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
mSuBvrUJFsF
Double Machine Learning Density Estimation for Local Treatment Effects with Instruments
https://openreview.net/forum?id=mSuBvrUJFsF
[ "Yonghan Jung", "Jin Tian", "Elias Bareinboim" ]
Spotlight
null
Local treatment effects are a common quantity found throughout the empirical sciences that measure the treatment effect among those who comply with what they are assigned. Most of the literature is focused on estimating the average of such quantity, which is called the ``local average treatment effect (LATE)'' [Imbens ...
[ "instrumental variables", "local treatment effect", "complier average causal effects", "density estimation", "double/debiased machine learning", "semiparametric inference" ]
We develop double/debiased machine learning (DML) density estimators for treatment effects among compliers in the presence of instruments.
10,716
null
null
Uj7pF-D-YvT
Deep Learning on a Data Diet: Finding Important Examples Early in Training
https://openreview.net/forum?id=Uj7pF-D-YvT
[ "Mansheej Paul", "Surya Ganguli", "Gintare Karolina Dziugaite" ]
Poster
null
Recent success in deep learning has partially been driven by training increasingly overparametrized networks on ever larger datasets. It is therefore natural to ask: how much of the data is superfluous, which examples are important for generalization, and how do we find them? In this work, we make the striking observat...
[ "data pruning", "neural network training dynamics", "understanding deep learning", "data compression", "coresets" ]
We propose a method for identifying a significantly smaller set of training data early in training, which can be used to achieve test accuracy comparable to training on the full dataset.
10,714
2107.07075
title_snapshot
AzmEMstdf3o
Passive attention in artificial neural networks predicts human visual selectivity
https://openreview.net/forum?id=AzmEMstdf3o
[ "Thomas A Langlois", "Haicheng Charles Zhao", "Erin Grant", "Ishita Dasgupta", "Thomas L. Griffiths", "Nori Jacoby" ]
Oral
null
Developments in machine learning interpretability techniques over the past decade have provided new tools to observe the image regions that are most informative for classification and localization in artificial neural networks (ANNs). Are the same regions similarly informative to human observers? Using data from 79 new...
[ "Cognition", "Attention", "Interpretable AI", "Computer Vision", "Human Visual Perception" ]
We compare ANNs to humans using attention visualization techniques. We find that gradient-based attention visualizations obtained from a select class of networks predict human attention and saliency estimates derived from 6 distinct behavioral tasks.
10,708
2107.07013
title_snapshot
DE8MOQIgFTK
Adversarial Examples Make Strong Poisons
https://openreview.net/forum?id=DE8MOQIgFTK
[ "Liam H Fowl", "Micah Goldblum", "Ping-yeh Chiang", "Jonas Geiping", "Wojciech Czaja", "Tom Goldstein" ]
Poster
null
The adversarial machine learning literature is largely partitioned into evasion attacks on testing data and poisoning attacks on training data. In this work, we show that adversarial examples, originally intended for attacking pre-trained models, are even more effective for data poisoning than recent methods designed ...
[ "Data Poisoning", "Robustness", "Security", "Adversarial Examples", "Data Release", "Availability Attack" ]
We find that adversarial examples make stronger availability poisons than other methods designed specifically for data poisoning.
10,706
2106.10807
title_snapshot
_eXwwWOyqT_
Practical Large-Scale Linear Programming using Primal-Dual Hybrid Gradient
https://openreview.net/forum?id=_eXwwWOyqT_
[ "David Applegate", "Mateo Diaz Diaz", "Oliver Hinder", "Haihao Lu", "Miles Lubin", "Brendan O'Donoghue", "Warren Schudy" ]
Poster
null
We present PDLP, a practical first-order method for linear programming (LP) that can solve to the high levels of accuracy that are expected in traditional LP applications. In addition, it can scale to very large problems because its core operation is matrix-vector multiplications. PDLP is derived by applying the primal...
[ "linear programming", "convex optimization", "operations research", "first-order methods", "large scale optimization" ]
We develop a practically effective first-order method for solving linear programming problems to high accuracy.
10,693
2106.04756
title_snapshot
BKeJmkspvc
Leveraging Spatial and Temporal Correlations in Sparsified Mean Estimation
https://openreview.net/forum?id=BKeJmkspvc
[ "Divyansh Jhunjhunwala", "Ankur Mallick", "Advait Harshal Gadhikar", "Swanand Kadhe", "Gauri Joshi" ]
Poster
null
We study the problem of estimating at a central server the mean of a set of vectors distributed across several nodes (one vector per node). When the vectors are high-dimensional, the communication cost of sending entire vectors may be prohibitive, and it may be imperative for them to use sparsification techniques. Whil...
[ "Communication-Efficient Mean Estimation", "Vector Sparsification", "Distributed Optimization" ]
We design estimators that leverage spatial and temporal correlations to improve sparsified mean estimation.
10,691
2110.07751
title_snapshot
405l3VpbqRA
Post-Contextual-Bandit Inference
https://openreview.net/forum?id=405l3VpbqRA
[ "Aurelien Bibaut", "Maria Dimakopoulou", "Nathan Kallus", "Antoine Chambaz", "Mark van der Laan" ]
Poster
null
Contextual bandit algorithms are increasingly replacing non-adaptive A/B tests in e-commerce, healthcare, and policymaking because they can both improve outcomes for study participants and increase the chance of identifying good or even best policies. To support credible inference on novel interventions at the end of t...
[ "contextual bandit", "causal inference", "confidence intervals", "adaptive data collection" ]
We propose a novel asymptotically normal estimator for evaluating a contextual policy from context-dependent adaptively collected data, leading to confidence intervals with correct coverage when existing estimators do not.
10,679
2106.00418
title_snapshot
KLS346_Asf
What can linearized neural networks actually say about generalization?
https://openreview.net/forum?id=KLS346_Asf
[ "Guillermo Ortiz-Jimenez", "Seyed-Mohsen Moosavi-Dezfooli", "Pascal Frossard" ]
Poster
null
For certain infinitely-wide neural networks, the neural tangent kernel (NTK) theory fully characterizes generalization, but for the networks used in practice, the empirical NTK only provides a rough first-order approximation. Still, a growing body of work keeps leveraging this approximation to successfully analyze impo...
[ "Deep learning theory", "NTK", "generalization", "inductive bias", "linearized networks", "neural tangent kernel" ]
We discover that neural networks do not always perform better than their kernel approximations, and reveal that the performance gap heavily depends on architecture, dataset size and training task.
10,673
2106.06770
title_snapshot
yGKi6deX8bX
Towards understanding retrosynthesis by energy-based models
https://openreview.net/forum?id=yGKi6deX8bX
[ "Ruoxi Sun", "Hanjun Dai", "Li Li", "Steven Kearnes", "Bo Dai" ]
Poster
null
Retrosynthesis is the process of identifying a set of reactants to synthesize a target molecule. It is of vital importance to material design and drug discovery. Existing machine learning approaches based on language models and graph neural networks have achieved encouraging results. However, the inner connections of t...
[ "chemical application", "energy-based model" ]
We explore various EBM designs for retrosynthesis, and show some designs outperform the state of the art.
10,672
null
null
Y8YqrYeFftd
Robustness between the worst and average case
https://openreview.net/forum?id=Y8YqrYeFftd
[ "Leslie Rice", "Anna Bair", "Huan Zhang", "J Zico Kolter" ]
Poster
null
Several recent works in machine learning have focused on evaluating the test-time robustness of a classifier: how well the classifier performs not just on the target domain it was trained upon, but upon perturbed examples. In these settings, the focus has largely been on two extremes of robustness: the robustness to p...
[ "adversarial robustness", "data augmentation", "Hamiltonian Monte Carlo", "path sampling" ]
We define a new metric for evaluating classifier robustness by introducing an interpolation between robustness over random and adversarial perturbations, and propose an MCMC-based sampler to effectively evaluate this metric.
10,671
null
null
gviX23L1bqw
Online learning in MDPs with linear function approximation and bandit feedback.
https://openreview.net/forum?id=gviX23L1bqw
[ "Gergely Neu", "Julia Olkhovskaya" ]
Poster
null
We consider the problem of online learning in an episodic Markov decision process, where the reward function is allowed to change between episodes in an adversarial manner and the learner only observes the rewards associated with its actions. We assume that rewards and the transition function can be represented as line...
[ "reinforcement learning", "online learning", "Markov decision processes", "bandit feedback" ]
null
10,670
2007.01612
title_snapshot
WtmMyno9Tq2
Multimodal Few-Shot Learning with Frozen Language Models
https://openreview.net/forum?id=WtmMyno9Tq2
[ "Maria Tsimpoukelli", "Jacob Menick", "Serkan Cabi", "S. M. Ali Eslami", "Oriol Vinyals", "Felix Hill" ]
Poster
null
When trained at sufficient scale, auto-regressive language models exhibit the notable ability to learn a new language task after being prompted with just a few examples. Here, we present a simple, yet effective, approach for transferring this few-shot learning ability to a multimodal setting (vision and language). Usin...
[ "Language Modeling", "Multimodal", "Few-shot Learning" ]
We present a simple approach for transferring abilities of a frozen language model to a multi-modal setting (vision and language).
10,662
2106.13884
title_snapshot
RHZs3GqLBwg
Time-series Generation by Contrastive Imitation
https://openreview.net/forum?id=RHZs3GqLBwg
[ "Daniel Jarrett", "Ioana Bica", "Mihaela van der Schaar" ]
Poster
null
Consider learning a generative model for time-series data. The sequential setting poses a unique challenge: Not only should the generator capture the *conditional* dynamics of (stepwise) transitions, but its open-loop rollouts should also preserve the *joint* distribution of (multi-step) trajectories. On one hand, auto...
[ "Time Series Data", "Synthetic Data Generation" ]
We invite an analogy between synthetic time-series generation and imitation learning, and explore a framework that fleshes out this connection.
10,659
2311.01388
title_snapshot
aedFIIRRfXr
Similarity and Matching of Neural Network Representations
https://openreview.net/forum?id=aedFIIRRfXr
[ "Adrián Csiszárik", "Péter Kőrösi-Szabó", "Ákos K. Matszangosz", "Gergely Papp", "Dániel Varga" ]
Poster
null
We employ a toolset --- dubbed Dr. Frankenstein --- to analyse the similarity of representations in deep neural networks. With this toolset we aim to match the activations on given layers of two trained neural networks by joining them with a stitching layer. We demonstrate that the inner representations emerging in dee...
[ "Representation learning", "Image representations" ]
We investigate notions of neural representation similarity by matching representations via stitching networks together.
10,640
2110.14633
title_snapshot
RcIorZrz88d
Stochastic bandits with groups of similar arms.
https://openreview.net/forum?id=RcIorZrz88d
[ "Fabien Pesquerel", "Hassan SABER", "Odalric-Ambrym Maillard" ]
Poster
null
We consider a variant of the stochastic multi-armed bandit problem where arms are known to be organized into different groups having the same mean. The groups are unknown but a lower bound $q$ on their size is known. This situation typically appears when each arm can be described with a list of categorical attributes, ...
[ "Multi-armed Bandit", "Indexed Minimum Empirical Divergence", "Structure", "Regret analysis", "dependent arms" ]
Structured bandit with group of arms having same mean. Efficient relaxation avoiding combinatorial optimization with controlled near-optimal regret.
10,634
null
null
VsUQQkpEXgr
Sifting through the noise: Universal first-order methods for stochastic variational inequalities
https://openreview.net/forum?id=VsUQQkpEXgr
[ "Kimon Antonakopoulos", "Thomas Pethick", "Ali Kavis", "Panayotis Mertikopoulos", "Volkan Cevher" ]
Poster
null
We examine a flexible algorithmic framework for solving monotone variational inequalities in the presence of randomness and uncertainty. The proposed template encompasses a wide range of popular first-order methods, including dual averaging, dual extrapolation and optimistic gradient algorithms – both adaptive and non-...
[ "monotone variational inequalities", "stochastic variational inequalities", "adaptive methods", "improved rates" ]
We propose a flexible algorithmic framework for solving monotone variational inequalities in the presence of randomness and uncertainty.
10,624
null
null
fzkU-UMKJIv
Directed Probabilistic Watershed
https://openreview.net/forum?id=fzkU-UMKJIv
[ "Enrique Fita Sanmartín", "Sebastian Damrich", "Fred A Hamprecht" ]
Poster
null
The Probabilistic Watershed is a semi-supervised learning algorithm applied on undirected graphs. Given a set of labeled nodes (seeds), it defines a Gibbs probability distribution over all possible spanning forests disconnecting the seeds. It calculates, for every node, the probability of sampling a forest connecting a...
[ "Directed Graph", "Matrix Tree Theorem", "Semi-Supervised Learning", "Watershed", "Random Walker" ]
Extension of the Probabilistic Watershed method to the directed setting
10,614
null
null
-1OkHh56c2m
Learning with Labeling Induced Abstentions
https://openreview.net/forum?id=-1OkHh56c2m
[ "Kareem Amin", "Giulia DeSalvo", "Afshin Rostamizadeh" ]
Poster
null
Consider a setting where we wish to automate an expensive task with a machine learning algorithm using a limited labeling resource. In such settings, examples routed for labeling are often out of scope for the machine learning algorithm. For example, in a spam detection setting, human reviewers not only provide labeled...
[ "active learning", "abstention learning", "labeling" ]
In real systems examples sent for labeling alter the distribution of examples needing evaluation by machine learning algorithms. We provide a framework for understanding these effects.
10,608
null
null
U5Af9S_RcI0
Faster Algorithms and Constant Lower Bounds for the Worst-Case Expected Error
https://openreview.net/forum?id=U5Af9S_RcI0
[ "Jonah Brown-Cohen" ]
Poster
null
The study of statistical estimation without distributional assumptions on data values, but with knowledge of data collection methods was recently introduced by Chen, Valiant and Valiant (NeurIPS 2020). In this framework, the goal is to design estimators that minimize the worst-case expected error. Here the expectation ...
[ "Learning Theory", "Convex Optimization", "Robust Estimation" ]
Faster algorithms based on online convex optimization for computing estimators minimizing worst-case expected error.
10,602
2112.13832
title_snapshot
2CQQ_C1i0b
SQALER: Scaling Question Answering by Decoupling Multi-Hop and Logical Reasoning
https://openreview.net/forum?id=2CQQ_C1i0b
[ "Mattia Atzeni", "Jasmina Bogojeska", "Andreas Loukas" ]
Poster
null
State-of-the-art approaches to reasoning and question answering over knowledge graphs (KGs) usually scale with the number of edges and can only be applied effectively on small instance-dependent subgraphs. In this paper, we address this issue by showing that multi-hop and more complex logical reasoning can be accomplis...
[ "knowledge graphs", "reasoning", "question answering" ]
null
10,596
2110.14266
title_snapshot
KpKWDyXq17d
SLOE: A Faster Method for Statistical Inference in High-Dimensional Logistic Regression
https://openreview.net/forum?id=KpKWDyXq17d
[ "Steve Yadlowsky", "Taedong Yun", "Cory Y McLean", "Alexander D'Amour" ]
Spotlight
null
Logistic regression remains one of the most widely used tools in applied statistics, machine learning and data science. However, in moderately high-dimensional problems, where the number of features $d$ is a non-negligible fraction of the sample size $n$, the logistic regression maximum likelihood estimator (MLE), and ...
[ "high dimensional", "logistic regression", "statistical inference" ]
Simpler corrections for logistic regression in high dimensions
10,592
2103.12725
title_snapshot
NhbFhfM960
Subquadratic Overparameterization for Shallow Neural Networks
https://openreview.net/forum?id=NhbFhfM960
[ "Chaehwan Song", "Ali Ramezani-Kebrya", "Thomas Pethick", "Armin Eftekhari", "Volkan Cevher" ]
Poster
null
Overparameterization refers to the important phenomenon where the width of a neural network is chosen such that learning algorithms can provably attain zero loss in nonconvex training. The existing theory establishes such global convergence using various initialization strategies, training modifications, and width scal...
[ "Overparameterization", "Gradient Descent", "Neural Networks", "Global Convergence", "Initialization" ]
This paper achieves the current best subquadratic scaling on the number of parameters for fully-trained shallow neural networks under standard initialization schemes.
10,588
2111.01875
title_snapshot
od-00q5T2vB
Reverse engineering recurrent neural networks with Jacobian switching linear dynamical systems
https://openreview.net/forum?id=od-00q5T2vB
[ "Jimmy T.H. Smith", "Scott Linderman", "David Sussillo" ]
Poster
null
Recurrent neural networks (RNNs) are powerful models for processing time-series data, but it remains challenging to understand how they function. Improving this understanding is of substantial interest to both the machine learning and neuroscience communities. The framework of reverse engineering a trained RNN by linea...
[ "recurrent neural networks", "RNNs", "switching linear dynamical systems", "SLDS", "interpretability", "dynamical systems", "reverse engineering", "fixed points" ]
We introduce a novel switching linear dynamical system that allows for the unambiguous study of the learned dynamics of recurrent neural networks.
10,586
2111.01256
title_snapshot
1QhRTsqYPB
Convergence and Alignment of Gradient Descent with Random Backpropagation Weights
https://openreview.net/forum?id=1QhRTsqYPB
[ "Ganlin Song", "Ruitu Xu", "John Lafferty" ]
Poster
null
Stochastic gradient descent with backpropagation is the workhorse of artificial neural networks. It has long been recognized that backpropagation fails to be a biologically plausible algorithm. Fundamentally, it is a non-local procedure---updating one neuron's synaptic weights requires knowledge of synaptic weights or ...
[ "feedback alignment", "deep learning" ]
null
10,585
2106.06044
title_snapshot
LEqVjnffcWo
Fair Classification with Adversarial Perturbations
https://openreview.net/forum?id=LEqVjnffcWo
[ "L. Elisa Celis", "Anay Mehrotra", "Nisheeth K Vishnoi" ]
Poster
null
We study fair classification in the presence of an omniscient adversary that, given an $\eta$, is allowed to choose an arbitrary $\eta$-fraction of the training samples and arbitrarily perturb their protected attributes. The motivation comes from settings in which protected attributes can be incorrect due to strategic ...
[ "Fair classification", "Algorithmic fairness", "Adversarial data" ]
Ensuring fairness in classification when a fraction of the protected attributes are adversarially perturbed
10,542
2106.05964
title_snapshot
Mfi0LZmFB5a
SOPE: Spectrum of Off-Policy Estimators
https://openreview.net/forum?id=Mfi0LZmFB5a
[ "Christina Yuan", "Yash Chandak", "Stephen Giguere", "Philip S. Thomas", "Scott Niekum" ]
Poster
null
Many sequential decision making problems are high-stakes and require off-policy evaluation (OPE) of a new policy using historical data collected using some other policy. One of the most common OPE techniques that provides unbiased estimates is trajectory based importance sampling (IS). However, due to the high variance...
[ "reinforcement learning", "off-policy policy evaluation" ]
null
10,540
2111.03936
title_snapshot
zMZPDwm3H3
Learning the optimal Tikhonov regularizer for inverse problems
https://openreview.net/forum?id=zMZPDwm3H3
[ "Giovanni S Alberti", "Ernesto De Vito", "Matti Lassas", "Luca Ratti", "Matteo Santacesaria" ]
Poster
null
In this work, we consider the linear inverse problem $y=Ax+\varepsilon$, where $A\colon X\to Y$ is a known linear operator between the separable Hilbert spaces $X$ and $Y$, $x$ is a random variable in $X$ and $\epsilon$ is a zero-mean random process in $Y$. This setting covers several inverse problems in imaging includ...
[ "inverse problems", "regularization", "MMSE", "optimal estimation", "supervised and unsupervised learning", "generalization estimates", "Hilbert spaces" ]
null
10,535
2106.06513
title_snapshot
LNXTIrMqyGz
Implicit Semantic Response Alignment for Partial Domain Adaptation
https://openreview.net/forum?id=LNXTIrMqyGz
[ "Wenxiao Xiao", "Zhengming Ding", "Hongfu Liu" ]
Poster
null
Partial Domain Adaptation (PDA) addresses the unsupervised domain adaptation problem where the target label space is a subset of the source label space. Most state-of-art PDA methods tackle the inconsistent label space by assigning weights to classes or individual samples, in an attempt to discard the source data that ...
[ "Partial Domain Adaptation", "Transfer Learning", "Implicit Semantic Discovery", "Feature-Level Weighting Schema", "Semantic Alignment" ]
null
10,534
null
null
x5hh6N9bUUb
Stochastic Solutions for Linear Inverse Problems using the Prior Implicit in a Denoiser
https://openreview.net/forum?id=x5hh6N9bUUb
[ "Zahra Kadkhodaie", "Eero P Simoncelli" ]
Poster
null
Deep neural networks have provided state-of-the-art solutions for problems such as image denoising, which implicitly rely on a prior probability model of natural images. Two recent lines of work – Denoising Score Matching and Plug-and-Play – propose methodologies for drawing samples from this implicit prior and using i...
[ "inverse problems", "image priors", "unsupervised learning", "semi-supervised learning", "deblurring", "superresolution", "compressive sensing" ]
We've described a framework for using the prior implicitly embedded in a denoiser to solve other linear inverse problems by sampling.
10,532
2007.13640
title_judge
l7Yjt_8WvJ
On Optimal Interpolation in Linear Regression
https://openreview.net/forum?id=l7Yjt_8WvJ
[ "Eduard Oravkin", "Patrick Rebeschini" ]
Poster
null
Understanding when and why interpolating methods generalize well has recently been a topic of interest in statistical learning theory. However, systematically connecting interpolating methods to achievable notions of optimality has only received partial attention. In this paper, we ask the question of what is the optim...
[ "interpolation", "optimality", "generalization", "linear regression", "statistical learning theory" ]
studying and designing optimal interpolating methods in linear regression
10,530
2110.11258
title_snapshot
YP1ham75vml
TriBERT: Human-centric Audio-visual Representation Learning
https://openreview.net/forum?id=YP1ham75vml
[ "Tanzila Rahman", "Mengyu Yang", "Leonid Sigal" ]
Poster
null
The recent success of transformer models in language, such as BERT, has motivated the use of such architectures for multi-modal feature learning and tasks. However, most multi-modal variants (e.g., ViLBERT) have limited themselves to visual-linguistic data. Relatively few have explored its use in audio-visual modalitie...
[ "BERT", "Transformer", "multi-modal", "joint representation" ]
A tri-modal VilBERT-inspired model, which we call TriBERT, that co-attends among visual, pose keypoint, and audio modalities to produce highly contextualized representations.
10,528
2110.13412
title_judge
Nl7VO_Y7K4Q
Learning Policies with Zero or Bounded Constraint Violation for Constrained MDPs
https://openreview.net/forum?id=Nl7VO_Y7K4Q
[ "Tao Liu", "Ruida Zhou", "Dileep Kalathil", "Panganamala Kumar", "Chao Tian" ]
Poster
null
We address the issue of safety in reinforcement learning. We pose the problem in an episodic framework of a constrained Markov decision process. Existing results have shown that it is possible to achieve a reward regret of $\tilde{\mathcal{O}}(\sqrt{K})$ while allowing an $\tilde{\mathcal{O}}(\sqrt{K})$ constraint viol...
[ "Safe Reinforcement Learning", "Constrained Markov Decision Process" ]
We design algorithms that guarantee zero or bounded constraint violation for episodic constrained MDPs, while attaining an \tilde{\mathcal{O}}(\sqrt{Number of episodes}) reward regret.
10,513
2106.02684
title_snapshot
Nfbe1usrgx4
Differential Privacy Dynamics of Langevin Diffusion and Noisy Gradient Descent
https://openreview.net/forum?id=Nfbe1usrgx4
[ "Rishav Chourasia", "Jiayuan Ye", "Reza Shokri" ]
Spotlight
null
What is the information leakage of an iterative randomized learning algorithm about its training data, when the internal state of the algorithm is \emph{private}? How much is the contribution of each specific training epoch to the information leakage through the released model? We study this problem for noisy gradient ...
[ "Differential Privacy", "Noisy Gradient Descent" ]
Tight converging differential privacy analysis for noisy gradient descent when only last iterate model is released.
10,507
2102.05855
title_snapshot
zjJyjQj1W7U
Hierarchical Clustering: $O(1)$-Approximation for Well-Clustered Graphs
https://openreview.net/forum?id=zjJyjQj1W7U
[ "Bogdan Adrian Manghiuc", "He Sun" ]
Poster
null
Hierarchical clustering studies a recursive partition of a data set into clusters of successively smaller size, and is a fundamental problem in data analysis. In this work we study the cost function for hierarchical clustering introduced by Dasgupta, and present two polynomial-time approximation algorithms: Our first ...
[ "Hierarchical clustering", "graph algorithms", "spectral methods" ]
We present $O(1)$-approximation algorithms for hierarchical clustering when the input graph exhibits a clears structure of clusters.
10,489
2112.09055
title_snapshot
kTy7bbm-4I4
Controlled Text Generation as Continuous Optimization with Multiple Constraints
https://openreview.net/forum?id=kTy7bbm-4I4
[ "Sachin Kumar", "Eric Malmi", "Aliaksei Severyn", "Yulia Tsvetkov" ]
Poster
null
As large-scale language model pretraining pushes the state-of-the-art in text generation, recent work has turned to controlling attributes of the text such models generate. While modifying the pretrained models via fine-tuning remains the popular approach, it incurs a significant computational cost and can be infeasibl...
[ "controllable text generation", "constrained optimization", "style transfer" ]
We present a method of controllable inference from pretrained language models via continuous optimization through gradient descent
10,488
2108.01850
title_snapshot
iBHiqlbFvLb
An Even More Optimal Stochastic Optimization Algorithm: Minibatching and Interpolation Learning
https://openreview.net/forum?id=iBHiqlbFvLb
[ "Blake Woodworth", "Nathan Srebro" ]
Poster
null
We present and analyze an algorithm for optimizing smooth and convex or strongly convex objectives using minibatch stochastic gradient estimates. The algorithm is optimal with respect to its dependence on both the minibatch size and minimum expected loss simultaneously. This improves over the optimal method of Lan, whi...
[ "Stochastic", "Optimization", "Learning", "Minibatch", "Optimistic", "Interpolation" ]
We present optimal minibatch stochastic optimization algorithm for interpolation learning.
10,484
2106.02720
title_snapshot
fIn4wLS2XzU
Behavior From the Void: Unsupervised Active Pre-Training
https://openreview.net/forum?id=fIn4wLS2XzU
[ "Hao Liu", "Pieter Abbeel" ]
Spotlight
null
We introduce a new unsupervised pre-training method for reinforcement learning called APT, which stands for Active Pre-Training. APT learns behaviors and representations by actively searching for novel states in reward-free environments. The key novel idea is to explore the environment by maximizing a non-parametric en...
[ "unsupervised exploration", "unsupervised pretraining", "reinforcement learning" ]
We present an unsupervised pre-training RL method that is based on particle-based state entropy maximization and it outperforms canonical RL methods.
10,477
2103.04551
title_snapshot
4c1EiEvivpx
Neural Scene Flow Prior
https://openreview.net/forum?id=4c1EiEvivpx
[ "Xueqian Li", "Jhony Kaesemodel Pontes", "Simon Lucey" ]
Spotlight
null
Before the deep learning revolution, many perception algorithms were based on runtime optimization in conjunction with a strong prior/regularization penalty. A prime example of this in computer vision is optical and scene flow. Supervised learning has largely displaced the need for explicit regularization. Instead, the...
[ "neural prior", "scene flow", "implicit regularizer" ]
We estimate scene flow through runtime optimization with a neural prior, which utilizes the architecture of neural networks as an implicit regularizer.
10,465
2111.01253
title_snapshot
dwY40cSK-dt
How Modular should Neural Module Networks Be for Systematic Generalization?
https://openreview.net/forum?id=dwY40cSK-dt
[ "Vanessa D'Amario", "Tomotake Sasaki", "Xavier Boix" ]
Poster
null
Neural Module Networks (NMNs) aim at Visual Question Answering (VQA) via composition of modules that tackle a sub-task. NMNs are a promising strategy to achieve systematic generalization, i.e., overcoming biasing factors in the training distribution. However, the aspects of NMNs that facilitate systematic generalizatio...
[ "Neural Module Networks", "Systematic Generalization", "Out of Distribution", "VQA" ]
We show that the degree of modularity of Neural Module Networks for VQA has a large impact in systematic generalization, especially in the image encoder.
10,464
2106.08170
title_snapshot
1ANcwXQuijU
Set Prediction in the Latent Space
https://openreview.net/forum?id=1ANcwXQuijU
[ "Konpat Preechakul", "Chawan Piansaddhayanon", "Burin Naowarat", "Tirasan Khandhawit", "Sira Sriswasdi", "Ekapol Chuangsuwanich" ]
Poster
null
Set prediction tasks require the matching between predicted set and ground truth set in order to propagate the gradient signal. Recent works have performed this matching in the original feature space thus requiring predefined distance functions. We propose a method for learning the distance function by performing the m...
[ "set prediction", "object detection", "image captioning", "medical report generation" ]
null
10,460
null
null
VA18aFPYfkd
Truncated Marginal Neural Ratio Estimation
https://openreview.net/forum?id=VA18aFPYfkd
[ "Benjamin Kurt Miller", "Alex Cole", "Patrick Forré", "Gilles Louppe", "Christoph Weniger" ]
Poster
null
Parametric stochastic simulators are ubiquitous in science, often featuring high-dimensional input parameters and/or an intractable likelihood. Performing Bayesian parameter inference in this context can be challenging. We present a neural simulation-based inference algorithm which simultaneously offers simulation effi...
[ "simulation-based inference", "likelihood-free inference", "implicit likelihood", "posterior", "marginal posterior", "likelihood-to-evidence ratio estimation", "neural ratio estimation", "empirically testable inference", "Bayesian inference", "inverse problem" ]
We present an empirically testable simulation-based inference algorithm which efficiently estimates any marginal posterior by approximating the likelihood-to-evidence ratio on the region of highest posterior density.
10,457
2107.01214
title_snapshot
Zsrn9wXWN0
Active Offline Policy Selection
https://openreview.net/forum?id=Zsrn9wXWN0
[ "Ksenia Konyushkova", "Yutian Chen", "Thomas Paine", "Caglar Gulcehre", "Cosmin Paduraru", "Daniel J Mankowitz", "Misha Denil", "Nando de Freitas" ]
Poster
null
This paper addresses the problem of policy selection in domains with abundant logged data, but with a restricted interaction budget. Solving this problem would enable safe evaluation and deployment of offline reinforcement learning policies in industry, robotics, and recommendation domains among others. Several off-pol...
[ "policy selection", "offline reinforcement learning", "off-policy policy evaluation", "active learning", "Bayesian optimisation" ]
Active offline policy selection helps to identify the best offline RL policy for deployment using logged data and limited interaction budget.
10,455
2106.10251
title_snapshot
WybjTtCKfGi
Shape your Space: A Gaussian Mixture Regularization Approach to Deterministic Autoencoders
https://openreview.net/forum?id=WybjTtCKfGi
[ "Amrutha Saseendran", "Kathrin Skubch", "Stefan Falkner", "Margret Keuper" ]
Poster
null
Variational Autoencoders (VAEs) are powerful probabilistic models to learn representations of complex data distributions. One important limitation of VAEs is the strong prior assumption that latent representations learned by the model follow a simple uni-modal Gaussian distribution. Further, the variational training pr...
[ "Generative Model", "Variational Autoencoder", "Gaussian mixture model", "Regularization" ]
A simple and end-to-end trainable Deterministic Gaussian Mixture based Autoencoder approach for efficient generative modelling
10,452
null
null
98zhe-xzviq
Drawing Robust Scratch Tickets: Subnetworks with Inborn Robustness Are Found within Randomly Initialized Networks
https://openreview.net/forum?id=98zhe-xzviq
[ "Yonggan Fu", "Qixuan Yu", "Yang Zhang", "Shang Wu", "Xu Ouyang", "David Daniel Cox", "Yingyan Lin" ]
Poster
null
Deep Neural Networks (DNNs) are known to be vulnerable to adversarial attacks, i.e., an imperceptible perturbation to the input can mislead DNNs trained on clean images into making erroneous predictions. To tackle this, adversarial training is currently the most effective defense method, by augmenting the training set ...
[ "model robustness", "adversarial training", "lottery ticket hypothesis" ]
We find subnetworks with inborn robustness hidden within randomly initialized networks and then extensively study their properties and applications.
10,440
2110.14068
title_snapshot
_WnGcwXLYOE
FLEX: Unifying Evaluation for Few-Shot NLP
https://openreview.net/forum?id=_WnGcwXLYOE
[ "Jonathan Bragg", "Arman Cohan", "Kyle Lo", "Iz Beltagy" ]
Poster
null
Few-shot NLP research is highly active, yet conducted in disjoint research threads with evaluation suites that lack challenging-yet-realistic testing setups and fail to employ careful experimental design. Consequently, the community does not know which techniques perform best or even if they outperform simple baselines...
[ "benchmarks", "evaluation", "few-shot", "zero-shot", "nlp", "prompt-based models", "pretrained language models", "statistical analysis" ]
FLEX Principles, benchmark, and leaderboard unifying best practices for evaluating few-shot NLP; and UniFew, a simple and strong prompt-based model by unifying pre-training and downstream task formats.
10,421
2107.07170
title_snapshot
GNFcszMtYvV
Convergence of adaptive algorithms for constrained weakly convex optimization
https://openreview.net/forum?id=GNFcszMtYvV
[ "Ahmet Alacaoglu", "Yura Malitsky", "Volkan Cevher" ]
Poster
null
We analyze the adaptive first order algorithm AMSGrad, for solving a constrained stochastic optimization problem with a weakly convex objective. We prove the $\mathcal{\tilde O}(t^{-1/2})$ rate of convergence for the squared norm of the gradient of Moreau envelope, which is the standard stationarity measure for this cl...
[ "adaptive gradient algorithms", "weakly convex optimization", "AMSGrad", "Adam" ]
We establish convergence of adaptive algorithms for a class of nonsmooth nonconvex problems, for the first time.
10,405
2006.06650
title_judge
lVmIjQiJJSr
Removing Inter-Experimental Variability from Functional Data in Systems Neuroscience
https://openreview.net/forum?id=lVmIjQiJJSr
[ "Dominic Gonschorek", "Larissa Höfling", "Klaudia P. Szatko", "Katrin Franke", "Timm Schubert", "Benjamin Adric Dunn", "Philipp Berens", "David A. Klindt", "Thomas Euler" ]
Spotlight
null
Integrating data from multiple experiments is common practice in systems neuroscience but it requires inter-experimental variability to be negligible compared to the biological signal of interest. This requirement is rarely fulfilled; systematic changes between experiments can drastically affect the outcome of complex ...
[ "systems neuroscience", "inter-experimental variability", "cell type classification", "retina", "domain adaptation", "autoencoder", "adversarial optimization" ]
We offer a flexible approach to remove inter-experimental variability and integrate datasets across experiments in systems neuroscience.
10,400
null
null
du_Rss0tW8
Certifying Robustness to Programmable Data Bias in Decision Trees
https://openreview.net/forum?id=du_Rss0tW8
[ "Anna P. Meyer", "Aws Albarghouthi", "Loris D'Antoni" ]
Poster
null
Datasets can be biased due to societal inequities, human biases, under-representation of minorities, etc. Our goal is to certify that models produced by a learning algorithm are pointwise-robust to dataset biases. This is a challenging problem: it entails learning models for a large, or even infinite, number of dataset...
[ "data bias", "data poisoning", "robustness", "decision trees", "fairness" ]
null
10,394
2110.04363
title_snapshot
8pOPKfibVN
Tailoring: encoding inductive biases by optimizing unsupervised objectives at prediction time
https://openreview.net/forum?id=8pOPKfibVN
[ "Ferran Alet", "Maria Bauza Villalonga", "Kenji Kawaguchi", "Nurullah Giray Kuru", "Tomas Perez", "Leslie Pack Kaelbling" ]
Poster
null
From CNNs to attention mechanisms, encoding inductive biases into neural networks has been a fruitful source of improvement in machine learning. Adding auxiliary losses to the main objective function is a general way of encoding biases that can help networks learn better representations. However, since auxiliary losses...
[ "meta-learning", "inductive biases", "self-supervised learning" ]
We optimize unsupervised losses for the current input. By optimizing where we act, we bypass generalization gaps and can impose a wide variety of inductive biases.
10,386
2009.10623
title_snapshot
nJqCQUzpvS
Learning State Representations from Random Deep Action-conditional Predictions
https://openreview.net/forum?id=nJqCQUzpvS
[ "Zeyu Zheng", "Vivek Veeriah", "Risto Vuorio", "Richard Lewis", "Satinder Singh" ]
Poster
null
Our main contribution in this work is an empirical finding that random General Value Functions (GVFs), i.e., deep action-conditional predictions---random both in what feature of observations they predict as well as in the sequence of actions the predictions are conditioned upon---form good auxiliary tasks for reinforce...
[ "reinforcement learning", "state representation learning", "auxiliary task" ]
null
10,375
2102.04897
title_snapshot
P9_gOq5w7Eb
Towards optimally abstaining from prediction with OOD test examples
https://openreview.net/forum?id=P9_gOq5w7Eb
[ "Adam Tauman Kalai", "Varun Kanade" ]
Spotlight
null
A common challenge across all areas of machine learning is that training data is not distributed like test data, due to natural shifts or adversarial examples; such examples are referred to as out-of-distribution (OOD) test examples. We consider a model where one may abstain from predicting, at a fixed cost. In particu...
[ "Selective classification", "covariate shift", "classification with a reject option" ]
Near-optimal guarantees for selective classification (with the option to abstain at a cost), with extreme covariate shift where train and test distributions do not even fully overlap
10,360
2105.14119
title_snapshot
7rYDxRb1eSa
Bias and variance of the Bayesian-mean decoder
https://openreview.net/forum?id=7rYDxRb1eSa
[ "Arthur Prat-Carrabin", "Michael Woodford" ]
Spotlight
null
Perception, in theoretical neuroscience, has been modeled as the encoding of external stimuli into internal signals, which are then decoded. The Bayesian mean is an important decoder, as it is optimal for purposes of both estimation and discrimination. We present widely-applicable approximations to the bias and to the ...
[ "neuroscience", "perception", "encoding-decoding models", "efficient coding", "Bayesian inference", "Bayesian mean", "perceptual biases" ]
We present an analytical approximation to the mean of the Bayesian posterior, as a function of the prior and of the encoding Fisher information, and study the resulting perceptual biases under different encoding strategies.
10,356
2105.13705
title_snapshot
_KhlwS9oFBp
Rate-Optimal Subspace Estimation on Random Graphs
https://openreview.net/forum?id=_KhlwS9oFBp
[ "Zhixin Zhou", "Fan Zhou", "Ping Li", "Cun-Hui Zhang" ]
Poster
null
We study the theory of random bipartite graph whose adjacency matrix is generated according to a connectivity matrix $M$. We consider the bipartite graph to be sparse, i.e., the entries of $M$ are upper bounded by certain sparsity parameter. We show that the performance of estimating the connectivity matrix $M$ depends...
[ "bipartite graph", "network analysis", "singular value thresholding", "minimax theory" ]
null
10,355
null
null
e2gqGkFjDHg
Redesigning the Transformer Architecture with Insights from Multi-particle Dynamical Systems
https://openreview.net/forum?id=e2gqGkFjDHg
[ "Subhabrata Dutta", "Tanya Gautam", "Soumen Chakrabarti", "Tanmoy Chakraborty" ]
Spotlight
null
The Transformer and its variants have been proven to be efficient sequence learners in many different domains. Despite their staggering success, a critical issue has been the enormous number of parameters that must be trained (ranging from $10^7$ to $10^{11}$) along with the quadratic complexity of dot-product attentio...
[ "Transformers", "Dynamical systems", "Time-evolution", "Self-attention" ]
A dynamical system motivated Transformer-variant that can achieves state-of-the-art performance with astoundingly fewer parameters.
10,344
2109.15142
title_snapshot
1bBF5Zq1YHz
Probabilistic Tensor Decomposition of Neural Population Spiking Activity
https://openreview.net/forum?id=1bBF5Zq1YHz
[ "Hugo Soulat", "Sepiedeh Keshavarzi", "Troy William Margrie", "Maneesh Sahani" ]
Spotlight
null
The firing of neural populations is coordinated across cells, in time, and across experimental conditions or repeated experimental trials; and so a full understanding of the computational significance of neural responses must be based on a separation of these different contributions to structured activity. Tensor deco...
[ "Probabilistic", "Tensor Decomposition", "Neuroscience", "Spike", "Population Activity", "Count" ]
null
10,341
null
null
ZdyLIxqgz29
An Axiomatic Theory of Provably-Fair Welfare-Centric Machine Learning
https://openreview.net/forum?id=ZdyLIxqgz29
[ "Cyrus Cousins" ]
Poster
null
We address an inherent difficulty in welfare-theoretic fair machine learning (ML), by proposing an equivalently-axiomatically justified alternative setting, and studying the resulting computational and statistical learning questions. Welfare metrics quantify overall wellbeing across a population of groups, and welfare-...
[ "PAC-Learning", "Fair Learning", "Welfare", "Computational Learning Theory", "Statistical Learning Theory" ]
We propose and develop a theory of fair machine learning, with roots in PAC learnability and welfare economics.
10,326
2104.14504
title_snapshot
ytke6qKpxtr
STORM+: Fully Adaptive SGD with Recursive Momentum for Nonconvex Optimization
https://openreview.net/forum?id=ytke6qKpxtr
[ "Kfir Yehuda Levy", "Ali Kavis", "Volkan Cevher" ]
Poster
null
In this work we investigate stochastic non-convex optimization problems where the objective is an expectation over smooth loss functions, and the goal is to find an approximate stationary point. The most popular approach to handling such problems is variance reduction techniques, which are also known to obtain tight co...
[ "adaptive methods", "recursive momentum", "nonconvex optimization", "stochastic optimization" ]
null
10,323
2111.01040
title_judge
yJqcM36Qvnu
Federated Graph Classification over Non-IID Graphs
https://openreview.net/forum?id=yJqcM36Qvnu
[ "Han Xie", "Jing Ma", "Li Xiong", "Carl Yang" ]
Poster
null
Federated learning has emerged as an important paradigm for training machine learning models in different domains. For graph-level tasks such as graph classification, graphs can also be regarded as a special type of data samples, which can be collected and stored in separate local systems. Similar to other domains, mul...
[ "federated learning", "graph classification", "non-iid graphs", "structure heterogeneity", "feature heterogeneity" ]
We propose new methods for cross-dataset/cross-domain graph classification with federated learning.
10,317
2106.13423
title_snapshot
XK4eVsG2LKw
Margin-Independent Online Multiclass Learning via Convex Geometry
https://openreview.net/forum?id=XK4eVsG2LKw
[ "Guru Guruganesh", "Allen Liu", "Jon Schneider", "Joshua Ruizhi Wang" ]
Poster
null
We consider the problem of multi-class classification, where a stream of adversarially chosen queries arrive and must be assigned a label online. Unlike traditional bounds which seek to minimize the misclassification rate, we minimize the total distance from each query to the region corresponding to its assigned label....
[ "Multiclass Classification", "Online Learning" ]
We design algorithms for multi-class classification that minimize the average size of mistakes instead of simply the number of mistakes.
10,316
2111.08057
title_snapshot
2LdBqxc1Yv
Variational Diffusion Models
https://openreview.net/forum?id=2LdBqxc1Yv
[ "Diederik P Kingma", "Tim Salimans", "Ben Poole", "Jonathan Ho" ]
Poster
null
Diffusion-based generative models have demonstrated a capacity for perceptually impressive synthesis, but can they also be great likelihood-based models? We answer this in the affirmative, and introduce a family of diffusion-based generative models that obtain state-of-the-art likelihoods on standard image density esti...
[ "Diffusion Models", "Generative Models", "Density Estimation" ]
SOTA likelihood using diffusion models with learnable noise schedule
10,302
2107.00630
title_snapshot
usxt30HpW66
What’s a good imputation to predict with missing values?
https://openreview.net/forum?id=usxt30HpW66
[ "Marine Le Morvan", "Julie Josse", "Erwan Scornet", "Gael Varoquaux" ]
Spotlight
null
How to learn a good predictor on data with missing values? Most efforts focus on first imputing as well as possible and second learning on the completed data to predict the outcome. Yet, this widespread practice has no theoretical grounding. Here we show that for almost all imputation functions, an impute-then-regress ...
[ "missing values", "neural networks", "Bayes optimality", "consistency" ]
null
10,299
2106.00311
title_snapshot
FyI2-YoHHd
Learning Nonparametric Volterra Kernels with Gaussian Processes
https://openreview.net/forum?id=FyI2-YoHHd
[ "Magnus Ross", "Michael Thomas Smith", "Mauricio A Álvarez" ]
Poster
null
This paper introduces a method for the nonparametric Bayesian learning of nonlinear operators, through the use of the Volterra series with kernels represented using Gaussian processes (GPs), which we term the nonparametric Volterra kernels model (NVKM). When the input function to the operator is unobserved and has a GP...
[ "Gaussian Processes" ]
null
10,297
2106.05582
title_snapshot
myJO35O7Gg
Assessing Fairness in the Presence of Missing Data
https://openreview.net/forum?id=myJO35O7Gg
[ "Yiliang Zhang", "Qi Long" ]
Poster
null
Missing data are prevalent and present daunting challenges in real data analysis. While there is a growing body of literature on fairness in analysis of fully observed data, there has been little theoretical work on investigating fairness in analysis of incomplete data. In practice, a popular analytical approach for de...
[ "Algorithmic fairness", "Missing data", "Domain adaptation" ]
null
10,289
2112.04899
title_snapshot
sNw3VBPL7rg
Calibration and Consistency of Adversarial Surrogate Losses
https://openreview.net/forum?id=sNw3VBPL7rg
[ "Pranjal Awasthi", "Natalie Frank", "Anqi Mao", "Mehryar Mohri", "Yutao Zhong" ]
Spotlight
null
Adversarial robustness is an increasingly critical property of classifiers in applications. The design of robust algorithms relies on surrogate losses since the optimization of the adversarial loss with most hypothesis sets is NP-hard. But, which surrogate losses should be used and when do they benefit from theoretical...
[ "Adversarial Robustness", "Learning Theory", "Consistency", "Calibration", "Statistical learning", "Classification." ]
null
10,283
2104.09658
title_snapshot
B6uDDaDoW4a
On the Periodic Behavior of Neural Network Training with Batch Normalization and Weight Decay
https://openreview.net/forum?id=B6uDDaDoW4a
[ "Ekaterina Lobacheva", "Maxim Kodryan", "Nadezhda Chirkova", "Andrey Malinin", "Dmitry P. Vetrov" ]
Poster
null
Training neural networks with batch normalization and weight decay has become a common practice in recent years. In this work, we show that their combined use may result in a surprising periodic behavior of optimization dynamics: the training process regularly exhibits destabilizations that, however, do not lead to com...
[ "neural network training", "batch normalization", "weight decay" ]
We study the periodic behavior of training dynamics caused by the interaction of batch normalization and weight decay.
10,268
2106.15739
title_snapshot
hzioAx8g9x
Neural optimal feedback control with local learning rules
https://openreview.net/forum?id=hzioAx8g9x
[ "Johannes Friedrich", "Siavash Golkar", "Shiva Farashahi", "Alexander Genkin", "Anirvan M. Sengupta", "Dmitri Chklovskii" ]
Spotlight
null
A major problem in motor control is understanding how the brain plans and executes proper movements in the face of delayed and noisy stimuli. A prominent framework for addressing such control problems is Optimal Feedback Control (OFC). OFC generates control actions that optimize behaviorally relevant criteria by integr...
[ "optimal feedback control", "Kalman filter", "motor control", "neural circuit" ]
The paper introduces a biologically plausible neural implementation of optimal feedback control, which considers that sensory feedback is delayed, and combines adaptive Kalman filtering and model free control.
10,265
2111.06920
title_snapshot
RJ7XFI15Q8f
Achieving Forgetting Prevention and Knowledge Transfer in Continual Learning
https://openreview.net/forum?id=RJ7XFI15Q8f
[ "Zixuan Ke", "Bing Liu", "Nianzu Ma", "Hu Xu", "Lei Shu" ]
Poster
null
Continual learning (CL) learns a sequence of tasks incrementally with the goal of achieving two main objectives: overcoming catastrophic forgetting (CF) and encouraging knowledge transfer (KT) across tasks. However, most existing techniques focus only on overcoming CF and have no mechanism to encourage KT, and thus do ...
[ "continual learning", "natural language processing applications of continual learning" ]
This paper presents an algorithm that deals with both catastrophic forgetting and knowledge transfer for learning a sequence of natural language tasks.
10,247
2112.02706
title_snapshot
um7zVEeyVH1
Terra: Imperative-Symbolic Co-Execution of Imperative Deep Learning Programs
https://openreview.net/forum?id=um7zVEeyVH1
[ "Taebum Kim", "Eunji Jeong", "Geon-Woo Kim", "Yunmo Koo", "Sehoon Kim", "Gyeong-In Yu", "Byung-Gon Chun" ]
Poster
null
Imperative programming allows users to implement their deep neural networks (DNNs) easily and has become an essential part of recent deep learning (DL) frameworks. Recently, several systems have been proposed to combine the usability of imperative programming with the optimized performance of symbolic graph execution. ...
[ "machine learning system" ]
null
10,236
2201.09210
title_snapshot
jTEGbvLjgp
Differentiable Spline Approximations
https://openreview.net/forum?id=jTEGbvLjgp
[ "Minsu Cho", "Aditya Balu", "Ameya Joshi", "Anjana Deva Prasad", "Biswajit Khara", "Soumik Sarkar", "Baskar Ganapathysubramanian", "Adarsh Krishnamurthy", "Chinmay Hegde" ]
Poster
null
The paradigm of differentiable programming has significantly enhanced the scope of machine learning via the judicious use of gradient-based optimization. However, standard differentiable programming methods (such as autodiff) typically require that the machine learning models be differentiable, limiting their applicabi...
[ "Differentiable programming", "Spline approximation", "NURBS", "k-histogram" ]
null
10,230
2110.01532
title_snapshot
DsWYWm6ozxx
Bias Out-of-the-Box: An Empirical Analysis of Intersectional Occupational Biases in Popular Generative Language Models
https://openreview.net/forum?id=DsWYWm6ozxx
[ "Hannah Rose Kirk", "Yennie Jun", "Filippo Volpin", "Haider Iqbal", "Elias Benussi", "Frederic A Dreyer", "Aleksandar Shtedritski", "Yuki Asano" ]
Poster
null
The capabilities of natural language models trained on large-scale data have increased immensely over the past few years. Open source libraries such as HuggingFace have made these models easily available and accessible. While prior research has identified biases in large language models, this paper considers biases con...
[ "Natural language processing (NLP)", "generative language models", "bias", "GPT", "occupations", "intersections" ]
We ask what biases exist in the most downloaded text generation model (GPT-2) by extracting occupational associations for gender intersected with five protected categories and comparing these predictions to US Labor Bureau data.
10,219
2102.04130
title_snapshot
_NOwVKCmSo
Noether Networks: meta-learning useful conserved quantities
https://openreview.net/forum?id=_NOwVKCmSo
[ "Ferran Alet", "Dylan Doblar", "Allan Zhou", "Joshua B. Tenenbaum", "Kenji Kawaguchi", "Chelsea Finn" ]
Poster
null
Progress in machine learning (ML) stems from a combination of data availability, computational resources, and an appropriate encoding of inductive biases. Useful biases often exploit symmetries in the prediction problem, such as convolutional networks relying on translation equivariance. Automatically discovering these...
[ "meta-learning", "sequential prediction", "physics", "video prediction" ]
We meta-learn inductive biases in the form of useful conserved quantities that improve predictions when optimized inside the prediction function.
10,214
2112.03321
title_snapshot
tjdHCnPqoo
Is Automated Topic Model Evaluation Broken? The Incoherence of Coherence
https://openreview.net/forum?id=tjdHCnPqoo
[ "Alexander Hoyle", "Pranav Goel", "Andrew Hian-Cheong", "Denis Peskoff", "Jordan Lee Boyd-Graber", "Philip Resnik" ]
Spotlight
null
Topic model evaluation, like evaluation of other unsupervised methods, can be contentious. However, the field has coalesced around automated estimates of topic coherence, which rely on the frequency of word co-occurrences in a reference corpus. Contemporary neural topic models surpass classical ones according to these ...
[ "topic model", "topic model evaluation", "npmi", "topic coherence", "human evaluation", "metric", "evaluation", "validation", "crowdsourcing", "model comparison", "automatic metric", "automated metric" ]
Yes, automated topic evaluation is broken.
10,209
2107.02173
title_snapshot
comGUyv5sac
Relaxed Marginal Consistency for Differentially Private Query Answering
https://openreview.net/forum?id=comGUyv5sac
[ "Ryan McKenna", "Siddhant Pradhan", "Daniel Sheldon", "Gerome Miklau" ]
Poster
null
Many differentially private algorithms for answering database queries involve a step that reconstructs a discrete data distribution from noisy measurements. This provides consistent query answers and reduces error, but often requires space that grows exponentially with dimension. PRIVATE-PGM is a recent approach that u...
[ "differential privacy", "convex optimization", "graphical models", "approximate inference", "local polytope" ]
We propose a post-processing technique that boosts utility by enforcing (local) consistency constraints. Our method is scalable to far more general settings than prior work.
10,208
2109.06153
title_snapshot
WVYzd7GvaOM
Attention Approximates Sparse Distributed Memory
https://openreview.net/forum?id=WVYzd7GvaOM
[ "Trenton Bricken", "Cengiz Pehlevan" ]
Poster
null
While Attention has come to be an important mechanism in deep learning, there remains limited intuition for why it works so well. Here, we show that Transformer Attention can be closely related under certain data conditions to Kanerva's Sparse Distributed Memory (SDM), a biologically plausible associative memory model....
[ "Transformer", "Attention", "Sparse Distributed Memory", "Associative Memory", "Cerebellum" ]
We show the heuristic Transformer Attention operation can be implemented with simple properties of high dimensional vectors, in a biologically plausible fashion using SDM.
10,207
2111.05498
title_snapshot
2vyiCxfb6el
From global to local MDI variable importances for random forests and when they are Shapley values
https://openreview.net/forum?id=2vyiCxfb6el
[ "Antonio Sutera", "Gilles Louppe", "Van Anh Huynh-Thu", "Louis Wehenkel", "Pierre Geurts" ]
Poster
null
Random forests have been widely used for their ability to provide so-called importance measures, which give insight at a global (per dataset) level on the relevance of input variables to predict a certain output. On the other hand, methods based on Shapley values have been introduced to refine the analysis of feature r...
[ "random forests", "variable importances", "shapley values", "feature ranking", "information theory", "explainability", "interpretable machine learning" ]
We propose a local MDI variable importance measure for random forests and analyse global and local MDI in the light of Shapley values.
10,186
2111.02218
title_snapshot
9IJLHPuLpvZ
Wasserstein Flow Meets Replicator Dynamics: A Mean-Field Analysis of Representation Learning in Actor-Critic
https://openreview.net/forum?id=9IJLHPuLpvZ
[ "Yufeng Zhang", "Siyu Chen", "Zhuoran Yang", "Michael Jordan", "Zhaoran Wang" ]
Poster
null
Actor-critic (AC) algorithms, empowered by neural networks, have had significant empirical success in recent years. However, most of the existing theoretical support for AC algorithms focuses on the case of linear function approximations, or linearized neural networks, where the feature representation is fixed through...
[ "reinforcement learning", "policy optimization", "optimal transport" ]
null
10,183
2112.13530
title_snapshot
NNZ0caVe2ak
Square Root Principal Component Pursuit: Tuning-Free Noisy Robust Matrix Recovery
https://openreview.net/forum?id=NNZ0caVe2ak
[ "Junhui Zhang", "Jingkai Yan", "John Wright" ]
Poster
null
We propose a new framework -- Square Root Principal Component Pursuit -- for low-rank matrix recovery from observations corrupted with noise and outliers. Inspired by the square root Lasso, this new formulation does not require prior knowledge of the noise level. We show that a single, universal choice of the regulariz...
[ "matrix recovery", "ADMM", "convex optimization" ]
We propose Square Root Principle Component Pursuit, which is capable of achieving low-rank plus sparse matrix recovery from noisy data without needing to know the noise level in advance.
10,180
2106.09211
title_snapshot
jScy7BjbZeQ
Grad2Task: Improved Few-shot Text Classification Using Gradients for Task Representation
https://openreview.net/forum?id=jScy7BjbZeQ
[ "Jixuan Wang", "Kuan-Chieh Wang", "Frank Rudzicz", "Michael Brudno" ]
Poster
null
Large pretrained language models (LMs) like BERT have improved performance in many disparate natural language processing (NLP) tasks. However, fine tuning such models requires a large number of training examples for each target task. Simultaneously, many realistic NLP problems are "few shot", without a sufficiently lar...
[ "few-shot text classification", "meta-learning", "transfer learning", "conditional neural process", "transformer", "pretrained language models" ]
We propose a novel model-based meta-learning approach for few-shot text classification using gradient information for task conditioning.
10,163
2201.11576
title_snapshot
R6U4-Qkcg21
Unsupervised Noise Adaptive Speech Enhancement by Discriminator-Constrained Optimal Transport
https://openreview.net/forum?id=R6U4-Qkcg21
[ "Hsin-Yi Lin", "Huan-Hsin Tseng", "Xugang Lu", "Yu Tsao" ]
Poster
null
This paper presents a novel discriminator-constrained optimal transport network (DOTN) that performs unsupervised domain adaptation for speech enhancement (SE), which is an essential regression task in speech processing. The DOTN aims to estimate clean references of noisy speech in a target domain, by exploiting the kn...
[ "Domain adaptation", "Optimal Transport", "adversarial training", "speech enhancement" ]
Unsupervised Noise Adaptive Speech Enhancement by Optimal Transport
10,158
2111.06316
title_snapshot
r-oRRT-ElX
On Provable Benefits of Depth in Training Graph Convolutional Networks
https://openreview.net/forum?id=r-oRRT-ElX
[ "Weilin Cong", "Morteza Ramezani", "Mehrdad Mahdavi" ]
Poster
null
Graph Convolutional Networks (GCNs) are known to suffer from performance degradation as the number of layers increases, which is usually attributed to over-smoothing. Despite the apparent consensus, we observe that there exists a discrepancy between the theoretical understanding of over-smoothing and the practical capa...
[ "graph neural network", "over-smoothing", "generalization", "expressive power", "optimization" ]
Reveal the discrepancy between the theoretical understanding of over-smoothing and the practical capabilities of GCNs.
10,149
2110.15174
title_snapshot
z3tlL2MeTK2
Nested Counterfactual Identification from Arbitrary Surrogate Experiments
https://openreview.net/forum?id=z3tlL2MeTK2
[ "Juan D. Correa", "Sanghack Lee", "Elias Bareinboim" ]
Poster
null
The Ladder of Causation describes three qualitatively different types of activities an agent may be interested in engaging in, namely, seeing (observational), doing (interventional), and imagining (counterfactual) (Pearl and Mackenzie, 2018). The inferential challenge imposed by the causal hierarchy is that data is col...
[ "counterfactuals", "identification", "causal inference", "experimental data" ]
Identification of nested counterfactuals from experimental distributions
10,135
2107.03190
title_snapshot
yKoZfSVFtAx
Efficient and Local Parallel Random Walks
https://openreview.net/forum?id=yKoZfSVFtAx
[ "Michael Kapralov", "Silvio Lattanzi", "Navid Nouri", "Jakab Tardos" ]
Poster
null
Random walks are a fundamental primitive used in many machine learning algorithms with several applications in clustering and semi-supervised learning. Despite their relevance, the first efficient parallel algorithm to compute random walks has been introduced very recently (Łącki et al.). Unfortunately their method has...
[ "personalized pagerank", "random walks", "MPC", "clustering" ]
Random walk simulation and clustering in the MPC model.
10,122
2112.00655
title_snapshot
6fmgB38rLI1
Multimodal and Multilingual Embeddings for Large-Scale Speech Mining
https://openreview.net/forum?id=6fmgB38rLI1
[ "Paul-Ambroise Duquenne", "Hongyu Gong", "Holger Schwenk" ]
Spotlight
null
We present an approach to encode a speech signal into a fixed-size representation which minimizes the cosine loss with the existing massively multilingual LASER text embedding space. Sentences are close in this embedding space, independently of their language and modality, either text or audio. Using a similarity metri...
[ "Speech Mining", "Large-scale mining", "Speech Translation" ]
We train a fixed-size speech embedding which is compatible with LASER text embedding. More than 20000h of mined speech translations, significant improvement of SOTA S2T system on CoVoST2. Proof of concept of direct speech-to-speech mining.
10,119
null
null
jcCatp6oWZK
Iterative Methods for Private Synthetic Data: Unifying Framework and New Methods
https://openreview.net/forum?id=jcCatp6oWZK
[ "Terrance Liu", "Giuseppe Vietri", "Steven Wu" ]
Poster
null
We study private synthetic data generation for query release, where the goal is to construct a sanitized version of a sensitive dataset, subject to differential privacy, that approximately preserves the answers to a large collection of statistical queries. We first present an algorithmic framework that unifies a long l...
[ "Privacy", "Differential Privacy", "Machine Learning", "Deep Learning", "Synthetic Data", "Query Release" ]
null
10,114
2106.07153
title_snapshot
zHj5fx11jQC
Marginalised Gaussian Processes with Nested Sampling
https://openreview.net/forum?id=zHj5fx11jQC
[ "Fergus Simpson", "Vidhi Lalchand", "Carl Edward Rasmussen" ]
Poster
null
Gaussian Process models are a rich distribution over functions with inductive biases controlled by a kernel function. Learning occurs through optimisation of the kernel hyperparameters using the marginal likelihood as the objective. This work proposes nested sampling as a means of marginalising kernel hyperparameters, ...
[ "Gaussian Processes", "nested sampling", "Bayesian inference" ]
We propose nested sampling as a promising means of marginalising kernel hyperparameters.
10,112
2010.16344
title_snapshot
4Il6i0jdrvP
Can we globally optimize cross-validation loss? Quasiconvexity in ridge regression
https://openreview.net/forum?id=4Il6i0jdrvP
[ "William T. Stephenson", "Zachary Frangella", "Madeleine Udell", "Tamara Broderick" ]
Poster
null
Models like LASSO and ridge regression are extensively used in practice due to their interpretability, ease of use, and strong theoretical guarantees. Cross-validation (CV) is widely used for hyperparameter tuning in these models, but do practical methods minimize the true out-of-sample loss? A recent line of researc...
[ "Cross-validation", "local optima", "optimization", "quasiconvexity", "ridge regression", "linear regression" ]
null
10,108
2107.09194
title_snapshot
bV89lw5OF8x
The Flip Side of the Reweighted Coin: Duality of Adaptive Dropout and Regularization
https://openreview.net/forum?id=bV89lw5OF8x
[ "Daniel LeJeune", "Hamid Javadi", "Richard Baraniuk" ]
Poster
null
Among the most successful methods for sparsifying deep (neural) networks are those that adaptively mask the network weights throughout training. By examining this masking, or dropout, in the linear case, we uncover a duality between such adaptive methods and regularization through the so-called “η-trick” that casts bot...
[ "sparsity", "deep learning", "dropout", "regularization", "adaptive" ]
We prove a duality between adaptive dropout sparsity methods and subquadratic regularization penalties.
10,107
2106.07769
title_snapshot
3h1iwXmYVVJ
Implicit Regularization in Matrix Sensing via Mirror Descent
https://openreview.net/forum?id=3h1iwXmYVVJ
[ "Fan Wu", "Patrick Rebeschini" ]
Poster
null
We study discrete-time mirror descent applied to the unregularized empirical risk in matrix sensing. In both the general case of rectangular matrices and the particular case of positive semidefinite matrices, a simple potential-based analysis in terms of the Bregman divergence allows us to establish convergence of mirr...
[ "mirror descent", "implicit regularization", "matrix sensing", "sparsity" ]
null
10,104
2105.13831
title_snapshot
5rm0b_fsNZ
Proportional Participatory Budgeting with Additive Utilities
https://openreview.net/forum?id=5rm0b_fsNZ
[ "Grzegorz Pierczyński", "Piotr Skowron", "Dominik Peters" ]
Poster
null
We study voting rules for participatory budgeting, where a group of voters collectively decides which projects should be funded using a common budget. We allow the projects to have arbitrary costs, and the voters to have arbitrary additive valuations over the projects. We formulate two axioms that guarantee proportiona...
[ "committee elections", "participatory budgeting", "aggregating preferences" ]
We study fairness of voting rules for participatory budgeting, where a group of voters collectively decides which projects should be funded using a common budget.
10,089
2008.13276
title_snapshot
af_hng9tuNj
GeoMol: Torsional Geometric Generation of Molecular 3D Conformer Ensembles
https://openreview.net/forum?id=af_hng9tuNj
[ "Octavian-Eugen Ganea", "Lagnajit Pattanaik", "Connor W. Coley", "Regina Barzilay", "Klavs Jensen", "William Green", "Tommi S. Jaakkola" ]
Spotlight
null
Prediction of a molecule’s 3D conformer ensemble from the molecular graph holds a key role in areas of cheminformatics and drug discovery. Existing generative models have several drawbacks including lack of modeling important molecular geometry elements (e.g., torsion angles), separate optimization stages prone to erro...
[ "molecular conformer generation", "molecular geometry", "molecular generative models", "message passing neural networks", "3D generation", "molecular conformations" ]
We generate ensembles of molecular 3D conformers from the input molecular graph in an end-to-end fashion by explicitly modeling local atomic 3D structures, torsion angles, chirality, and other geometric elements.
10,081
2106.07802
title_snapshot
dwJyEMPZ04I
CrypTen: Secure Multi-Party Computation Meets Machine Learning
https://openreview.net/forum?id=dwJyEMPZ04I
[ "Brian Knott", "Shobha Venkataraman", "Awni Hannun", "Shubhabrata Sengupta", "Mark Ibrahim", "Laurens van der Maaten" ]
Poster
null
Secure multi-party computation (MPC) allows parties to perform computations on data while keeping that data private. This capability has great potential for machine-learning applications: it facilitates training of machine-learning models on private data sets owned by different parties, evaluation of one party's privat...
[ "Secure multi-party computation", "deep-learning framework." ]
This paper presents CrypTen: a new framework for private and secure deep learning via secure multi-party computation.
10,078
2109.00984
title_snapshot
f-ggKIDTu5D
Exploiting Chain Rule and Bayes' Theorem to Compare Probability Distributions
https://openreview.net/forum?id=f-ggKIDTu5D
[ "Huangjie Zheng", "Mingyuan Zhou" ]
Poster
null
To measure the difference between two probability distributions, referred to as the source and target, respectively, we exploit both the chain rule and Bayes' theorem to construct conditional transport (CT), which is constituted by both a forward component and a backward one. The forward CT is the expected cost of movi...
[ "Bayesian methods", "conditional distribution", "generative models", "mode-covering", "mode-seeking" ]
Bayesian techniques are utilized to compare probability distributions, resulting in generative models that can control their mode-covering and mode-seeking behaviors.
10,074
2012.14100
title_snapshot
tMFTT3BDEK9
Look at What I’m Doing: Self-Supervised Spatial Grounding of Narrations in Instructional Videos
https://openreview.net/forum?id=tMFTT3BDEK9
[ "Reuben Tan", "Bryan A. Plummer", "Kate Saenko", "Hailin Jin", "Bryan Russell" ]
Spotlight
null
We introduce the task of spatially localizing narrated interactions in videos. Key to our approach is the ability to learn to spatially localize interactions with self-supervision on a large corpus of videos with accompanying transcribed narrations. To achieve this goal, we propose a multilayer cross-modal attention n...
[ "Self-supervision", "video understanding", "natural language" ]
Self-supervised interaction grounding in videos
10,067
2110.10596
title_snapshot
6veB3MCD-bu
A Law of Iterated Logarithm for Multi-Agent Reinforcement Learning
https://openreview.net/forum?id=6veB3MCD-bu
[ "Gugan Thoppe", "Bhumesh Kumar" ]
Poster
null
In Multi-Agent Reinforcement Learning (MARL), multiple agents interact with a common environment, as also with each other, for solving a shared problem in sequential decision-making. It has wide-ranging applications in gaming, robotics, finance, communication, etc. In this work, we derive a novel law of iterated logari...
[ "multi-agent", "distributed", "stochastic approximation", "almost sure", "convergence rates", "row stochastic", "gossip", "law of iterated logarithm", "concentration", "martingale" ]
Almost sure convergence rates for multi-agent reinforcement learning algorithms
10,066
2110.15092
title_snapshot
h2E5OYma5U
Causal-BALD: Deep Bayesian Active Learning of Outcomes to Infer Treatment-Effects from Observational Data
https://openreview.net/forum?id=h2E5OYma5U
[ "Andrew Jesson", "Panagiotis Tigas", "Joost van Amersfoort", "Andreas Kirsch", "Uri Shalit", "Yarin Gal" ]
Poster
null
Estimating personalized treatment effects from high-dimensional observational data is essential in situations where experimental designs are infeasible, unethical, or expensive. Existing approaches rely on fitting deep models on outcomes observed for treated and control populations. However, when measuring individual o...
[ "Causal Inference", "Active Learning", "Bayesian Deep Learning" ]
null
10,065
2111.02275
title_snapshot
fhDSTihtiB6
Finding Bipartite Components in Hypergraphs
https://openreview.net/forum?id=fhDSTihtiB6
[ "Peter Macgregor", "He Sun" ]
Poster
null
Hypergraphs are important objects to model ternary or higher-order relations of objects, and have a number of applications in analysing many complex datasets occurring in practice. In this work we study a new heat diffusion process in hypergraphs, and employ this process to design a polynomial-time algorithm that appr...
[ "hypergraph clustering", "spectral methods" ]
In this work we study a new heat diffusion process in hypergraphs, and design a polynomial-time algorithm that approximately finds bipartite components in a hypergraph.
10,064
2205.02771
title_snapshot
96ULbah4DC
Hierarchical Reinforcement Learning with Timed Subgoals
https://openreview.net/forum?id=96ULbah4DC
[ "Nico Gürtler", "Dieter Büchler", "Georg Martius" ]
Poster
null
Hierarchical reinforcement learning (HRL) holds great potential for sample-efficient learning on challenging long-horizon tasks. In particular, letting a higher level assign subgoals to a lower level has been shown to enable fast learning on difficult problems. However, such subgoal-based methods have been designed wit...
[ "hierarchical reinforcement learning", "reinforcement learning", "subgoals", "dynamic environments", "long-horizon task", "sparse reward", "non-stationary SMDP" ]
We present a hierarchical reinforcement learning algorithm that enables sample-efficient learning on sparse-reward tasks in dynamic environments.
10,056
2112.03100
title_snapshot
dsmxf7FKiaY
Revisiting ResNets: Improved Training and Scaling Strategies
https://openreview.net/forum?id=dsmxf7FKiaY
[ "Irwan Bello", "William Fedus", "Xianzhi Du", "Ekin Dogus Cubuk", "Aravind Srinivas", "Tsung-Yi Lin", "Jonathon Shlens", "Barret Zoph" ]
Spotlight
null
Novel computer vision architectures monopolize the spotlight, but the impact of the model architecture is often conflated with simultaneous changes to training methodology and scaling strategies. Our work revisits the canonical ResNet and studies these three aspects in an effort to disentangle them. Perhaps surprisingl...
[ "Scaling", "Architectures", "ResNet", "EfficientNet", "ViT", "Image Classification", "Vision", "ImageNet" ]
Improved training and scaling strategies make ResNet architectures competitive on image/video classification, suggesting that proper training and scaling matters more than architectural changes
10,054
2103.07579
title_snapshot
GPYHMC-MXl
Unifying Gradient Estimators for Meta-Reinforcement Learning via Off-Policy Evaluation
https://openreview.net/forum?id=GPYHMC-MXl
[ "Yunhao Tang", "Tadashi Kozuno", "Mark Rowland", "Remi Munos", "Michal Valko" ]
Poster
null
Model-agnostic meta-reinforcement learning requires estimating the Hessian matrix of value functions. This is challenging from an implementation perspective, as repeatedly differentiating policy gradient estimates may lead to biased Hessian estimates. In this work, we provide a unifying framework for estimating higher-...
[ "meta learning", "reinforcement learning", "off-policy learning", "off-policy evaluation" ]
off-policy evaluations provide a unified framework for estimating high-order derivatives of value functions in meta RL
10,039
2106.13125
title_snapshot
ZEhDWKLTvt7
Near Optimal Policy Optimization via REPS
https://openreview.net/forum?id=ZEhDWKLTvt7
[ "Aldo Pacchiano", "Jonathan Lee", "Peter Bartlett", "Ofir Nachum" ]
Poster
null
Since its introduction a decade ago, relative entropy policy search (REPS) has demonstrated successful policy learning on a number of simulated and real-world robotic domains, not to mention providing algorithmic components used by many recently proposed reinforcement learning (RL) algorithms. While REPS is commonly kn...
[ "reinforcement learning", "REPS", "linear programming" ]
We show convergence rates for REPS.
10,030
2103.09756
title_snapshot