net/forum?id=WBuLBaoEKNK
[ "Trung Dang", "Om Thakkar", "Swaroop Ramaswamy", "Rajiv Mathews", "Sang Chin", "Françoise Beaufays" ]
Poster
null
Distributed learning paradigms such as federated learning often involve transmission of model updates, or gradients, over a network, thereby avoiding transmission of private data. However, it is possible for sensitive information about the training data to be revealed from such gradients. Prior works have demonstrated ...
[ "privacy", "distributed training" ]
We propose a method to discover the set of labels of training samples from only the gradient of the last layer and the id to label mapping
10,754
2111.00556
title_snapshot
M5j42PvY65V
Intermediate Layers Matter in Momentum Contrastive Self Supervised Learning
https://openreview.net/forum?id=M5j42PvY65V
[ "Aakash Kaku", "Sahana Upadhya", "Narges Razavian" ]
Poster
null
We show that bringing intermediate layers' representations of two augmented versions of an image closer together in self-supervised learning helps to improve the momentum contrastive (MoCo) method. To this end, in addition to the contrastive loss, we minimize the mean squared error between the intermediate layer repres...
[ "Deep learning", "Self-supervised learning", "MoCo", "Momentum contrastive self supervised learning", "Histopathology", "Chest Xray", "Diabetic Retinopathy", "Medical imaging" ]
We improve momentum contrastive self-supervised learning for medical imaging datasets by having additional loss terms that brings the intermediate layer representations of two augmented versions of an image closer together
10,743
2110.14805
title_snapshot
yCA2i3bGbfC
Identification of Partially Observed Linear Causal Models: Graphical Conditions for the Non-Gaussian and Heterogeneous Cases
https://openreview.net/forum?id=yCA2i3bGbfC
[ "Jeffrey Adams", "Niels Richard Hansen", "Kun Zhang" ]
Poster
null
In causal discovery, linear non-Gaussian acyclic models (LiNGAMs) have been studied extensively. While the causally sufficient case is well understood, in many real problems the observed variables are not causally related. Rather, they are generated by latent variables, such as confounders and mediators, which may them...
[ "Causal Discovery", "Structural Equation Models", "Latent Variable Modeling" ]
We provide graphical conditions which are necessary and sufficient for the identification of partially observed linear non-Gaussian causal models.
10,730
null
null
Wp3we5kv6P
EDGE: Explaining Deep Reinforcement Learning Policies
https://openreview.net/forum?id=Wp3we5kv6P
[ "Wenbo Guo", "Xian Wu", "Usmann Khan", "Xinyu Xing" ]
Poster
null
With the rapid development of deep reinforcement learning (DRL) techniques, there is an increasing need to understand and interpret DRL policies. While recent research has developed explanation methods to interpret how an agent determines its moves, they cannot capture the importance of actions/states to a game's final...
[ "Explainable AI", "Deep reinforcement learing", "Explainable Deep reinforcement learing", "Policy debugging" ]
This paper proposes EDGE, a novel DRL explanation method, to identify the critical time steps within a target DRL agent's game episodes and utilizes it to understand agent behavior, discover policy weakness, and remediate policy errors.
10,728
null
null
zEuLFJCRk4X
Imitating Deep Learning Dynamics via Locally Elastic Stochastic Differential Equations
https://openreview.net/forum?id=zEuLFJCRk4X
[ "Jiayao Zhang", "Hua Wang", "Weijie J Su" ]
Poster
null
Understanding the training dynamics of deep learning models is perhaps a necessary step toward demystifying the effectiveness of these models. In particular, how do training data from different classes gradually become separable in their feature spaces when training neural networks using stochastic gradient descent? In...
[ "deep learning", "stochastic differential equations", "ordinary differential equations", "local elasticity" ]
We proposed an SDE model that captures the local elasticity phenomenon to imitate the dynamics of features in deep neural nets.
10,724
2110.05960
title_snapshot