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
PPh6lqP5BO
Teachable Reinforcement Learning via Advice Distillation
https://openreview.net/forum?id=PPh6lqP5BO
[ "Olivia Watkins", "Abhishek Gupta", "Trevor Darrell", "Pieter Abbeel", "Jacob Andreas" ]
Poster
null
Training automated agents to complete complex tasks in interactive environments is challenging: reinforcement learning requires careful hand-engineering of reward functions, imitation learning requires specialized infrastructure and access to a human expert, and learning from intermediate forms of supervision (like bin...
[ "Reinforcement Learning", "Human in the Loop RL" ]
Enabling agents to interpret human in the loop advice to learn new tasks quickly
8,562
2203.11197
title_snapshot
KAFyFabsK88
Storchastic: A Framework for General Stochastic Automatic Differentiation
https://openreview.net/forum?id=KAFyFabsK88
[ "Emile van Krieken", "Jakub Mikolaj Tomczak", "Annette Ten Teije" ]
Poster
null
Modelers use automatic differentiation (AD) of computation graphs to implement complex Deep Learning models without defining gradient computations. Stochastic AD extends AD to stochastic computation graphs with sampling steps, which arise when modelers handle the intractable expectations common in Reinforcement Learnin...
[ "gradient estimation", "automatic differentiation", "optimization", "stochastic computation graphs" ]
We present a framework for gradient estimation in stochastic computation graphs that incorporates many estimators and extends to any-order differentiation.
8,549
2104.00428
title_snapshot
T1f0YKPP_K
Robust Visual Reasoning via Language Guided Neural Module Networks
https://openreview.net/forum?id=T1f0YKPP_K
[ "Arjun Reddy Akula", "Varun Jampani", "Soravit Changpinyo", "Song-Chun Zhu" ]
Poster
null
Neural module networks (NMN) are a popular approach for solving multi-modal tasks such as visual question answering (VQA) and visual referring expression recognition (REF). A key limitation in prior implementations of NMN is that the neural modules do not effectively capture the association between the visual input and...
[ "Neural Module Networks", "Adaptive Convolutions" ]
Robust Visual Reasoning via Language Guided Neural Module Networks
8,545
null
null
mf9XiRCEgZu
Noisy Recurrent Neural Networks
https://openreview.net/forum?id=mf9XiRCEgZu
[ "Soon Hoe Lim", "N. Benjamin Erichson", "Liam Hodgkinson", "Michael W. Mahoney" ]
Poster
null
We provide a general framework for studying recurrent neural networks (RNNs) trained by injecting noise into hidden states. Specifically, we consider RNNs that can be viewed as discretizations of stochastic differential equations driven by input data. This framework allows us to study the implicit regularization effect...
[ "Recurrent neural networks", "implicit regularization", "stochastic differential equations", "stability analysis", "model robustness" ]
This paper provides a stochastic differential equation framework to study implicit regularization, model stability and robustness of recurrent neural networks trained by injecting noise into hidden states.
8,544
2102.04877
title_snapshot
GrZmKDYCp6H
An Information-theoretic Approach to Distribution Shifts
https://openreview.net/forum?id=GrZmKDYCp6H
[ "Marco Federici", "Ryota Tomioka", "Patrick Forré" ]
Poster
null
Safely deploying machine learning models to the real world is often a challenging process. For example, models trained with data obtained from a specific geographic location tend to fail when queried with data obtained elsewhere, agents trained in a simulation can struggle to adapt when deployed in the real world or no...
[ "Information Theory", "Selection Bias", "Representation Learning", "Distribution Shift", "Fair Machine Learning" ]
We analyze the problem of distribution shift from an information-theoretical perspective and describe different strategies in literature to correct for selection bias; empirically underlying their strengths and weaknesses.
8,541
2106.03783
title_snapshot
ZYX1ff6H0Bs
Compressive Visual Representations
https://openreview.net/forum?id=ZYX1ff6H0Bs
[ "Kuang-Huei Lee", "Anurag Arnab", "Sergio Guadarrama", "John Canny", "Ian Fischer" ]
Poster
null
Learning effective visual representations that generalize well without human supervision is a fundamental problem in order to apply Machine Learning to a wide variety of tasks. Recently, two families of self-supervised methods, contrastive learning and latent bootstrapping, exemplified by SimCLR and BYOL respectively, ...
[ "Self-Supervised Learning", "Information Theory", "Compression", "Visual Representations", "Vision" ]
We introduce information-compressive variants of the state-of-the-art SimCLR and BYOL self-supervised algorithms, and achieve linear evaluation results comparable to supervised models and better model robustness.
8,524
2109.12909
title_snapshot
jfDaBf8PAE
Fast Minimum-norm Adversarial Attacks through Adaptive Norm Constraints
https://openreview.net/forum?id=jfDaBf8PAE
[ "Maura Pintor", "Fabio Roli", "Wieland Brendel", "Battista Biggio" ]
Poster
null
Evaluating adversarial robustness amounts to finding the minimum perturbation needed to have an input sample misclassified. The inherent complexity of the underlying optimization requires current gradient-based attacks to be carefully tuned, initialized, and possibly executed for many computationally-demanding iterati...
[ "machine learning", "adversarial machine learning", "computer vision", "adversarial robustness", "adversarial attacks" ]
Minimum-norm gradient-based adversarial attack that works with multiple $\ell_p$ norms.
8,520
2102.12827
title_snapshot
x6z8J_17LP3
Lip to Speech Synthesis with Visual Context Attentional GAN
https://openreview.net/forum?id=x6z8J_17LP3
[ "Minsu Kim", "Joanna Hong", "Yong Man Ro" ]
Poster
null
In this paper, we propose a novel lip-to-speech generative adversarial network, Visual Context Attentional GAN (VCA-GAN), which can jointly model local and global lip movements during speech synthesis. Specifically, the proposed VCA-GAN synthesizes the speech from local lip visual features by finding a mapping function...
[ "video driven speech synthesis", "speech reconstruction from silent video", "lip reading", "audio-visual attention", "context attentional gan" ]
In this paper, we propose a novel lip-to-speech generative adversarial network, Visual Context Attentional GAN (VCA-GAN), which can jointly model local and global lip movements during speech synthesis.
8,518
2204.01726
title_snapshot
F93Z9Au6HxE
SSAL: Synergizing between Self-Training and Adversarial Learning for Domain Adaptive Object Detection
https://openreview.net/forum?id=F93Z9Au6HxE
[ "Muhammad Akhtar Munir", "Muhammad Haris Khan", "M. Saquib Sarfraz", "Mohsen Ali" ]
Poster
null
We study adapting trained object detectors to unseen domains manifesting significant variations of object appearance, viewpoints and backgrounds. Most current methods align domains by either using image or instance-level feature alignment in an adversarial fashion. This often suffers due to the presence of unwanted bac...
[ "domain adaptation", "uncertainty", "deep learning" ]
Domain adaptive object detection with feature level alignment and self training based on model’s predictive uncertainty.
8,517
null
null
H6y7EAf7s4P
Disentangling the Roles of Curation, Data-Augmentation and the Prior in the Cold Posterior Effect
https://openreview.net/forum?id=H6y7EAf7s4P
[ "Lorenzo Noci", "Kevin Roth", "Gregor Bachmann", "Sebastian Nowozin", "Thomas Hofmann" ]
Poster
null
The “cold posterior effect” (CPE) in Bayesian deep learning describes the disturbing observation that the predictive performance of Bayesian neural networks can be significantly improved if the Bayes posterior is artificially sharpened using a temperature parameter T <1. The CPE is problematic in theory and practice a...
[ "Bayesian Neural Networks", "Cold Posterior", "Curation", "Data-Augmentation", "Prior" ]
We provide novel and nuanced evidence relevant to existing explanations for the cold posterior effect, disentangling the roles of curation, data-augmentation and the prior.
8,516
2106.06596
title_snapshot
qKRr_rNCEPz
BooVI: Provably Efficient Bootstrapped Value Iteration
https://openreview.net/forum?id=qKRr_rNCEPz
[ "Boyi Liu", "Qi Cai", "Zhuoran Yang", "Zhaoran Wang" ]
Poster
null
Despite the tremendous success of reinforcement learning (RL) with function approximation, efficient exploration remains a significant challenge, both practically and theoretically. In particular, existing theoretically grounded RL algorithms based on upper confidence bounds (UCBs), such as optimistic least-squares val...
[ "Reinforcement Learning", "Exploration" ]
null
8,514
null
null
sR1XB9-F-rv
Antipodes of Label Differential Privacy: PATE and ALIBI
https://openreview.net/forum?id=sR1XB9-F-rv
[ "Mani Malek Esmaeili", "Ilya Mironov", "Karthik Prasad", "Igor Shilov", "Florian Tramer" ]
Poster
null
We consider the privacy-preserving machine learning (ML) setting where the trained model must satisfy differential privacy (DP) with respect to the labels of the training examples. We propose two novel approaches based on, respectively, the Laplace mechanism and the PATE framework, and demonstrate their effectiveness o...
[ "differential privacy", "label differential privacy", "PATE", "ALIBI", "memorization attacks" ]
We propose two very distinct approaches for training models with label-only differential privacy (Label DP) and evaluate them with memorization attacks.
8,502
2106.03408
title_snapshot
REXvo_lsQS9
Credit Assignment in Neural Networks through Deep Feedback Control
https://openreview.net/forum?id=REXvo_lsQS9
[ "Alexander Meulemans", "Matilde Tristany Farinha", "Javier Garcia Ordonez", "Pau Vilimelis Aceituno", "Joao Sacramento", "Benjamin F Grewe" ]
Spotlight
null
The success of deep learning sparked interest in whether the brain learns by using similar techniques for assigning credit to each synaptic weight for its contribution to the network output. However, the majority of current attempts at biologically-plausible learning methods are either non-local in time, require highly...
[ "Biologically-plausible Deep Learning", "Neuroscience", "Deep Learning", "Control theory", "Optimization theory" ]
We introduce Deep Feedback Control, a new bio-plausible learning method that uses a feedback controller to drive a deep neural network to match a desired output target and which approximates Gauss-Newton optimization.
8,501
2106.07887
title_snapshot
1W2WuCYbz_C
A Comprehensively Tight Analysis of Gradient Descent for PCA
https://openreview.net/forum?id=1W2WuCYbz_C
[ "zhiqiang xu", "Ping Li" ]
Poster
null
We study the Riemannian gradient method for PCA on which a crucial fact is that despite the simplicity of the considered setting, i.e., deterministic version of Krasulina's method, the convergence rate has not been well-understood yet. In this work, we provide a general tight analysis for the gap-dependent rate at $O(\...
[ "PCA", "Riemannian gradient descent", "tight rate", "gap-free rate" ]
A Comprehensively Tight Analysis of Gradient Descent for PCA
8,500
null
null
Hox8lKfr82L
Rethinking conditional GAN training: An approach using geometrically structured latent manifolds
https://openreview.net/forum?id=Hox8lKfr82L
[ "Sameera Ramasinghe", "Moshiur R Farazi", "Salman Khan", "Nick Barnes", "Stephen Gould" ]
Poster
null
Conditional GANs (cGAN), in their rudimentary form, suffer from critical drawbacks such as the lack of diversity in generated outputs and distortion between the latent and output manifolds. Although efforts have been made to improve results, they can suffer from unpleasant side-effects such as the topology mismatch be...
[ "Generative modeling", "Multimodal output spaces", "Conditional generative models" ]
A novel generic training algorithm for training conditional GANs for improved diversity, realism and structure
8,493
2011.13055
title_snapshot
DTA7Bgrai-Q
Precise characterization of the prior predictive distribution of deep ReLU networks
https://openreview.net/forum?id=DTA7Bgrai-Q
[ "Lorenzo Noci", "Gregor Bachmann", "Kevin Roth", "Sebastian Nowozin", "Thomas Hofmann" ]
Spotlight
null
Recent works on Bayesian neural networks (BNNs) have highlighted the need to better understand the implications of using Gaussian priors in combination with the compositional structure of the network architecture. Similar in spirit to the kind of analysis that has been developed to devise better initialization schemes...
[ "Bayesian Neural Networks", "Priors", "Meijer-G", "NNGP", "finite width corrections" ]
We derive a precise characterization of the prior predictive distribution of deep ReLU networks using Meijer-G functions.
8,491
2106.06615
title_snapshot
Qijzj3WqUl3
Adaptive Proximal Gradient Methods for Structured Neural Networks
https://openreview.net/forum?id=Qijzj3WqUl3
[ "Jihun Yun", "Aurelie Lozano", "Eunho Yang" ]
Poster
null
We consider the training of structured neural networks where the regularizer can be non-smooth and possibly non-convex. While popular machine learning libraries have resorted to stochastic (adaptive) subgradient approaches, the use of proximal gradient methods in the stochastic setting has been little explored and warr...
[ "Proximal Gradient Descent for Adaptive Methods", "Stochastic Optimization", "Non-convex Optimization" ]
Optimization for Regularized Deep Networks; Stochastic Optimization; Proximal Gradient Descent
8,482
null
null
bGfDnD7xo-v
Continuized Accelerations of Deterministic and Stochastic Gradient Descents, and of Gossip Algorithms
https://openreview.net/forum?id=bGfDnD7xo-v
[ "Mathieu Even", "Raphaël Berthier", "Francis Bach", "Nicolas Flammarion", "Hadrien Hendrikx", "Pierre Gaillard", "Laurent Massoulié", "Adrien Taylor" ]
Oral
null
We introduce the ``continuized'' Nesterov acceleration, a close variant of Nesterov acceleration whose variables are indexed by a continuous time parameter. The two variables continuously mix following a linear ordinary differential equation and take gradient steps at random times. This continuized variant benefits fro...
[ "Convex optimization", "Nesterov acceleration" ]
null
8,470
null
null
dL8p6rLFTS3
PiRank: Scalable Learning To Rank via Differentiable Sorting
https://openreview.net/forum?id=dL8p6rLFTS3
[ "Robin Marcel Edwin Swezey", "Aditya Grover", "Bruno Charron", "Stefano Ermon" ]
Poster
null
A key challenge with machine learning approaches for ranking is the gap between the performance metrics of interest and the surrogate loss functions that can be optimized with gradient-based methods. This gap arises because ranking metrics typically involve a sorting operation which is not differentiable w.r.t. the mod...
[ "learning-to-rank", "differentiable sorting", "continuous relaxations" ]
We propose a framework for learning-to-rank based on a scalable, differentiable relaxation to sorting.
8,466
2012.06731
title_snapshot
2zCRcTafea
Focal Attention for Long-Range Interactions in Vision Transformers
https://openreview.net/forum?id=2zCRcTafea
[ "Jianwei Yang", "Chunyuan Li", "Pengchuan Zhang", "Xiyang Dai", "Bin Xiao", "Lu Yuan", "Jianfeng Gao" ]
Spotlight
null
Recently, Vision Transformer and its variants have shown great promise on various computer vision tasks. The ability to capture local and global visual dependencies through self-attention is the key to its success. But it also brings challenges due to quadratic computational overhead, especially for the high-resolution...
[ "Focal Attention", "Self-Attention", "Long-range Interactions", "Local-Global Interactions", "Vision Transformer", "Image Classification", "Object Detection" ]
An effective focal attention mechanism for modeling short- and long-range visual dependencies in Vision Transformers
8,461
2107.00641
title_judge
Pgv4fwfh63L
Catalytic Role Of Noise And Necessity Of Inductive Biases In The Emergence Of Compositional Communication
https://openreview.net/forum?id=Pgv4fwfh63L
[ "Łukasz Kuciński", "Tomasz Korbak", "Paweł Kołodziej", "Piotr Miłoś" ]
Poster
null
Communication is compositional if complex signals can be represented as a combination of simpler subparts. In this paper, we theoretically show that inductive biases on both the training framework and the data are needed to develop a compositional communication. Moreover, we prove that compositionality spontaneously a...
[ "compositionality", "signaling games", "noisy channel", "deep learning" ]
We show that inductive biases are needed for the emergence of compositional communication and that it emerges in signaling games in which agents communicate over a noisy channel.
8,460
2111.06464
title_snapshot
8CaXYuLlZ6o
Activation Sharing with Asymmetric Paths Solves Weight Transport Problem without Bidirectional Connection
https://openreview.net/forum?id=8CaXYuLlZ6o
[ "Sunghyeon Woo", "Jeongwoo Park", "Jiwoo Hong", "Dongsuk Jeon" ]
Poster
null
One of the reasons why it is difficult for the brain to perform backpropagation (BP) is the weight transport problem, which argues forward and feedback neurons cannot share the same synaptic weights during learning in biological neural networks. Recently proposed algorithms address the weight transport problem while pr...
[ "activation sharing", "weight transport problem", "bidirectional connection", "biologically plausible algorithms", "neuromorphic algorithms", "deep convolutional neural network", "biological neural network" ]
The proposed biologically plausible algorithm supports training deep convolutional networks without the weight transport problem and bidirectional connections.
8,450
null
null
qGn3Rlgul5F
Parametrized Quantum Policies for Reinforcement Learning
https://openreview.net/forum?id=qGn3Rlgul5F
[ "Sofiene Jerbi", "Casper Gyurik", "Simon Callum Marshall", "Hans J Briegel", "Vedran Dunjko" ]
Poster
null
With the advent of real-world quantum computing, the idea that parametrized quantum computations can be used as hypothesis families in a quantum-classical machine learning system is gaining increasing traction. Such hybrid systems have already shown the potential to tackle real-world tasks in supervised and generative ...
[ "reinforcement learning", "quantum computing", "parametrized quantum circuits", "quantum neural networks", "policy gradient", "quantum machine learning", "quantum reinforcement learning", "quantum", "variational quantum circuits" ]
We investigate the potential of parametrized quantum computations when trained as reinforcement learning policies in classical environments.
8,446
2103.05577
title_snapshot
UZm2IQhgIyB
Label-Imbalanced and Group-Sensitive Classification under Overparameterization
https://openreview.net/forum?id=UZm2IQhgIyB
[ "Ganesh Ramachandra Kini", "Orestis Paraskevas", "Samet Oymak", "Christos Thrampoulidis" ]
Poster
null
The goal in label-imbalanced and group-sensitive classification is to optimize relevant metrics such as balanced error and equal opportunity. Classical methods, such as weighted cross-entropy, fail when training deep nets to the terminal phase of training (TPT), that is training beyond zero training error. This observa...
[ "fairness", "overparameterization", "cost-sensitive methods" ]
null
8,422
2103.01550
title_snapshot
AnJUTpZiiWD
Instance-Dependent Partial Label Learning
https://openreview.net/forum?id=AnJUTpZiiWD
[ "Ning Xu", "Congyu Qiao", "Xin Geng", "Min-Ling Zhang" ]
Spotlight
null
Partial label learning (PLL) is a typical weakly supervised learning problem, where each training example is associated with a set of candidate labels among which only one is true. Most existing PLL approaches assume that the incorrect labels in each training example are randomly picked as the candidate labels. However...
[ "partial label learning", "label enhancement", "label distribution", "pseudo label" ]
We consider instance-dependent PLL and assume that each example is associated with a latent label distribution.
8,421
2110.12911
title_snapshot
4vUZPUKZsr5
D2C: Diffusion-Decoding Models for Few-Shot Conditional Generation
https://openreview.net/forum?id=4vUZPUKZsr5
[ "Abhishek Sinha", "Jiaming Song", "Chenlin Meng", "Stefano Ermon" ]
Poster
null
Conditional generative models of high-dimensional images have many applications, but supervision signals from conditions to images can be expensive to acquire. This paper describes Diffusion-Decoding models with Contrastive representations (D2C), a paradigm for training unconditional variational autoencoders (VAE) for ...
[ "generative models" ]
We propose a paradigm for training unconditional variational autoencoders (VAEs) for few-shot conditional image generation
8,414
2106.06819
title_judge
FTt28RYj5Pc
Training Certifiably Robust Neural Networks with Efficient Local Lipschitz Bounds
https://openreview.net/forum?id=FTt28RYj5Pc
[ "Yujia Huang", "Huan Zhang", "Yuanyuan Shi", "J Zico Kolter", "Anima Anandkumar" ]
Poster
null
Certified robustness is a desirable property for deep neural networks in safety-critical applications, and popular training algorithms can certify robustness of a neural network by computing a global bound on its Lipschitz constant. However, such a bound is often loose: it tends to over-regularize the neural network an...
[ "lipschitz constant", "local lipschitz constant", "certified defense", "adversarial examples", "robustness" ]
We propose an efficient and trainable local Lipscthiz bound for training certifibly robust neural networks.
8,401
2111.01395
title_snapshot
cksOcsjnXh
Revenue maximization via machine learning with noisy data
https://openreview.net/forum?id=cksOcsjnXh
[ "Ellen Vitercik", "Tom Yan" ]
Poster
null
Increasingly, copious amounts of consumer data are used to learn high-revenue mechanisms via machine learning. Existing research on mechanism design via machine learning assumes that there is a distribution over the buyers' values for the items for sale and that the learning algorithm's input is a training set sampled ...
[ "Mechanism design", "auction design", "revenue maximization", "machine learning theory", "sample complexity", "generalization guarantees", "adversarial noise" ]
null
8,399
null
null
YCqx6zhEzRp
Federated Multi-Task Learning under a Mixture of Distributions
https://openreview.net/forum?id=YCqx6zhEzRp
[ "Othmane MARFOQ", "Giovanni Neglia", "Aurélien Bellet", "Laetitia Kameni", "Richard Vidal" ]
Poster
null
The increasing size of data generated by smartphones and IoT devices motivated the development of Federated Learning (FL), a framework for on-device collaborative training of machine learning models. First efforts in FL focused on learning a single global model with good average performance across clients, but the glob...
[ "Federated learning", "multi-task learning", "personalized models", "expectation-maximization", "personalized federated learning" ]
A novel personalized federated learning method based on the flexible assumption that each local data distribution is a mixture of unknown underlying distributions.
8,397
2108.10252
title_snapshot
DEsIX_D_vR
Neural Bellman-Ford Networks: A General Graph Neural Network Framework for Link Prediction
https://openreview.net/forum?id=DEsIX_D_vR
[ "Zhaocheng Zhu", "Zuobai Zhang", "Louis-Pascal A. C. Xhonneux", "Jian Tang" ]
Poster
null
Link prediction is a very fundamental task on graphs. Inspired by traditional path-based methods, in this paper we propose a general and flexible representation learning framework based on paths for link prediction. Specifically, we define the representation of a pair of nodes as the generalized sum of all path represe...
[ "graph neural networks", "link prediction", "knowledge graph completion", "Bellman-Ford algorithm" ]
An inductive, interpretable, flexible and scalable framework for link prediction inspired by the generalized Bellman-Ford algorithm.
8,396
2106.06935
title_snapshot
83A-0x6Pfi_
Adversarial Robustness of Streaming Algorithms through Importance Sampling
https://openreview.net/forum?id=83A-0x6Pfi_
[ "Vladimir Braverman", "Avinatan Hassidim", "Yossi Matias", "Mariano Schain", "Sandeep Silwal", "Samson Zhou" ]
Poster
null
Robustness against adversarial attacks has recently been at the forefront of algorithmic design for machine learning tasks. In the adversarial streaming model, an adversary gives an algorithm a sequence of adaptively chosen updates $u_1,\ldots,u_n$ as a data stream. The goal of the algorithm is to compute or approximat...
[ "adversarial robustness", "streaming algorithms", "coresets" ]
We show that importance sampling leads to adversarially robust algorithms for many streaming problems.
8,392
2106.14952
title_snapshot
0Kb33DHJ1g
Adaptive Diffusion in Graph Neural Networks
https://openreview.net/forum?id=0Kb33DHJ1g
[ "Jialin Zhao", "Yuxiao Dong", "Ming Ding", "Evgeny Kharlamov", "Jie Tang" ]
Poster
null
The success of graph neural networks (GNNs) largely relies on the process of aggregating information from neighbors defined by the input graph structures. Notably, message passing based GNNs, e.g., graph convolutional networks, leverage the immediate neighbors of each node during the aggregation process, and recently, ...
[ "Graph Convolutional Network" ]
null
8,388
null
null
vecLnc6g6iQ
Neural Rule-Execution Tracking Machine For Transformer-Based Text Generation
https://openreview.net/forum?id=vecLnc6g6iQ
[ "Yufei Wang", "Can Xu", "Huang Hu", "Chongyang Tao", "Stephen Wan", "Mark Dras", "Mark Johnson", "Daxin Jiang" ]
Poster
null
Sequence-to-Sequence (Seq2Seq) neural text generation models, especially the pre-trained ones (e.g., BART and T5), have exhibited compelling performance on various natural language generation tasks. However, the black-box nature of these models limits their application in tasks where specific rules (e.g., controllable ...
[ "Sequence-to-Sequence Text Generation", "Rule Execution Tracking", "Pre-trained Transformer-based Language Models" ]
In this paper, we propose a novel module named Neural Rule-Execution Tracking Machine to leverage multiple rules simultaneously to guide the neural generation model for superior generation performance in a unified and scalable way.
8,371
2107.13077
title_snapshot
0HW7A5YZjq7
Self-Supervised Learning with Kernel Dependence Maximization
https://openreview.net/forum?id=0HW7A5YZjq7
[ "Yazhe Li", "Roman Pogodin", "Danica J. Sutherland", "Arthur Gretton" ]
Poster
null
We approach self-supervised learning of image representations from a statistical dependence perspective, proposing Self-Supervised Learning with the Hilbert-Schmidt Independence Criterion (SSL-HSIC). SSL-HSIC maximizes dependence between representations of transformations of an image and the image identity, while minim...
[ "self-supervised learning", "representation learning", "dependence maximization", "HSIC", "deep kernel methods" ]
Kernel measures of dependence help us understand self-supervised models, and can get better performance.
8,368
2106.08320
title_snapshot
Ydlco-tfIG
Explanation-based Data Augmentation for Image Classification
https://openreview.net/forum?id=Ydlco-tfIG
[ "Sandareka Wickramanayake", "Wynne Hsu", "Mong-Li Lee" ]
Poster
null
Existing works have generated explanations for deep neural network decisions to provide insights into model behavior. We observe that these explanations can also be used to identify concepts that caused misclassifications. This allows us to understand the possible limitations of the dataset used to train the model, par...
[ "Explainability", "Data Augmentation", "Image Classification" ]
An explanation based data augmentation approach is proposed to improve accuracy of image classifiers.
8,364
null
null
4VAp_PL9yKs
Never Go Full Batch (in Stochastic Convex Optimization)
https://openreview.net/forum?id=4VAp_PL9yKs
[ "Idan Amir", "Yair Carmon", "Tomer Koren", "Roi Livni" ]
Poster
null
We study the generalization performance of $\text{\emph{full-batch}}$ optimization algorithms for stochastic convex optimization: these are first-order methods that only access the exact gradient of the empirical risk (rather than gradients with respect to individual data points), that include a wide range of algorithm...
[ "Stochastic Convex Optimization", "First-order Optimization", "Gradient Methods", "Stochastic Gradient Descent", "Generalization" ]
We show that any first-order full-batch optimization algorithm must perform at least O(1/eps^4) iterations in order to generalize.
8,362
2107.00469
title_snapshot
G7W2mriQLxf
Learned Robust PCA: A Scalable Deep Unfolding Approach for High-Dimensional Outlier Detection
https://openreview.net/forum?id=G7W2mriQLxf
[ "HanQin Cai", "Jialin Liu", "Wotao Yin" ]
Poster
null
Robust principal component analysis (RPCA) is a critical tool in modern machine learning, which detects outliers in the task of low-rank matrix reconstruction. In this paper, we propose a scalable and learnable non-convex approach for high-dimensional RPCA problems, which we call Learned Robust PCA (LRPCA). LRPCA is hi...
[ "Robust PCA", "Outlier Detection", "Deep Unfolding", "Learning to Optimize", "Video Background Subtraction", "Ultrasound Imaging" ]
null
8,361
2110.05649
title_snapshot
wLsA3nurh9W
Analysis of one-hidden-layer neural networks via the resolvent method
https://openreview.net/forum?id=wLsA3nurh9W
[ "Vanessa Piccolo", "Dominik Schröder" ]
Poster
null
In this work, we investigate the asymptotic spectral density of the random feature matrix $M = Y Y^*$ with $Y = f(WX)$ generated by a single-hidden-layer neural network, where $W$ and $X$ are random rectangular matrices with i.i.d. centred entries and $f$ is a non-linear smooth function which is applied entry-wise. We ...
[ "Neural network", "Resolvent method", "Cumulant expansion" ]
We derive the singular value distribution of fully random 1-layer neural networks using the resolvent method.
8,350
2105.05115
title_snapshot
cVwc7IHWEWi
Near-Optimal No-Regret Learning in General Games
https://openreview.net/forum?id=cVwc7IHWEWi
[ "Constantinos Costis Daskalakis", "Maxwell Fishelson", "Noah Golowich" ]
Oral
null
We show that Optimistic Hedge -- a common variant of multiplicative-weights-updates with recency bias -- attains ${\rm poly}(\log T)$ regret in multi-player general-sum games. In particular, when every player of the game uses Optimistic Hedge to iteratively update her action in response to the history of play so far, t...
[ "No-regret learning", "coarse correlated equilibrium", "Optimistic Hedge" ]
We prove a poly-logarithmic regret bound for no-regret learners in general-sum games.
8,349
2108.06924
title_snapshot
FyOhThdDBM
Uniform Convergence of Interpolators: Gaussian Width, Norm Bounds and Benign Overfitting
https://openreview.net/forum?id=FyOhThdDBM
[ "Frederic Koehler", "Lijia Zhou", "Danica J. Sutherland", "Nathan Srebro" ]
Oral
null
We consider interpolation learning in high-dimensional linear regression with Gaussian data, and prove a generic uniform convergence guarantee on the generalization error of interpolators in an arbitrary hypothesis class in terms of the class’s Gaussian width. Applying the generic bound to Euclidean norm balls recover...
[ "interpolation learning", "uniform convergence", "double descent", "statistical learning theory" ]
Uniform convergence of interpolating predictors can explain consistency for high-dimensional linear regression.
8,343
2106.09276
title_snapshot
8Yrcy55iHE
A Trainable Spectral-Spatial Sparse Coding Model for Hyperspectral Image Restoration
https://openreview.net/forum?id=8Yrcy55iHE
[ "Theo Bodrito", "Alexandre Zouaoui", "Jocelyn Chanussot", "Julien Mairal" ]
Poster
null
Hyperspectral imaging offers new perspectives for diverse applications, ranging from the monitoring of the environment using airborne or satellite remote sensing, precision farming, food safety, planetary exploration, or astrophysics. Unfortunately, the spectral diversity of information comes at the expense of various ...
[ "Hyperspectral denoising", "trainable sparse coding models for image denoising" ]
A Trainable Spectral-Spatial Sparse Coding Model for Hyperspectral Image Restoration
8,332
2111.09708
title_snapshot
Owggnutk6lE
Alias-Free Generative Adversarial Networks
https://openreview.net/forum?id=Owggnutk6lE
[ "Tero Karras", "Miika Aittala", "Samuli Laine", "Erik Härkönen", "Janne Hellsten", "Jaakko Lehtinen", "Timo Aila" ]
Oral
null
We observe that despite their hierarchical convolutional nature, the synthesis process of typical generative adversarial networks depends on absolute pixel coordinates in an unhealthy manner. This manifests itself as, e.g., detail appearing to be glued to image coordinates instead of the surfaces of depicted objects. W...
[ "GAN", "CNN", "signal processing", "aliasing", "image synthesis" ]
Careless signal processing causes "texture sticking" artifacts in GANs; our redesign solves them and dramatically improves visual quality under animation.
8,331
2106.12423
title_snapshot
RwASmRpLp-
NORESQA: A Framework for Speech Quality Assessment using Non-Matching References
https://openreview.net/forum?id=RwASmRpLp-
[ "Pranay Manocha", "Buye Xu", "Anurag Kumar" ]
Poster
null
The perceptual task of speech quality assessment (SQA) is a challenging task for machines to do. Objective SQA methods that rely on the availability of the corresponding clean reference have been the primary go-to approaches for SQA. Clearly, these methods fail in real-world scenarios where the ground truth clean refer...
[ "no-reference", "speech quality", "unsupervised learning", "speech enhancement", "perceptual similarity", "speech processing" ]
novel framework for speech quality assessment using random non-matched reference recordings
8,327
2109.08125
title_snapshot
Hcr9mgBG6ds
Labeling Trick: A Theory of Using Graph Neural Networks for Multi-Node Representation Learning
https://openreview.net/forum?id=Hcr9mgBG6ds
[ "Muhan Zhang", "Pan Li", "Yinglong Xia", "Kai Wang", "Long Jin" ]
Poster
null
In this paper, we provide a theory of using graph neural networks (GNNs) for multi-node representation learning (where we are interested in learning a representation for a set of more than one node, such as link). We know that GNN is designed to learn single-node representations. When we want to learn a node set repres...
[ "theory", "GNN", "link prediction" ]
null
8,325
2010.16103
title_snapshot
CYUzpnOkFJp
NovelD: A Simple yet Effective Exploration Criterion
https://openreview.net/forum?id=CYUzpnOkFJp
[ "Tianjun Zhang", "Huazhe Xu", "Xiaolong Wang", "Yi Wu", "Kurt Keutzer", "Joseph E. Gonzalez", "Yuandong Tian" ]
Poster
null
Efficient exploration under sparse rewards remains a key challenge in deep reinforcement learning. Previous exploration methods (e.g., RND) have achieved strong results in multiple hard tasks. However, if there are multiple novel areas to explore, these methods often focus quickly on one without sufficiently trying oth...
[ "RL", "Exploration" ]
We propose a novelty exploration strategy NovelD and show strong performance.
8,321
null
null
j5NrN8ffXC
Exploring the Limits of Out-of-Distribution Detection
https://openreview.net/forum?id=j5NrN8ffXC
[ "Stanislav Fort", "Jie Ren", "Balaji Lakshminarayanan" ]
Poster
null
Near out-of-distribution detection (OOD) is a major challenge for deep neural networks. We demonstrate that large-scale pre-trained transformers can significantly improve the state-of-the-art (SOTA) on a range of near OOD tasks across different data modalities. For instance, on CIFAR-100 vs CIFAR-10 OOD detection, we i...
[ "out-of-distribution detection", "outlier exposure" ]
We significantly improve the state-of-the-art on difficult out-of-distribution detection benchmarks (improving AUROC from 85% to 96% on CIFAR-100 vs CIFAR-10, 66% to 77% on genomics) and show that outlier exposure can further improve performance.
8,319
2106.03004
title_snapshot
giEMdtueyZn
Damped Anderson Mixing for Deep Reinforcement Learning: Acceleration, Convergence, and Stabilization
https://openreview.net/forum?id=giEMdtueyZn
[ "Ke Sun", "Yafei Wang", "Yi Liu", "Yingnan Zhao", "Bo Pan", "SHANGLING JUI", "Bei Jiang", "Linglong Kong" ]
Poster
null
Anderson mixing has been heuristically applied to reinforcement learning (RL) algorithms for accelerating convergence and improving the sampling efficiency of deep RL. Despite its heuristic improvement of convergence, a rigorous mathematical justification for the benefits of Anderson mixing in RL has not yet been put f...
[ "Anderson Acceleration", "Reinforcement Learning" ]
null
8,301
2110.08896
title_snapshot
Ruw3MHL9jAO
Corruption Robust Active Learning
https://openreview.net/forum?id=Ruw3MHL9jAO
[ "Yifang Chen", "Simon Shaolei Du", "Kevin Jamieson" ]
Poster
null
We conduct theoretical studies on streaming-based active learning for binary classification under unknown adversarial label corruptions. In this setting, every time before the learner observes a sample, the adversary decides whether to corrupt the label ornot. First, we show that, in a benign corruption setting (which ...
[ "active learning", "robustness" ]
We design a new corruption robust algorithm for non-oblivious adversary under the stream-based active learning setting.
8,298
2106.11220
title_snapshot
N0Pigj5tpHE
Graph Posterior Network: Bayesian Predictive Uncertainty for Node Classification
https://openreview.net/forum?id=N0Pigj5tpHE
[ "Maximilian Stadler", "Bertrand Charpentier", "Simon Geisler", "Daniel Zügner", "Stephan Günnemann" ]
Poster
null
The interdependence between nodes in graphs is key to improve class prediction on nodes, utilized in approaches like Label Probagation (LP) or in Graph Neural Networks (GNNs). Nonetheless, uncertainty estimation for non-independent node-level predictions is under-explored. In this work, we explore uncertainty quantifi...
[ "Uncertainty", "Bayesian rule", "Posterior", "Graph", "Classification" ]
null
8,290
2110.14012
title_snapshot
FeFIzwifdoL
Hindsight Task Relabelling: Experience Replay for Sparse Reward Meta-RL
https://openreview.net/forum?id=FeFIzwifdoL
[ "Charles Packer", "Pieter Abbeel", "Joseph E. Gonzalez" ]
Poster
null
Meta-reinforcement learning (meta-RL) has proven to be a successful framework for leveraging experience from prior tasks to rapidly learn new related tasks, however, current meta-RL approaches struggle to learn in sparse reward environments. Although existing meta-RL algorithms can learn strategies for adapting to new ...
[ "Meta Reinforcement Learning", "Meta-RL", "Deep RL", "Hindsight", "Sparse Reward" ]
Meta-RL algorithms struggle with sparse reward environments and often require a dense reward function for training - we fix this by applying ideas from Hindsight Experience Replay to meta-RL.
8,289
2112.00901
title_snapshot
A_TVp2HtxPS
Physics-Aware Downsampling with Deep Learning for Scalable Flood Modeling
https://openreview.net/forum?id=A_TVp2HtxPS
[ "Niv Giladi", "Zvika Ben-Haim", "Sella Nevo", "Yossi Matias", "Daniel Soudry" ]
Poster
null
Background. Floods are the most common natural disaster in the world, affecting the lives of hundreds of millions. Flood forecasting is therefore a vitally important endeavor, typically achieved using physical water flow simulations, which rely on accurate terrain elevation maps. However, such simulations, based on sol...
[ "Partial differential equation", "Inundation modeling", "Physics-informed", "Downsampling", "Flood modeling" ]
Finding an optimized coarse grid representation of terrain maps with a physics aware deep neural network, for accurate flood prediction.
8,278
2106.07218
title_snapshot
7_eLEvFjCi3
Nested Graph Neural Networks
https://openreview.net/forum?id=7_eLEvFjCi3
[ "Muhan Zhang", "Pan Li" ]
Poster
null
Graph neural network (GNN)'s success in graph classification is closely related to the Weisfeiler-Lehman (1-WL) algorithm. By iteratively aggregating neighboring node features to a center node, both 1-WL and GNN obtain a node representation that encodes a rooted subtree around the center node. These rooted subtree repr...
[ "graph neural network", "theory", "expressive power", "Weisfeiler-Lehman" ]
null
8,273
2110.13197
title_snapshot
HjFtRc83eBB
A Near-Optimal Algorithm for Stochastic Bilevel Optimization via Double-Momentum
https://openreview.net/forum?id=HjFtRc83eBB
[ "Prashant Khanduri", "Siliang Zeng", "Mingyi Hong", "Hoi To Wai", "Zhaoran Wang", "Zhuoran Yang" ]
Poster
null
This paper proposes a new algorithm -- the \underline{S}ingle-timescale Do\underline{u}ble-momentum \underline{St}ochastic \underline{A}pprox\underline{i}matio\underline{n} (SUSTAIN) -- for tackling stochastic unconstrained bilevel optimization problems. We focus on bilevel problems where the lower level subproblem is...
[ "stochastic bilevel optimization", "non-convex optimization", "single time-scale stochastic approximation" ]
null
8,265
2102.07367
title_snapshot
uZJJFpFl60W
Best-case lower bounds in online learning
https://openreview.net/forum?id=uZJJFpFl60W
[ "Cristóbal A Guzmán", "Nishant A Mehta", "Ali Mortazavi" ]
Poster
null
Much of the work in online learning focuses on the study of sublinear upper bounds on the regret. In this work, we initiate the study of best-case lower bounds in online convex optimization, wherein we bound the largest \emph{improvement} an algorithm can obtain relative to the single best action in hindsight. This pro...
[ "online learning", "online convex optimization", "learning theory" ]
null
8,264
2106.12688
title_snapshot
ioyq7NsR1KJ
Adversarial Graph Augmentation to Improve Graph Contrastive Learning
https://openreview.net/forum?id=ioyq7NsR1KJ
[ "Susheel Suresh", "Pan Li", "Cong Hao", "Jennifer Neville" ]
Poster
null
Self-supervised learning of graph neural networks (GNN) is in great need because of the widespread label scarcity issue in real-world graph/network data. Graph contrastive learning (GCL), by training GNNs to maximize the correspondence between the representations of the same graph in its different augmented forms, may ...
[ "graph contrastive learning", "self-supervised graph representation learning" ]
Adversarial training to learn augmentation strategies for better self-supervised graph representations.
8,263
2106.05819
title_snapshot
Jyxmk4wUoQV
Foundations of Symbolic Languages for Model Interpretability
https://openreview.net/forum?id=Jyxmk4wUoQV
[ "Marcelo Arenas", "Daniel Báez", "Pablo Barcelo", "Jorge Pérez", "Bernardo Subercaseaux" ]
Spotlight
null
Several queries and scores have recently been proposed to explain individual predictions over ML models. Examples include queries based on “anchors”, which are parts of an instance that are sufficient to justify its classification, and “feature-perturbation” scores such as SHAP. Given the need for flexible, reliable, a...
[ "language", "interpretability", "logic", "complexity", "decision trees", "explainability" ]
We propose a systematic study of model interpretability by considering a minimalistic symbolic query language tailored for interpretability, studying its complexity of evaluation and presenting a prototype with high-level syntax.
8,261
2110.02376
title_snapshot
LZDiWaC9CGL
Multiwavelet-based Operator Learning for Differential Equations
https://openreview.net/forum?id=LZDiWaC9CGL
[ "Gaurav Gupta", "Xiongye Xiao", "Paul Bogdan" ]
Spotlight
null
The solution of a partial differential equation can be obtained by computing the inverse operator map between the input and the solution space. Towards this end, we introduce a $\textit{multiwavelet-based neural operator learning scheme}$ that compresses the associated operator's kernel using fine-grained wavelets. By ...
[ "Neural operators", "multiwavelet transform", "partial differential equations" ]
We propose a novel multiwavelet-based Neural Operator scheme for learning partial differential equations, where the fundamental property of kernel smoothness enables efficient learning by multiwavelets.
8,245
2109.13459
title_snapshot
MjNFN44NbZm
Policy Finetuning: Bridging Sample-Efficient Offline and Online Reinforcement Learning
https://openreview.net/forum?id=MjNFN44NbZm
[ "Tengyang Xie", "Nan Jiang", "Huan Wang", "Caiming Xiong", "Yu Bai" ]
Poster
null
Recent theoretical work studies sample-efficient reinforcement learning (RL) extensively in two settings: learning interactively in the environment (online RL), or learning from an offline dataset (offline RL). However, existing algorithms and theories for learning near-optimal policies in these two settings are rather...
[ "reinforcement learning theory", "offline RL", "online RL", "policy finetuning" ]
We establish sharp upper and lower bounds for sample-efficient RL in policy finetuning, a new RL setting that bridges online and offline reinforcement learning.
8,232
2106.04895
title_snapshot
MQQeeDiO5vv
Combiner: Full Attention Transformer with Sparse Computation Cost
https://openreview.net/forum?id=MQQeeDiO5vv
[ "Hongyu Ren", "Hanjun Dai", "Zihang Dai", "Mengjiao Yang", "Jure Leskovec", "Dale Schuurmans", "Bo Dai" ]
Spotlight
null
Transformers provide a class of expressive architectures that are extremely effective for sequence modeling. However, the key limitation of transformers is their quadratic memory and time complexity $\mathcal{O}(L^2)$ with respect to the sequence length in attention layers, which restricts application in extremely long...
[ "Transformer", "long sequence modeling", "conditional expectation factorization", "full attention" ]
We propose Combiner, a drop-in replacement of attention, achieving the full attention with sub-quadratic cost using structured factorization. The proposed Combiner achieves SOTA on variety of tasks.
8,230
2107.05768
title_snapshot
uJGObgFU0lU
Differentiable Quality Diversity
https://openreview.net/forum?id=uJGObgFU0lU
[ "Matthew Christopher Fontaine", "Stefanos Nikolaidis" ]
Oral
null
Quality diversity (QD) is a growing branch of stochastic optimization research that studies the problem of generating an archive of solutions that maximize a given objective function but are also diverse with respect to a set of specified measure functions. However, even when these functions are differentiable, QD algo...
[ "quality diversity optimization", "generative adversarial network", "latent space exploration" ]
We present the differentiable quality diversity (DQD) problem and the first DQD algorithm.
8,222
2106.03894
title_snapshot
3GpcwM1slH8
Uniform Sampling over Episode Difficulty
https://openreview.net/forum?id=3GpcwM1slH8
[ "Sebastien Arnold", "Guneet Singh Dhillon", "Avinash Ravichandran", "Stefano Soatto" ]
Spotlight
null
Episodic training is a core ingredient of few-shot learning to train models on tasks with limited labelled data. Despite its success, episodic training remains largely understudied, prompting us to ask the question: what is the best way to sample episodes? In this paper, we first propose a method to approximate episode...
[ "episode sampling", "few-shot learning", "meta-learning" ]
How should we sample episodes to improve generalization? Our empirical study suggests that sampling uniformly over episode difficulty outperforms curriculum learning, easy, and hard mining.
8,213
2108.01662
title_snapshot
pMvBiSLGTeU
A Topological Perspective on Causal Inference
https://openreview.net/forum?id=pMvBiSLGTeU
[ "Duligur Ibeling", "Thomas F Icard" ]
Poster
null
This paper presents a topological learning-theoretic perspective on causal inference by introducing a series of topologies defined on general spaces of structural causal models (SCMs). As an illustration of the framework we prove a topological causal hierarchy theorem, showing that substantive assumption-free causal in...
[ "causal inference", "statistical learning theory", "topology" ]
null
8,210
2107.08558
title_snapshot
P4W74BXoyBy
Rot-Pro: Modeling Transitivity by Projection in Knowledge Graph Embedding
https://openreview.net/forum?id=P4W74BXoyBy
[ "Tengwei Song", "Jie Luo", "Lei Huang" ]
Poster
null
Knowledge graph embedding models learn the representations of entities and relations in the knowledge graphs for predicting missing links (relations) between entities. Their effectiveness are deeply affected by the ability of modeling and inferring different relation patterns such as symmetry, asymmetry, inversion, com...
[ "knowledge graph embedding", "transitivity", "projection", "relation pattern" ]
null
8,207
2110.14450
title_snapshot
B2cyX_ht4VI
Sampling with Trusthworthy Constraints: A Variational Gradient Framework
https://openreview.net/forum?id=B2cyX_ht4VI
[ "Xingchao Liu", "Xin Tong", "qiang liu" ]
Poster
null
Sampling-based inference and learning techniques, especially Bayesian inference, provide an essential approach to handling uncertainty in machine learning (ML). As these techniques are increasingly used in daily life, it becomes essential to safeguard the ML systems with various trustworthy-related constraints, such as...
[ "Trustworthy Machine Learning", "Sampling", "Bayesian Inference" ]
null
8,201
null
null
GYr3qnFKgU
Adversarial Intrinsic Motivation for Reinforcement Learning
https://openreview.net/forum?id=GYr3qnFKgU
[ "Ishan Durugkar", "Mauricio Tec", "Scott Niekum", "Peter Stone" ]
Poster
null
Learning with an objective to minimize the mismatch with a reference distribution has been shown to be useful for generative modeling and imitation learning. In this paper, we investigate whether one such objective, the Wasserstein-1 distance between a policy's state visitation distribution and a target distribution, c...
[ "Reinforcement Learning", "Intrinsic Motivation", "Distribution Mismatch", "Wasserstein distance" ]
An intrinsic reward function based on the Wasserstein-1 distance of the agent's state visitation distribution to the goal accelerates goal-conditioned reinforcement learning
8,200
2105.13345
title_snapshot
5J9sbGwZ9bC
Indexed Minimum Empirical Divergence for Unimodal Bandits
https://openreview.net/forum?id=5J9sbGwZ9bC
[ "Hassan SABER", "Pierre MENARD", "Odalric-Ambrym Maillard" ]
Poster
null
We consider a stochastic multi-armed bandit problem specified by a set of one-dimensional family exponential distributions endowed with a unimodal structure. The unimodal structure is of practical relevance for several applications. We introduce IMED-UB, an algorithm that exploits provably optimally the unimodal-struct...
[ "Multi-Armed Bandit", "Indexed Minimum Empirical Divergence", "Unimodal Bandits", "Optimal Algorithm", "One-Dimensional Exponential Family Distributions", "Regret analysis" ]
Optimal strategy for the unimodal bandit problem with one-dimentional family distributions. Elegant proof and improved practical performances.
8,195
2112.01452
title_snapshot
BYrJYl1rexa
Unsupervised Representation Transfer for Small Networks: I Believe I Can Distill On-the-Fly
https://openreview.net/forum?id=BYrJYl1rexa
[ "Hee Min Choi", "Hyoa Kang", "Dokwan Oh" ]
Poster
null
A current remarkable improvement of unsupervised visual representation learning is based on heavy networks with large-batch training. While recent methods have greatly reduced the gap between supervised and unsupervised performance of deep models such as ResNet-50, this development has been relatively limited for small...
[ "Representation learning", "knowledge distillation", "unsupervised learning" ]
We propose a novel unsupervised representation learning framework for small networks that combines deep self-supervised representation learning and knowledge distillation within one-phase training.
8,187
null
null
-1rrzmJCp4
SPANN: Highly-efficient Billion-scale Approximate Nearest Neighborhood Search
https://openreview.net/forum?id=-1rrzmJCp4
[ "Qi Chen", "Bing Zhao", "Haidong Wang", "Mingqin Li", "Chuanjie Liu", "Zengzhong Li", "Mao Yang", "Jingdong Wang" ]
Spotlight
null
The in-memory algorithms for approximate nearest neighbor search (ANNS) have achieved great success for fast high-recall search, but are extremely expensive when handling very large scale database. Thus, there is an increasing request for the hybrid ANNS solutions with small memory and inexpensive solid-state drive (SS...
[ "billion-scale", "vector search", "inverted index solution" ]
SPANN: Highly-efficient Billion-scale Approximate Nearest Neighbor Search
8,181
2111.08566
title_judge
l2UWXn5iBQI
Exponential Graph is Provably Efficient for Decentralized Deep Training
https://openreview.net/forum?id=l2UWXn5iBQI
[ "Bicheng Ying", "Kun Yuan", "Yiming Chen", "Hanbin Hu", "Pan Pan", "Wotao Yin" ]
Poster
null
Decentralized SGD is an emerging training method for deep learning known for its much less (thus faster) communication per iteration, which relaxes the averaging step in parallel SGD to inexact averaging. The less exact the averaging is, however, the more the total iterations the training needs to take. Therefore, the ...
[ "Decentralized optimization", "distributed and parallel algorithms", "stochastic optimization", "deep learning" ]
We theoretically establish that static and one-peer exponential graphs endow decentralized SGD with fast and high-quality training performance
8,177
2110.13363
title_snapshot
Yc4AdP1M9kT
Learning to Simulate Self-driven Particles System with Coordinated Policy Optimization
https://openreview.net/forum?id=Yc4AdP1M9kT
[ "Zhenghao Peng", "Quanyi Li", "Ka Ming Hui", "Chunxiao Liu", "Bolei Zhou" ]
Poster
null
Self-Driven Particles (SDP) describe a category of multi-agent systems common in everyday life, such as flocking birds and traffic flows. In a SDP system, each agent pursues its own goal and constantly changes its cooperative or competitive behaviors with its nearby agents. Manually designing the controllers for such S...
[ "Multi-agent Reinforcement Learning", "Traffic Simulation" ]
We propose Coordinated Policy Optimization (CoPO) which achieves superior performance in 5 new traffic simulation tasks and learns complex and diverse social behaviors.
8,171
2110.13827
title_snapshot
CuQoImkKkIj
Robust and differentially private mean estimation
https://openreview.net/forum?id=CuQoImkKkIj
[ "Xiyang Liu", "Weihao Kong", "Sham M. Kakade", "Sewoong Oh" ]
Poster
null
In statistical learning and analysis from shared data, which is increasingly widely adopted in platforms such as federated learning and meta-learning, there are two major concerns: privacy and robustness. Each participating individual should be able to contribute without the fear of leaking one's sensitive information....
[ "differential privacy", "robust mean estimation" ]
We provide the first family of algorithms for mean estimation achieving both robustness and differential privacy.
8,166
2102.09159
title_snapshot
zdmF437BCB
Domain Adaptation with Invariant Representation Learning: What Transformations to Learn?
https://openreview.net/forum?id=zdmF437BCB
[ "Petar Stojanov", "Zijian Li", "Mingming Gong", "Ruichu Cai", "Jaime G. Carbonell", "Kun Zhang" ]
Poster
null
Unsupervised domain adaptation, as a prevalent transfer learning setting, spans many real-world applications. With the increasing representational power and applicability of neural networks, state-of-the-art domain adaptation methods make use of deep architectures to map the input features $X$ to a latent representatio...
[ "Domain adaptation", "transfer learning", "deep learning", "adversarial training", "autencoders" ]
We make use of the data-generating process in order to perform unsupervised domain adaptation via neural architectures and invariant representation learning..
8,165
null
null
nqutwR1WDBY
On The Structure of Parametric Tournaments with Application to Ranking from Pairwise Comparisons
https://openreview.net/forum?id=nqutwR1WDBY
[ "Vishnu Veerathu", "Arun Rajkumar" ]
Poster
null
We consider the classical problem of finding the minimum feedback arc set on tournaments (MFAST). The problem is NP-hard in general and we study it for important classes of tournaments that arise naturally in the problem of learning to rank from pairwise comparisons. Specifically, we consider tournaments classes that ...
[ "ranking", "pairwise", "intransitive", "low-rank" ]
We solve the minimum feedback arc set in poly-time for a class of tournaments that arise in learning to rank and use it to develop flexible parametric pairwise preference models that can model intransitivity.
8,161
null
null
1fr3bOX2t69
Curriculum Learning for Vision-and-Language Navigation
https://openreview.net/forum?id=1fr3bOX2t69
[ "Jiwen Zhang", "zhongyu wei", "Jianqing Fan", "Jiajie Peng" ]
Poster
null
Vision-and-Language Navigation (VLN) is a task where an agent navigates in an embodied indoor environment under human instructions. Previous works ignore the distribution of sample difficulty and we argue that this potentially degrade their agent performance. To tackle this issue, we propose a novel curriculum- based t...
[ "Vision-and-language navigation", "curriculum learning", "multimodal learning" ]
We propose to use an efficient curriculum learning method for vision-and-language navigation task.
8,156
2111.07228
title_snapshot
h596lT4RAH4
Batched Thompson Sampling
https://openreview.net/forum?id=h596lT4RAH4
[ "Cem Kalkanli", "Ayfer Ozgur" ]
Poster
null
We introduce a novel anytime batched Thompson sampling policy for multi-armed bandits where the agent observes the rewards of her actions and adjusts her policy only at the end of a small number of batches. We show that this policy simultaneously achieves a problem dependent regret of order $O(\log(T))$ and a minimax r...
[ "Thompson Sampling", "Batched Multi-Armed Bandits", "Stochastic Multi-Armed Bandits" ]
null
8,155
2110.00202
title_snapshot
HfpNVDg3ExA
Probabilistic Transformer For Time Series Analysis
https://openreview.net/forum?id=HfpNVDg3ExA
[ "Binh Tang", "David S. Matteson" ]
Poster
null
Generative modeling of multivariate time series has remained challenging partly due to the complex, non-deterministic dynamics across long-distance timesteps. In this paper, we propose deep probabilistic methods that combine state-space models (SSMs) with transformer architectures. In contrast to previously proposed SS...
[ "state space models", "transformer", "time series forecasting", "human motion prediction", "video prediction" ]
null
8,150
null
null
fj6rFciApc
The staircase property: How hierarchical structure can guide deep learning
https://openreview.net/forum?id=fj6rFciApc
[ "Emmanuel Abbe", "Enric Boix-Adserà", "Matthew Stewart Brennan", "Guy Bresler", "Dheeraj Mysore Nagaraj" ]
Poster
null
This paper identifies a structural property of data distributions that enables deep neural networks to learn hierarchically. We define the ``staircase'' property for functions over the Boolean hypercube, which posits that high-order Fourier coefficients are reachable from lower-order Fourier coefficients along increasi...
[ "deep learning", "hierarchical", "stochastic gradient descent" ]
This paper identifies a structural property of data distributions that enables deep neural networks to learn hierarchically.
8,140
2108.10573
title_snapshot
1Av2E0EugkA
Local Explanation of Dialogue Response Generation
https://openreview.net/forum?id=1Av2E0EugkA
[ "Yi-Lin Tuan", "Connor Pryor", "Wenhu Chen", "Lise Getoor", "William Yang Wang" ]
Poster
null
In comparison to the interpretation of classification models, the explanation of sequence generation models is also an important problem, however it has seen little attention. In this work, we study model-agnostic explanations of a representative text generation task -- dialogue response generation. Dialog response gen...
[ "explainable artificial intelligence", "dialogue response generation", "text generation" ]
null
8,134
2106.06528
title_snapshot
o2mbl-Hmfgd
Intriguing Properties of Vision Transformers
https://openreview.net/forum?id=o2mbl-Hmfgd
[ "Muzammal Naseer", "Kanchana Ranasinghe", "Salman Khan", "Munawar Hayat", "Fahad Khan", "Ming-Hsuan Yang" ]
Spotlight
null
Vision transformers (ViT) have demonstrated impressive performance across numerous machine vision tasks. These models are based on multi-head self-attention mechanisms that can flexibly attend to a sequence of image patches to encode contextual cues. An important question is how such flexibility (in attending image-wid...
[ "Vision Transformers", "Auto-segmentation", "Off-the-shelf-features", "Robustness", "Shape-Modeling" ]
Analysis of content-dependent long-range interaction modeling capabilities of Vision Transformers in terms of robustness against image nuisances such as severe occlusions, domain shifts, spatial permutations, adversarial and natural perturbations.
8,129
2105.10497
title_snapshot
IQdzjJtUGFl
Extracting Deformation-Aware Local Features by Learning to Deform
https://openreview.net/forum?id=IQdzjJtUGFl
[ "Guilherme Potje", "Renato Martins", "Felipe C. Chamone", "Erickson Rangel Nascimento" ]
Poster
null
Despite the advances in extracting local features achieved by handcrafted and learning-based descriptors, they are still limited by the lack of invariance to non-rigid transformations. In this paper, we present a new approach to compute features from still images that are robust to non-rigid deformations to circumvent ...
[ "Local descriptors", "Non-rigid matching", "Deformation", "Spatial Transformers" ]
An end-to-end learned deformation-aware descriptor to extract local image features robust to non-rigid deformations.
8,127
2111.10617
title_snapshot
ZBhZDNaiww
3DP3: 3D Scene Perception via Probabilistic Programming
https://openreview.net/forum?id=ZBhZDNaiww
[ "Nishad Gothoskar", "Marco Cusumano-Towner", "Ben Zinberg", "Matin Ghavamizadeh", "Falk Pollok", "Austin James Garrett", "Joshua B. Tenenbaum", "Dan Gutfreund", "Vikash Mansinghka" ]
Poster
null
We present 3DP3, a framework for inverse graphics that uses inference in a structured generative model of objects, scenes, and images. 3DP3 uses (i) voxel models to represent the 3D shape of objects, (ii) hierarchical scene graphs to decompose scenes into objects and the contacts between them, and (iii) depth image lik...
[ "probabilistic programming", "approximate inference", "scene understanding", "object pose estimation" ]
3D Vision Architecture based on Probabilistic Programs
8,126
2111.00312
title_snapshot
2BbDxFtDht7
Duplex Sequence-to-Sequence Learning for Reversible Machine Translation
https://openreview.net/forum?id=2BbDxFtDht7
[ "Zaixiang Zheng", "Hao Zhou", "Shujian Huang", "Jiajun Chen", "Jingjing Xu", "Lei Li" ]
Poster
null
Sequence-to-sequence learning naturally has two directions. How to effectively utilize supervision signals from both directions? Existing approaches either require two separate models, or a multitask-learned model but with inferior performance. In this paper, we propose REDER (Reversible Duplex Transformer), a paramete...
[ "duplex network", "sequence-to-sequence learning", "reversible machine translation" ]
null
8,121
2105.03458
title_snapshot
6YL_BntJrz6
Dirichlet Energy Constrained Learning for Deep Graph Neural Networks
https://openreview.net/forum?id=6YL_BntJrz6
[ "Kaixiong Zhou", "Xiao Huang", "Daochen Zha", "Rui Chen", "Li Li", "Soo-Hyun Choi", "Xia Hu" ]
Poster
null
Graph neural networks (GNNs) integrate deep architectures and topological structure modeling in an effective way. However, the performance of existing GNNs would decrease significantly when they stack many layers, because of the over-smoothing issue. Node embeddings tend to converge to similar vectors when GNNs keep re...
[ "Deep graph neural networks", "Over-smoothing", "Node classification", "Dirichlet energy", "Orthogonal weight" ]
Theoretical and empirical framework to regularize the training of deep graph neural networks
8,113
2107.02392
title_snapshot
o6s1b_-nDOE
Counterfactual Maximum Likelihood Estimation for Training Deep Networks
https://openreview.net/forum?id=o6s1b_-nDOE
[ "Xinyi Wang", "Wenhu Chen", "Michael Saxon", "William Yang Wang" ]
Poster
null
Although deep learning models have driven state-of-the-art performance on a wide array of tasks, they are prone to spurious correlations that should not be learned as predictive clues. To mitigate this problem, we propose a causality-based training framework to reduce the spurious correlations caused by observed confou...
[ "causality", "spurious correlation", "deep learning", "data augmentation", "out-of-distribution generalization", "natural language processing" ]
We propose a causality-based training framework for deep learning models to reduce the spurious correlations caused by observable confounders, with theoretical analysis as well as experiments on both simulated and real-world datasets.
8,109
2106.03831
title_snapshot
x8qirBbT9xp
Causal Inference for Event Pairs in Multivariate Point Processes
https://openreview.net/forum?id=x8qirBbT9xp
[ "Tian Gao", "Dharmashankar Subramanian", "Debarun Bhattacharjya", "Xiao Shou", "Nicholas Mattei", "Kristin Bennett" ]
Poster
null
Causal inference and discovery from observational data has been extensively studied across multiple fields. However, most prior work has focused on independent and identically distributed (i.i.d.) data. In this paper, we propose a formalization for causal inference between pairs of event variables in multivariate recur...
[ "causal inference", "point process" ]
We propose a framework for causal inference for multivariate point process for event datasets.
8,104
null
null
RwdHpzTTGl
Fast Tucker Rank Reduction for Non-Negative Tensors Using Mean-Field Approximation
https://openreview.net/forum?id=RwdHpzTTGl
[ "Kazu Ghalamkari", "Mahito Sugiyama" ]
Poster
null
We present an efficient low-rank approximation algorithm for non-negative tensors. The algorithm is derived from our two findings: First, we show that rank-1 approximation for tensors can be viewed as a mean-field approximation by treating each tensor as a probability distribution. Second, we theoretically provide a s...
[ "Low-rank approximation", "Tucker rank", "tensor", "mean-field theory", "Information geometry" ]
A novel fast Tucker rank reduction method based on mean-field approximation for non-negative tensors
8,099
2103.02898
title_snapshot
iQICgKcrGpE
On the Existence of The Adversarial Bayes Classifier
https://openreview.net/forum?id=iQICgKcrGpE
[ "Pranjal Awasthi", "Natalie Frank", "Mehryar Mohri" ]
Spotlight
null
Adversarial robustness is a critical property in a variety of modern machine learning applications. While it has been the subject of several recent theoretical studies, many important questions related to adversarial robustness are still open. In this work, we study a fundamental question regarding Bayes optimality fo...
[ "Adversarial Learning", "Learning Theory", "Consistency", "Calibration", "Bayes classifiers" ]
We prove that in many typical adversarial learning scenarios, there exists a classifier that minimizes the adversarial loss
8,088
2112.01694
title_judge
DTVfEJIL3DB
MADE: Exploration via Maximizing Deviation from Explored Regions
https://openreview.net/forum?id=DTVfEJIL3DB
[ "Tianjun Zhang", "Paria Rashidinejad", "Jiantao Jiao", "Yuandong Tian", "Joseph E. Gonzalez", "Stuart Russell" ]
Poster
null
In online reinforcement learning (RL), efficient exploration remains particularly challenging in high-dimensional environments with sparse rewards. In low-dimensional environments, where tabular parameterization is possible, count-based upper confidence bound (UCB) exploration methods achieve minimax near-optimal rates...
[ "Deep reinforcement learning", "exploration", "intrinsic reward" ]
We propose a simple and practical exploration algorithm for deep reinforcement learning.
8,085
2106.10268
title_snapshot
DHnThtAyoPj
Gone Fishing: Neural Active Learning with Fisher Embeddings
https://openreview.net/forum?id=DHnThtAyoPj
[ "Jordan T. Ash", "Surbhi Goel", "Akshay Krishnamurthy", "Sham M. Kakade" ]
Poster
null
There is an increasing need for effective active learning algorithms that are compatible with deep neural networks. This paper motivates and revisits a classic, Fisher-based active selection objective, and proposes BAIT, a practical, tractable, and high-performing algorithm that makes it viable for use with neural mode...
[ "deep active learning", "deep learning", "active learning" ]
We motivate and revisit a classic objective for active learning and show its effectiveness in neural networks for both classification and regression.
8,081
2106.09675
title_snapshot
pDgN3l3EAF
Learning Collaborative Policies to Solve NP-hard Routing Problems
https://openreview.net/forum?id=pDgN3l3EAF
[ "Minsu Kim", "Jinkyoo Park", "Joungho Kim" ]
Poster
null
Recently, deep reinforcement learning (DRL) frameworks have shown potential for solving NP-hard routing problems such as the traveling salesman problem (TSP) without problem-specific expert knowledge. Although DRL can be used to solve complex problems, DRL frameworks still struggle to compete with state-of-the-art heur...
[ "Deep reinforcement learning", "NP-hard", "Policy collaboration", "Hierarchical solving strategy" ]
We propose a novel method of hierarchical problem-solving strategy for NP-hard routing problems with two collaborative policies that substantially gain performance.
8,078
2110.13987
title_snapshot
pSitk34qYit
Risk Bounds and Calibration for a Smart Predict-then-Optimize Method
https://openreview.net/forum?id=pSitk34qYit
[ "Heyuan Liu", "Paul Grigas" ]
Poster
null
The predict-then-optimize framework is fundamental in practical stochastic decision-making problems: first predict unknown parameters of an optimization model, then solve the problem using the predicted values. A natural loss function in this setting is defined by measuring the decision error induced by the predicted ...
[ "contextual stochastic optimization", "predict-then-optimize", "learning theory", "risk bounds" ]
null
8,077
2108.08887
title_snapshot
9-sCrvMbL9
Offline Constrained Multi-Objective Reinforcement Learning via Pessimistic Dual Value Iteration
https://openreview.net/forum?id=9-sCrvMbL9
[ "Runzhe Wu", "Yufeng Zhang", "Zhuoran Yang", "Zhaoran Wang" ]
Poster
null
In constrained multi-objective RL, the goal is to learn a policy that achieves the best performance specified by a multi-objective preference function under a constraint. We focus on the offline setting where the RL agent aims to learn the optimal policy from a given dataset. This scenario is common in real-world appli...
[ "reinforcement learning", "offline learning", "multiple objectives" ]
null
8,070
null
null
4h4oqp-ATxb
Neural Symplectic Form: Learning Hamiltonian Equations on General Coordinate Systems
https://openreview.net/forum?id=4h4oqp-ATxb
[ "Yuhan Chen", "Takashi Matsubara", "Takaharu Yaguchi" ]
Spotlight
null
In recent years, substantial research on the methods for learning Hamiltonian equations has been conducted. Although these approaches are very promising, the commonly used representation of the Hamilton equation uses the generalized momenta, which are generally unknown. Therefore, the training data must be represented ...
[ "Hamiltonian Neural Networks", "Hamiltonian Mechanics", "Symplectic Geometry" ]
We propose an efficient method for learning Hamiltonian equations in general coordinate systems by using a coordinate-free representation used in symplectic geometry.
8,066
null
null
9CPc4EIr2t1
Stable Neural ODE with Lyapunov-Stable Equilibrium Points for Defending Against Adversarial Attacks
https://openreview.net/forum?id=9CPc4EIr2t1
[ "QIYU KANG", "Yang Song", "Qinxu Ding", "Wee Peng Tay" ]
Poster
null
Deep neural networks (DNNs) are well-known to be vulnerable to adversarial attacks, where malicious human-imperceptible perturbations are included in the input to the deep network to fool it into making a wrong classification. Recent studies have demonstrated that neural Ordinary Differential Equations (ODEs) are intri...
[ "ODE", "stable", "Ordinary Differential Equation", "adversarial attack" ]
null
8,056
2110.12976
title_snapshot
zL1szwVKdwc
The Elastic Lottery Ticket Hypothesis
https://openreview.net/forum?id=zL1szwVKdwc
[ "Xiaohan Chen", "Yu Cheng", "Shuohang Wang", "Zhe Gan", "Jingjing Liu", "Zhangyang Wang" ]
Poster
null
Lottery Ticket Hypothesis (LTH) raises keen attention to identifying sparse trainable subnetworks, or winning tickets, which can be trained in isolation to achieve similar or even better performance compared to the full models. Despite many efforts being made, the most effective method to identify such winning tickets ...
[ "Lottery Ticket Hypothesis" ]
null
8,053
2103.16547
title_snapshot
AWadl8GeCEG
Dynamic Normalization and Relay for Video Action Recognition
https://openreview.net/forum?id=AWadl8GeCEG
[ "Dongqi Cai", "Anbang Yao", "Yurong Chen" ]
Poster
null
Convolutional Neural Networks (CNNs) have been the dominant model for video action recognition. Due to the huge memory and compute demand, popular action recognition networks need to be trained with small batch sizes, which makes learning discriminative spatial-temporal representations for videos become a challenging p...
[ "Video Action Recognition", "Batch Normormalization", "Group Normalization", "Convolutional Neural Networks", "Recurrent Neural Networks" ]
This paper presents a dynamic normalization and relay method to improve deep convolutional neural networks for video action recognition.
8,043
null
null
mvcIGGWXPOV
Infinite Time Horizon Safety of Bayesian Neural Networks
https://openreview.net/forum?id=mvcIGGWXPOV
[ "Mathias Lechner", "Đorđe Žikelić", "Krishnendu Chatterjee", "Thomas A Henzinger" ]
Poster
null
Bayesian neural networks (BNNs) place distributions over the weights of a neural network to model uncertainty in the data and the network's prediction. We consider the problem of verifying safety when running a Bayesian neural network policy in a feedback loop with infinite time horizon systems. Compared to the existin...
[ "Bayesian neural networks", "BNN", "safety verification" ]
Verification of Bayesian neural networks in a feedback loop with infinite time horizon systems
8,040
2111.03165
title_snapshot
IuH1bVRgvHH
Iterative Connecting Probability Estimation for Networks
https://openreview.net/forum?id=IuH1bVRgvHH
[ "Yichen Qin", "Linhan Yu", "Yang Li" ]
Poster
null
Estimating the probabilities of connections between vertices in a random network using an observed adjacency matrix is an important task for network data analysis. Many existing estimation methods are based on certain assumptions on network structure, which limit their applicability in practice. Without making strong a...
[ "Graphon", "vertex similarity", "link probability", "edge probability" ]
We propose an iterative connecting probability estimation method for random networks.
8,039
null
null
6vWuYzkp8d
Discovering and Achieving Goals via World Models
https://openreview.net/forum?id=6vWuYzkp8d
[ "Russell Mendonca", "Oleh Rybkin", "Kostas Daniilidis", "Danijar Hafner", "Deepak Pathak" ]
Poster
null
How can artificial agents learn to solve many diverse tasks in complex visual environments without any supervision? We decompose this question into two challenges: discovering new goals and learning to reliably achieve them. Our proposed agent, Latent Explorer Achiever (LEXA), addresses both challenges by learning a wo...
[ "unsupervised goal reaching", "unsupervised rl", "goal-conditioned rl", "exploration", "model-based rl", "world models" ]
null
8,034
2110.09514
title_snapshot
5-iRjd9FreV
Analysis of Sensing Spectral for Signal Recovery under a Generalized Linear Model
https://openreview.net/forum?id=5-iRjd9FreV
[ "Junjie Ma", "Ji Xu", "Arian Maleki" ]
Poster
null
We consider a nonlinear inverse problem $\mathbf{y}= f(\mathbf{Ax})$, where observations $\mathbf{y} \in \mathbb{R}^m$ are the componentwise nonlinear transformation of $\mathbf{Ax} \in \mathbb{R}^m$, $\mathbf{x} \in \mathbb{R}^n$ is the signal of interest and $\mathbf{A}$ is a known linear mapping. By properly specify...
[ "phase retrieval", "compressed sensing", "approximate message passing", "statistical physics" ]
null
8,029
null
null