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
j6KoGtzPYa
A Unified View of cGANs with and without Classifiers
https://openreview.net/forum?id=j6KoGtzPYa
[ "Si-An Chen", "Chun-Liang Li", "Hsuan-Tien Lin" ]
Poster
null
Conditional Generative Adversarial Networks (cGANs) are implicit generative models which allow to sample from class-conditional distributions. Existing cGANs are based on a wide range of different discriminator designs and training objectives. One popular design in earlier works is to include a classifier during traini...
[ "generative model", "generative adversarial network", "GAN", "conditional generation", "conditional GAN", "classifier" ]
We propose a new cGAN framework which provides a unified view of existing cGANs and achieves better results with multiple architectures and datasets.
10,028
2111.01035
title_snapshot
519VBzfEaKW
InfoGCL: Information-Aware Graph Contrastive Learning
https://openreview.net/forum?id=519VBzfEaKW
[ "Dongkuan Xu", "Wei Cheng", "Dongsheng Luo", "Haifeng Chen", "Xiang Zhang" ]
Poster
null
Various graph contrastive learning models have been proposed to improve the performance of tasks on graph datasets in recent years. While effective and prevalent, these models are usually carefully customized. In particular, despite all recent work create two contrastive views, they differ in a variety of view augmenta...
[ "Graph Representation Learning", "Contrastive Learning", "Information Bottleneck" ]
We study how graph information is transformed and transferred during the contrastive learning process, and propose an information-aware graph contrastive learning framework called InfoGCL.
10,026
2110.15438
title_snapshot
qGvMv3undNJ
Moser Flow: Divergence-based Generative Modeling on Manifolds
https://openreview.net/forum?id=qGvMv3undNJ
[ "Noam Rozen", "Aditya Grover", "Maximilian Nickel", "Yaron Lipman" ]
Oral
null
We are interested in learning generative models for complex geometries described via manifolds, such as spheres, tori, and other implicit surfaces. Current extensions of existing (Euclidean) generative models are restricted to specific geometries and typically suffer from high computational costs. We introduce Moser F...
[ "generative models", "manifolds", "normalizing flows" ]
Introducing a novel generative model on manifolds based on a classical flow by Moser.
10,022
2108.08052
title_snapshot
ZRcjSOmYraB
Learning Distilled Collaboration Graph for Multi-Agent Perception
https://openreview.net/forum?id=ZRcjSOmYraB
[ "Yiming Li", "Shunli Ren", "Pengxiang Wu", "Siheng Chen", "Chen Feng", "Wenjun Zhang" ]
Poster
null
To promote better performance-bandwidth trade-off for multi-agent perception, we propose a novel distilled collaboration graph (DiscoGraph) to model trainable, pose-aware, and adaptive collaboration among agents. Our key novelties lie in two aspects. First, we propose a teacher-student framework to train DiscoGraph via...
[ "knowledge distillation", "matrix-valued weighted graph", "multi-agent perception", "3d object detection" ]
Knowledge distillation on a collaboration graph achieves better performance-bandwidth trade-off for multi-agent perception.
10,021
2111.00643
title_snapshot
DXJl9826dm
Causal Influence Detection for Improving Efficiency in Reinforcement Learning
https://openreview.net/forum?id=DXJl9826dm
[ "Maximilian Seitzer", "Bernhard Schölkopf", "Georg Martius" ]
Poster
null
Many reinforcement learning (RL) environments consist of independent entities that interact sparsely. In such environments, RL agents have only limited influence over other entities in any particular situation. Our idea in this work is that learning can be efficiently guided by knowing when and what the agent can influ...
[ "reinforcement learning", "causal inference", "exploration", "intrinsic motivation", "prioritized replay" ]
We propose a method of detecting the causal influence of RL agents on the environment and use it to improve the sample efficiency of RL algorithms.
10,020
2106.03443
title_snapshot
5t5FPwzE6mq
Contrastive Active Inference
https://openreview.net/forum?id=5t5FPwzE6mq
[ "Pietro Mazzaglia", "Tim Verbelen", "Bart Dhoedt" ]
Poster
null
Active inference is a unifying theory for perception and action resting upon the idea that the brain maintains an internal model of the world by minimizing free energy. From a behavioral perspective, active inference agents can be seen as self-evidencing beings that act to fulfill their optimistic predictions, namely p...
[ "active", "inference", "contrastive", "learning", "free", "energy", "control" ]
Contrastive Active Inference, an active inference-based framework that aims to both reduce the complexity of the agent's internal model and to provide a more suitable objective to fulfill preferred outcomes, by exploiting contrastive learning.
10,019
2110.10083
title_snapshot
yxHPRAqCqn
Convergence Rates of Stochastic Gradient Descent under Infinite Noise Variance
https://openreview.net/forum?id=yxHPRAqCqn
[ "Hongjian Wang", "Mert Gurbuzbalaban", "Lingjiong Zhu", "Umut Simsekli", "Murat A Erdogdu" ]
Poster
null
Recent studies have provided both empirical and theoretical evidence illustrating that heavy tails can emerge in stochastic gradient descent (SGD) in various scenarios. Such heavy tails potentially result in iterates with diverging variance, which hinders the use of conventional convergence analysis techniques that rel...
[ "SGD", "heavy tailed noise", "infinite variance", "Polyak-Ruppert averaging" ]
We prove L_p convergence rates (p<2) for SGD under infinite noise variance, and establish the alpha-stable limit for Polyak-Ruppert averaging.
10,007
2102.10346
title_snapshot
AADxnPG-PR
Discovery of Options via Meta-Learned Subgoals
https://openreview.net/forum?id=AADxnPG-PR
[ "Vivek Veeriah", "Tom Zahavy", "Matteo Hessel", "Zhongwen Xu", "Junhyuk Oh", "Iurii Kemaev", "Hado van Hasselt", "David Silver", "Satinder Singh" ]
Poster
null
Temporal abstractions in the form of options have been shown to help reinforcement learning (RL) agents learn faster. However, despite prior work on this topic, the problem of discovering options through interaction with an environment remains a challenge. In this paper, we introduce a novel meta-gradient approach for ...
[ "reinforcement learning", "meta-gradients" ]
null
9,999
2102.06741
title_snapshot
41QJ--DLjoD
Landmark-RxR: Solving Vision-and-Language Navigation with Fine-Grained Alignment Supervision
https://openreview.net/forum?id=41QJ--DLjoD
[ "Keji He", "Yan Huang", "Qi Wu", "Jianhua Yang", "Dong An", "Shuanglin Sima", "Liang Wang" ]
Poster
null
In Vision-and-Language Navigation (VLN) task, an agent is asked to navigate inside 3D indoor environments following given instructions. Cross-modal alignment is one of the most critical challenges in VLN because the predicted trajectory needs to match the given instruction accurately. In this paper, we address the cros...
[ "Vision-and-Language Navigation", "Fine-grained dataset", "Cross-modal alignment", "Reward Shaping", "Re-initialization mechanism" ]
null
9,995
null
null
OgtWS4bkNO8
Photonic Differential Privacy with Direct Feedback Alignment
https://openreview.net/forum?id=OgtWS4bkNO8
[ "Ruben Ohana", "Hamlet Jesse Medina Ruiz", "Julien Launay", "Alessandro Cappelli", "Iacopo Poli", "Liva Ralaivola", "Alain Rakotomamonjy" ]
Poster
null
Optical Processing Units (OPUs) -- low-power photonic chips dedicated to large scale random projections -- have been used in previous work to train deep neural networks using Direct Feedback Alignment (DFA), an effective alternative to backpropagation. Here, we demonstrate how to leverage the intrinsic noise of optical...
[ "differential privacy", "direct feedback alignment", "optical computing", "deep learning", "alternative training methods" ]
We show that optical processors induce differential privacy for deep architectures learned through Direct Feedback Alignment.
9,992
2106.03645
title_snapshot
4hBXGTdS6Lc
Finding Regions of Heterogeneity in Decision-Making via Expected Conditional Covariance
https://openreview.net/forum?id=4hBXGTdS6Lc
[ "Justin Lim", "Christina X Ji", "Michael Oberst", "Saul Blecker", "Leora Horwitz", "David Sontag" ]
Poster
null
Individuals often make different decisions when faced with the same context, due to personal preferences and background. For instance, judges may vary in their leniency towards certain drug-related offenses, and doctors may vary in their preference for how to start treatment for certain types of patients. With these ...
[ "heterogeneity", "causal inference", "decision making", "criminal justice", "healthcare" ]
We present an algorithm for identifying regions of decisions with high inter-decision-maker disagreement.
9,984
2110.14508
title_snapshot
5KCvuCYGi7G
Subgoal Search For Complex Reasoning Tasks
https://openreview.net/forum?id=5KCvuCYGi7G
[ "Konrad Czechowski", "Tomasz Odrzygóźdź", "Marek Zbysiński", "Michał Zawalski", "Krzysztof Olejnik", "Yuhuai Wu", "Łukasz Kuciński", "Piotr Miłoś" ]
Poster
null
Humans excel in solving complex reasoning tasks through a mental process of moving from one idea to a related one. Inspired by this, we propose Subgoal Search (kSubS) method. Its key component is a learned subgoal generator that produces a diversity of subgoals that are both achievable and closer to the solution. Using...
[ "search", "deep learning", "hierarchical planning" ]
We propose a hierarchical search method for complex reasoning domains, based on learned subgoal generator.
9,981
2108.11204
title_snapshot
6VMXq5GCB9R
Entropy-based adaptive Hamiltonian Monte Carlo
https://openreview.net/forum?id=6VMXq5GCB9R
[ "Marcel Hirt", "Michalis Titsias", "Petros Dellaportas" ]
Poster
null
Hamiltonian Monte Carlo (HMC) is a popular Markov Chain Monte Carlo (MCMC) algorithm to sample from an unnormalized probability distribution. A leapfrog integrator is commonly used to implement HMC in practice, but its performance can be sensitive to the choice of mass matrix used therein. We develop a gradient-based a...
[ "Hamiltonian Monte Carlo", "MCMC" ]
We develop a gradient-based adaptation of the mass matrix within HMC based on an approximation of the marginal proposal entropy.
9,962
2110.14625
title_snapshot
01884FCwbNf
Robust Auction Design in the Auto-bidding World
https://openreview.net/forum?id=01884FCwbNf
[ "Santiago Balseiro", "Yuan Deng", "Jieming Mao", "Vahab Mirrokni", "Song Zuo" ]
Poster
null
In classic auction theory, reserve prices are known to be effective for improving revenue for the auctioneer against quasi-linear utility maximizing bidders. The introduction of reserve prices, however, usually do not help improve total welfare of the auctioneer and the bidders. In this paper, we focus on value maximiz...
[ "auto-bidding", "auction design", "reserve pricing" ]
We show how to robustly introduce reserve prices and boosts to increase welfare and revenue of auctions in the auto-bidding world.
9,960
2111.02468
title_snapshot
lLP77dROaJ
Understanding Deflation Process in Over-parametrized Tensor Decomposition
https://openreview.net/forum?id=lLP77dROaJ
[ "Rong Ge", "Yunwei Ren", "Xiang Wang", "Mo Zhou" ]
Poster
null
In this paper we study the training dynamics for gradient flow on over-parametrized tensor decomposition problems. Empirically, such training process often first fits larger components and then discovers smaller components, which is similar to a tensor deflation process that is commonly used in tensor decomposition alg...
[ "tensor decomposition", "overparametrization", "gradient flow", "implicit regularization" ]
The paper analyzes gradient flow on overparametrized tensor decomposition and proves that its dynamics is similar to a tensor deflation process.
9,950
2106.06573
title_snapshot
DtXBYsSOxCD
A first-order primal-dual method with adaptivity to local smoothness
https://openreview.net/forum?id=DtXBYsSOxCD
[ "Maria-Luiza Vladarean", "Yura Malitsky", "Volkan Cevher" ]
Poster
null
We consider the problem of finding a saddle point for the convex-concave objective $\min_x \max_y f(x) + \langle Ax, y\rangle - g^*(y)$, where $f$ is a convex function with locally Lipschitz gradient and $g$ is convex and possibly non-smooth. We propose an adaptive version of the Condat-Vũ algorithm, which alternates b...
[ "adaptive", "primal-dual", "convex-concave", "composite optimization", "local smoothness", "local Lipschitz continuity" ]
We propose a first-order primal-dual method with adaptivity to the local geometry of one of the objective's components.
9,933
2110.15148
title_snapshot
uholDBWSVP
Beyond Smoothness: Incorporating Low-Rank Analysis into Nonparametric Density Estimation
https://openreview.net/forum?id=uholDBWSVP
[ "Robert A. Vandermeulen", "Antoine Ledent" ]
Poster
null
The construction and theoretical analysis of the most popular universally consistent nonparametric density estimators hinge on one functional property: smoothness. In this paper we investigate the theoretical implications of incorporating a multi-view latent variable model, a type of low-rank model, into nonparametric ...
[ "Low-rank", "nonparametric", "density estimation", "statistical learning theory", "matrix factorization", "tensor factorization", "multi-view model" ]
We propose and analyze a low rank approach to nonparametric density estimation and theoretically find that such an approach can offer huge improvements in sample complexity.
9,931
2204.00930
title_snapshot
HwGNkx1WcIs
Object DGCNN: 3D Object Detection using Dynamic Graphs
https://openreview.net/forum?id=HwGNkx1WcIs
[ "Yue Wang", "Justin Solomon" ]
Poster
null
3D object detection often involves complicated training and testing pipelines, which require substantial domain knowledge about individual datasets. Inspired by recent non-maximum suppression-free 2D object detection models, we propose a 3D object detection architecture on point clouds. Our method models 3D object dete...
[ "3D object detection", "point clouds", "dynamic graph CNN" ]
A 3D object detection model on point clouds without post-processing.
9,929
2110.06923
title_snapshot
DfGu8WwT0d
Large Scale Learning on Non-Homophilous Graphs: New Benchmarks and Strong Simple Methods
https://openreview.net/forum?id=DfGu8WwT0d
[ "Derek Lim", "Felix Matthew Hohne", "Xiuyu Li", "Sijia Linda Huang", "Vaishnavi Gupta", "Omkar Prasad Bhalerao", "Ser-Nam Lim" ]
Poster
null
Many widely used datasets for graph machine learning tasks have generally been homophilous, where nodes with similar labels connect to each other. Recently, new Graph Neural Networks (GNNs) have been developed that move beyond the homophily regime; however, their evaluation has often been conducted on small graphs with...
[ "graph neural network", "homophily", "graphs", "datasets", "benchmarks" ]
We propose new datasets and a strong simple method for scalable learning on non-homophilous graphs.
9,919
2110.14446
title_snapshot
eNB4WXnNczJ
CANITA: Faster Rates for Distributed Convex Optimization with Communication Compression
https://openreview.net/forum?id=eNB4WXnNczJ
[ "Zhize Li", "Peter Richtárik" ]
Poster
null
Due to the high communication cost in distributed and federated learning, methods relying on compressed communication are becoming increasingly popular. Besides, the best theoretically and practically performing gradient-type methods invariably rely on some form of acceleration/momentum to reduce the number of communic...
[ "distributed optimization", "federated learning", "communication compression", "acceleration/momentum" ]
We provide the first gradient-type method provably combining the benefits of both compression (compressed communication in each round) and acceleration (much fewer communication rounds) for distributed convex optimization.
9,917
2107.09461
title_snapshot
UUds0Jr_XWk
Systematic Generalization with Edge Transformers
https://openreview.net/forum?id=UUds0Jr_XWk
[ "Leon Bergen", "Timothy J. O'Donnell", "Dzmitry Bahdanau" ]
Poster
null
Recent research suggests that systematic generalization in natural language understanding remains a challenge for state-of-the-art neural models such as Transformers and Graph Neural Networks. To tackle this challenge, we propose Edge Transformer, a new model that combines inspiration from Transformers and rule-based s...
[ "systematic generalization", "transformer", "semantic parsing", "language understanding" ]
We propose a new model that is inspired by transformers and logical reasoning. The proposed model generalizes more systematically.
9,915
2112.00578
title_snapshot
U7SBcmRf65
Training for the Future: A Simple Gradient Interpolation Loss to Generalize Along Time
https://openreview.net/forum?id=U7SBcmRf65
[ "Anshul Nasery", "Soumyadeep Thakur", "Vihari Piratla", "Abir De", "Sunita Sarawagi" ]
Poster
null
In several real world applications, machine learning models are deployed to make predictions on data whose distribution changes gradually along time, leading to a drift between the train and test distributions. Such models are often re-trained on new data periodically, and they hence need to generalize to data not too ...
[ "continuous domain generalization", "time-evolving classifiers", "gradient interpolation" ]
We propose a simple gradient interpolation loss to enable models to generalize to unseen future data.
9,911
2108.06721
title_snapshot
OSLVL-tIBei
Exploration-Exploitation in Multi-Agent Competition: Convergence with Bounded Rationality
https://openreview.net/forum?id=OSLVL-tIBei
[ "Stefanos Leonardos", "Georgios Piliouras", "Kelly Spendlove" ]
Spotlight
null
The interplay between exploration and exploitation in competitive multi-agent learning is still far from being well understood. Motivated by this, we study smooth Q-learning, a prototypical learning model that explicitly captures the balance between game rewards and exploration costs. We show that Q-learning always con...
[ "Exploration-Exploitation", "Q-learning", "Zero-Sum Polymatrix Games", "Quantal Response Equilibria", "Bounded Rationality" ]
Complementing recent results about convergence in weighted potential games, we show that Q-learning converges both in competitive as well as cooperative settings, regardless of the number of agents and without any need for parameter fine-tuning.
9,896
2106.12928
title_snapshot
hhU9TEvB6AF
Meta Learning Backpropagation And Improving It
https://openreview.net/forum?id=hhU9TEvB6AF
[ "Louis Kirsch", "Jürgen Schmidhuber" ]
Poster
null
Many concepts have been proposed for meta learning with neural networks (NNs), e.g., NNs that learn to reprogram fast weights, Hebbian plasticity, learned learning rules, and meta recurrent NNs. Our Variable Shared Meta Learning (VSML) unifies the above and demonstrates that simple weight-sharing and sparsity in an NN ...
[ "meta-learning", "general-purpose meta-learning", "learned learning rules", "fast weights", "distributed memory", "backpropagation", "gradient descent", "modularity", "self-organization" ]
Implementing backpropagation in recurrent neural networks and discovering novel general-purpose learning algorithms.
9,877
2012.14905
title_snapshot
ZqabiikWeyt
Learning to Learn Graph Topologies
https://openreview.net/forum?id=ZqabiikWeyt
[ "Xingyue Pu", "Tianyue Cao", "Xiaoyun Zhang", "Xiaowen Dong", "Siheng Chen" ]
Poster
null
Learning a graph topology to reveal the underlying relationship between data entities plays an important role in various machine learning and data analysis tasks. Under the assumption that structured data vary smoothly over a graph, the problem can be formulated as a regularised convex optimisation over a positive semi...
[ "topology inference", "graph learning", "algorithm unrolling", "learning to optimise" ]
Learning to Learn Graph Topologies
9,872
2110.09807
title_snapshot
jqjsLUrB8F
SGD: The Role of Implicit Regularization, Batch-size and Multiple-epochs
https://openreview.net/forum?id=jqjsLUrB8F
[ "Ayush Sekhari", "Karthik Sridharan", "Satyen Kale" ]
Poster
null
Multi-epoch, small-batch, Stochastic Gradient Descent (SGD) has been the method of choice for learning with large over-parameterized models. A popular theory for explaining why SGD works well in practice is that the algorithm has an implicit regularization that biases its output towards a good solution. Perhaps the the...
[ "Theory", "Implicit regularization", "Stochastic convex optimization", "SGD", "RERM", "Regularized ERM", "Multiple-passes", "GD" ]
null
9,870
2107.05074
title_snapshot
ekVPXh9tYkL
Deep Markov Factor Analysis: Towards Concurrent Temporal and Spatial Analysis of fMRI Data
https://openreview.net/forum?id=ekVPXh9tYkL
[ "Amirreza Farnoosh", "Sarah Ostadabbas" ]
Poster
null
Factor analysis methods have been widely used in neuroimaging to transfer high dimensional imaging data into low dimensional, ideally interpretable representations. However, most of these methods overlook the highly nonlinear and complex temporal dynamics of neural processes when factorizing their imaging data. In this...
[ "Deep generative models", "Bayesian dynamical matrix factorization", "Variational inference", "Cognitive neuroscience", "Functional MRI (fMRI)" ]
null
9,866
null
null
oepSB9bsoCF
Learning Domain Invariant Representations in Goal-conditioned Block MDPs
https://openreview.net/forum?id=oepSB9bsoCF
[ "Beining Han", "Chongyi Zheng", "Harris Chan", "Keiran Paster", "Michael R. Zhang", "Jimmy Ba" ]
Poster
null
Deep Reinforcement Learning (RL) is successful in solving many complex Markov Decision Processes (MDPs) problems. However, agents often face unanticipated environmental changes after deployment in the real world. These changes are often spurious and unrelated to the underlying problem, such as background shifts for vis...
[ "goal-conditioned reinforcement learning", "learning invariants", "representation learning", "domain generalization", "block mdp", "deep reinforcement learning" ]
We propose a new method for learning domain invariant policies for Goal-conditioned RL agents.
9,864
2110.14248
title_snapshot
Dzy8YEm5dX
Towards a Unified Information-Theoretic Framework for Generalization
https://openreview.net/forum?id=Dzy8YEm5dX
[ "Mahdi Haghifam", "Gintare Karolina Dziugaite", "Shay Moran", "Daniel M. Roy" ]
Spotlight
null
In this work, we investigate the expressiveness of the "conditional mutual information" (CMI) framework of Steinke and Zakynthinou (2020) and the prospect of using it to provide a unified framework for proving generalization bounds in the realizable setting. We first demonstrate that one can use this framework to ex...
[ "information-theoretic generalization", "conditional mutual information", "leave-one-out", "sample compression schemes", "SVM" ]
We show that the CMI framework can be used to obtain optimal or near-optimal bounds for the expected excess risk for a wide range of algorithms.
9,862
2111.05275
title_snapshot
zdTW91r2wKO
Active 3D Shape Reconstruction from Vision and Touch
https://openreview.net/forum?id=zdTW91r2wKO
[ "Edward J. Smith", "David Meger", "Luis Pineda", "Roberto Calandra", "Jitendra Malik", "Adriana Romero", "Michal Drozdzal" ]
Poster
null
Humans build 3D understandings of the world through active object exploration, using jointly their senses of vision and touch. However, in 3D shape reconstruction, most recent progress has relied on static datasets of limited sensory data such as RGB images, depth maps or haptic readings, leaving the active exploration...
[ "active touch", "active sensing", "3D reconstruction", "3D perception", "robotics" ]
This paper learns optimal grasps for touch driven 3D object reconstruction through proposed active learning methods over a contributed reconstruction model and visuotactile simulator.
9,848
2107.09584
title_snapshot
ZAOrF0mYSYU
Agnostic Reinforcement Learning with Low-Rank MDPs and Rich Observations
https://openreview.net/forum?id=ZAOrF0mYSYU
[ "Ayush Sekhari", "Christoph Dann", "Mehryar Mohri", "Yishay Mansour", "Karthik Sridharan" ]
Spotlight
null
There have been many recent advances on provably efficient Reinforcement Learning (RL) in problems with rich observation spaces. However, all these works share a strong realizability assumption about the optimal value function of the true MDP. Such realizability assumptions are often too strong to hold in practice. In ...
[ "Sample complexity", "Agnostic learning", "low rank MDP", "RL theory" ]
Provide algorithms for low rank MDPs with rich observations that do not require a realizable value function class, and instead focus on completing with a given policy class.
9,847
2106.11519
title_snapshot
JZK9uP4Fev
Learning Models for Actionable Recourse
https://openreview.net/forum?id=JZK9uP4Fev
[ "Alexis Ross", "Himabindu Lakkaraju", "Osbert Bastani" ]
Poster
null
As machine learning models are increasingly deployed in high-stakes domains such as legal and financial decision-making, there has been growing interest in post-hoc methods for generating counterfactual explanations. Such explanations provide individuals adversely impacted by predicted outcomes (e.g., an applicant deni...
[ "recourse", "actionable recourse", "machine learning", "adversarial training" ]
null
9,843
2011.06146
title_snapshot
JQznhE5mdyv
Reinforcement learning for optimization of variational quantum circuit architectures
https://openreview.net/forum?id=JQznhE5mdyv
[ "Mateusz Ostaszewski", "Lea Marion Trenkwalder", "Wojciech Masarczyk", "Eleanor Scerri", "Vedran Dunjko" ]
Poster
null
The study of Variational Quantum Eigensolvers (VQEs) has been in the spotlight in recent times as they may lead to real-world applications of near-term quantum devices. However, their performance depends on the structure of the used variational ansatz, which requires balancing the depth and expressivity of the corresp...
[ "Quantum Computing", "Reinforcement Learning", "Quantum Chemistry" ]
null
9,824
2103.16089
title_snapshot
Arn2E4IRjEB
Flow Network based Generative Models for Non-Iterative Diverse Candidate Generation
https://openreview.net/forum?id=Arn2E4IRjEB
[ "Emmanuel Bengio", "Moksh Jain", "Maksym Korablyov", "Doina Precup", "Yoshua Bengio" ]
Poster
null
This paper is about the problem of learning a stochastic policy for generating an object (like a molecular graph) from a sequence of actions, such that the probability of generating an object is proportional to a given positive reward for that object. Whereas standard return maximization tends to converge to a single r...
[ "sequential generative model", "temporal difference", "diversity", "molecules" ]
A sequential generative model based on flow networks, treating the root as a source and all possible generated objects as sinks, which generates objects proportionally to their reward.
9,820
2106.04399
title_snapshot
KCsNBfdYI7E
Towards Context-Agnostic Learning Using Synthetic Data
https://openreview.net/forum?id=KCsNBfdYI7E
[ "Charles Jin", "Martin Rinard" ]
Poster
null
We propose a novel setting for learning, where the input domain is the image of a map defined on the product of two sets, one of which completely determines the labels. We derive a new risk bound for this setting that decomposes into a bias and an error term, and exhibits a surprisingly weak dependence on the true labe...
[ "synthetic data", "context agnostic learning" ]
We train image classifiers using only a single synthetic image per class, which are also invariant to changes in context.
9,816
2005.14707
title_snapshot
Z8mLxlpSyrJ
An Empirical Investigation of Domain Generalization with Empirical Risk Minimizers
https://openreview.net/forum?id=Z8mLxlpSyrJ
[ "Shanmukha Ramakrishna Vedantam", "David Lopez-Paz", "David J. Schwab" ]
Poster
null
Recent work demonstrates that deep neural networks trained using Empirical Risk Minimization (ERM) can generalize under distribution shift, outperforming specialized training algorithms for domain generalization. The goal of this paper is to further understand this phenomenon. In particular, we study the extent to whic...
[ "Domain Generalization", "Empirical Risk", "Generalization Measures", "Deep Learning", "OOD", "Implicit Bias" ]
Explaining and Understanding Domain Adaptation Capabilities of Empirical Risk Minimization
9,810
null
null
_89s8ViNwwj
Parameterized Knowledge Transfer for Personalized Federated Learning
https://openreview.net/forum?id=_89s8ViNwwj
[ "Jie Zhang", "Song Guo", "Xiaosong Ma", "Haozhao Wang", "Wenchao Xu", "Feijie Wu" ]
Poster
null
In recent years, personalized federated learning (pFL) has attracted increasing attention for its potential in dealing with statistical heterogeneity among clients. However, the state-of-the-art pFL methods rely on model parameters aggregation at the server side, which require all models to have the same structure and ...
[ "Personalized Federated Learning", "Knowledge Transfer", "Parameterized Distillation" ]
We present, KT-pFL, a novel training method for personalized Federated Learning which can accommodate heterogeneous model structure for each client and achieve personalized knowledge transfer in each FL training round.
9,809
2111.02862
title_snapshot
p7GujbewmRY
Grounding inductive biases in natural images: invariance stems from variations in data
https://openreview.net/forum?id=p7GujbewmRY
[ "Diane Bouchacourt", "Mark Ibrahim", "Ari S. Morcos" ]
Poster
null
To perform well on unseen and potentially out-of-distribution samples, it is desirable for machine learning models to have a predictable response with respect to transformations affecting the factors of variation of the input. Here, we study the relative importance of several types of inductive biases towards such pred...
[ "data augmentation", "invariance", "transformations", "factors of variation" ]
ImageNet models' invariance generally stems from the data itself, and aligns with the data factors of variations, whereas architectural bias or data augmentation matter less.
9,797
2106.05121
title_snapshot
ahYIlRBeCFw
Beta-CROWN: Efficient Bound Propagation with Per-neuron Split Constraints for Neural Network Robustness Verification
https://openreview.net/forum?id=ahYIlRBeCFw
[ "Shiqi Wang", "Huan Zhang", "Kaidi Xu", "Xue Lin", "Suman Jana", "Cho-Jui Hsieh", "J Zico Kolter" ]
Poster
null
Bound propagation based incomplete neural network verifiers such as CROWN are very efficient and can significantly accelerate branch-and-bound (BaB) based complete verification of neural networks. However, bound propagation cannot fully handle the neuron split constraints introduced by BaB commonly handled by expensive...
[ "neural network verification", "branch and bound", "adversarial robustness", "adversarial examples", "certified robustness" ]
We proposed an efficient branch and bound based neural network verifier using a new bound propagation based method that can fully encode neuron split constraints, outperforming existing baselines in both complete and incomplete verification settings.
9,784
2103.06624
title_judge
-VjKyYX-PI9
Sparsely Changing Latent States for Prediction and Planning in Partially Observable Domains
https://openreview.net/forum?id=-VjKyYX-PI9
[ "Christian Gumbsch", "Martin V. Butz", "Georg Martius" ]
Poster
null
A common approach to prediction and planning in partially observable domains is to use recurrent neural networks (RNNs), which ideally develop and maintain a latent memory about hidden, task-relevant factors. We hypothesize that many of these hidden factors in the physical world are constant over time, changing only sp...
[ "recurrent neural networks", "partial observability", "sparsity", "explainability", "generalization", "model-based planning", "reinforcement learning" ]
We present GateL0RD, an RNN that sparsely updates its latent states, for prediction and control with long-term memorization, better generalization, and explainable latent states.
9,780
2110.15949
title_snapshot
kB8eks2Edt8
Faster Matchings via Learned Duals
https://openreview.net/forum?id=kB8eks2Edt8
[ "Michael Dinitz", "Sungjin Im", "Thomas Lavastida", "Benjamin Moseley", "Sergei Vassilvitskii" ]
Oral
null
A recent line of research investigates how algorithms can be augmented with machine-learned predictions to overcome worst case lower bounds. This area has revealed interesting algorithmic insights into problems, with particular success in the design of competitive online algorithms. However, the question of improving...
[ "Bipartite matching", "run time", "learning" ]
null
9,776
2107.09770
title_snapshot
m72s2rDrm3G
Exploiting Opponents Under Utility Constraints in Sequential Games
https://openreview.net/forum?id=m72s2rDrm3G
[ "Martino Bernasconi", "Federico Cacciamani", "Simone Fioravanti", "Nicola Gatti", "Alberto Marchesi", "Francesco Trovò" ]
Poster
null
Recently, game-playing agents based on AI techniques have demonstrated super-human performance in several sequential games, such as chess, Go, and poker. Surprisingly, the multi-agent learning techniques that allowed to reach these achievements do not take into account the actual behavior of the human player, potential...
[ "Sequential games", "online learning", "opponent exploitation" ]
null
9,765
null
null
b-88mXTMg4J
Dual Parameterization of Sparse Variational Gaussian Processes
https://openreview.net/forum?id=b-88mXTMg4J
[ "Vincent ADAM", "Paul Edmund Chang", "Mohammad Emtiyaz Khan", "Arno Solin" ]
Poster
null
Sparse variational Gaussian process (SVGP) methods are a common choice for non-conjugate Gaussian process inference because of their computational benefits. In this paper, we improve their computational efficiency by using a dual parameterization where each data example is assigned dual parameters, similarly to site pa...
[ "Gaussian processes", "sparse variational inference", "natural gradients" ]
Leveraging dual-parameterization for efficient inference and learning of hyperparameters in sparse variational GP models
9,762
2111.03412
title_snapshot
ibD-yZEVBUX
Towards Biologically Plausible Convolutional Networks
https://openreview.net/forum?id=ibD-yZEVBUX
[ "Roman Pogodin", "Yash Mehta", "Timothy P Lillicrap", "Peter E. Latham" ]
Poster
null
Convolutional networks are ubiquitous in deep learning. They are particularly useful for images, as they reduce the number of parameters, reduce training time, and increase accuracy. However, as a model of the brain they are seriously problematic, since they require weight sharing - something real neurons simply cannot...
[ "biologically plausible deep learning", "convolutional networks", "Hebbian plasticity" ]
Two approaches for convolutional-like architectures without explicit weight sharing
9,756
2106.13031
title_snapshot
L8-54wkift
Lower Bounds and Optimal Algorithms for Smooth and Strongly Convex Decentralized Optimization Over Time-Varying Networks
https://openreview.net/forum?id=L8-54wkift
[ "Dmitry Kovalev", "Elnur Gasanov", "Alexander Gasnikov", "Peter Richtárik" ]
Poster
null
We consider the task of minimizing the sum of smooth and strongly convex functions stored in a decentralized manner across the nodes of a communication network whose links are allowed to change in time. We solve two fundamental problems for this task. First, we establish {\em the first lower bounds} on the number of de...
[ "decentralized optimization", "time-varying networks", "ADOM", "optimal algorithms", "lower bounds", "convex optimization" ]
We establish the first lower bounds for decentralized optimization over time-varying networks, and propose two optimal algorithms that match these bounds.
9,754
2106.04469
title_snapshot
uPWdkoZHgba
Two-sided fairness in rankings via Lorenz dominance
https://openreview.net/forum?id=uPWdkoZHgba
[ "Virginie Do", "Sam Corbett-Davies", "Jamal Atif", "Nicolas Usunier" ]
Poster
null
We consider the problem of generating rankings that are fair towards both users and item producers in recommender systems. We address both usual recommendation (e.g., of music or movies) and reciprocal recommendation (e.g., dating). Following concepts of distributive justice in welfare economics, our notion of fairness...
[ "fairness", "ranking", "recommendation", "welfare" ]
A unified framework for two-sided fairness in recommendation from a welfare economics perspective.
9,747
2110.15781
title_snapshot
-h6Ldc0MO-
Out-of-Distribution Generalization in Kernel Regression
https://openreview.net/forum?id=-h6Ldc0MO-
[ "Abdulkadir Canatar", "Blake Bordelon", "Cengiz Pehlevan" ]
Poster
null
In real word applications, data generating process for training a machine learning model often differs from what the model encounters in the test stage. Understanding how and whether machine learning models generalize under such distributional shifts have been a theoretical challenge. Here, we study generalization in ...
[ "kernel regression", "generalization", "out-of-distribution" ]
null
9,740
2106.02261
title_snapshot
d1FHmxHPEQ0
Curriculum Disentangled Recommendation with Noisy Multi-feedback
https://openreview.net/forum?id=d1FHmxHPEQ0
[ "Hong Chen", "Yudong Chen", "Xin Wang", "Ruobing Xie", "Rui Wang", "Feng Xia", "Wenwu Zhu" ]
Poster
null
Learning disentangled representations for user intentions from multi-feedback (i.e., positive and negative feedback) can enhance the accuracy and explainability of recommendation algorithms. However, learning such disentangled representations from multi-feedback data is challenging because i) multi-feedback is complex...
[ "Recommender system", "Disentangled representation learning", "Curriculum learning", "Multi-feedback" ]
Learn disentangled user intentions from user multi-feedback with a newly-proposed curriculum training strategy.
9,734
null
null
9c-IsSptbmA
Be Confident! Towards Trustworthy Graph Neural Networks via Confidence Calibration
https://openreview.net/forum?id=9c-IsSptbmA
[ "Xiao Wang", "Hongrui Liu", "Chuan Shi", "Cheng Yang" ]
Poster
null
Despite Graph Neural Networks (GNNs) have achieved remarkable accuracy, whether the results are trustworthy is still unexplored. Previous studies suggest that many modern neural networks are over-confident on the predictions, however, surprisingly, we discover that GNNs are primarily in the opposite direction, i.e., GN...
[ "Graph Neural Networks", "Confidence Calibration" ]
null
9,730
2109.14285
title_snapshot
Pf9RjFoUdLZ
LSH-SMILE: Locality Sensitive Hashing Accelerated Simulation and Learning
https://openreview.net/forum?id=Pf9RjFoUdLZ
[ "Chonghao Sima", "Yexiang Xue" ]
Poster
null
The advancement of deep neural networks over the last decade has enabled progress in scientific knowledge discovery in the form of learning Partial Differential Equations (PDEs) directly from experiment data. Nevertheless, forward simulation and backward learning of large-scale dynamic systems require handling billions...
[ "Partial Differential Equation", "Locality Sensitive Hashing", "Learning-driven Scientific Discovery", "Learning Physics Systems", "Acceleration" ]
Camera ready for LSH-SMILE: Locality Sensitive Hashing Accelerated Simulation and Learning
9,721
null
null
oAxm0Wz7Bv
Make Sure You're Unsure: A Framework for Verifying Probabilistic Specifications
https://openreview.net/forum?id=oAxm0Wz7Bv
[ "Leonard Berrada", "Sumanth Dathathri", "Krishnamurthy Dj Dvijotham", "Robert Stanforth", "Rudy R Bunel", "Jonathan Uesato", "Sven Gowal", "M. Pawan Kumar" ]
Spotlight
null
Most real world applications require dealing with stochasticity like sensor noise or predictive uncertainty, where formal specifications of desired behavior are inherently probabilistic. Despite the promise of formal verification in ensuring the reliability of neural networks, progress in the direction of probabilisti...
[ "Verification", "Safety", "Optimization" ]
Verification of probabilistic specifications
9,716
2102.09479
title_snapshot
1gLyEmOsKE8
Towards Best-of-All-Worlds Online Learning with Feedback Graphs
https://openreview.net/forum?id=1gLyEmOsKE8
[ "Liad Erez", "Tomer Koren" ]
Poster
null
We study the online learning with feedback graphs framework introduced by Mannor and Shamir (2011), in which the feedback received by the online learner is specified by a graph $G$ over the available actions. We develop an algorithm that simultaneously achieves regret bounds of the form: $O(\sqrt{\theta(G) T})$ with a...
[ "Online Learning", "Multi-armed Bandit", "Feedback Graphs", "Adversarial Corruptions" ]
We provide the first best-of-all-worlds regret guarantee for online learning with graph-structured feedback via a novel Tsallis-Shannon regularization.
9,703
2107.09572
title_judge
UmCsy3C4xj
Learning Fast-Inference Bayesian Networks
https://openreview.net/forum?id=UmCsy3C4xj
[ "Vaidyanathan Peruvemba Ramaswamy", "Stefan Szeider" ]
Poster
null
We propose new methods for learning Bayesian networks (BNs) that reliably support fast inference. We utilize maximum state space size as a more fine-grained measure for the BN's reasoning complexity than the standard treewidth measure, thereby accommodating the possibility that variables range over domains of different...
[ "Bayesian Network Structure Learning", "Exact Probabilistic Reasoning", "MaxSAT", "Propositional Satisfiability" ]
We propose a new Max-SAT based approach for learning BNs that admit reliably fast exact probabilistic reasoning
9,689
null
null
Tc6Uk03Te7g
Optimizing Conditional Value-At-Risk of Black-Box Functions
https://openreview.net/forum?id=Tc6Uk03Te7g
[ "Quoc Phong Nguyen", "Zhongxiang Dai", "Bryan Kian Hsiang Low", "Patrick Jaillet" ]
Poster
null
This paper presents two Bayesian optimization (BO) algorithms with theoretical performance guarantee to maximize the conditional value-at-risk (CVaR) of a black-box function: CV-UCB and CV-TS which are based on the well-established principle of optimism in the face of uncertainty and Thompson sampling, respectively. To...
[ "Bayesian optimization", "conditional value-at-risk", "CVaR", "UCB", "Thompson sampling" ]
To optimize the conditional value-at-risk of a black-box function, we develop two Bayesian optimization algorithms with performance guarantees, one of which can handle batch queries.
9,687
null
null
HEVfOwxrmQh
Better Algorithms for Individually Fair $k$-Clustering
https://openreview.net/forum?id=HEVfOwxrmQh
[ "Maryam Negahbani", "Deeparnab Chakrabarty" ]
Poster
null
We study data clustering problems with $\ell_p$-norm objectives (e.g. \textsc{$k$-Median} and \textsc{$k$-Means}) in the context of individual fairness. The dataset consists of $n$ points, and we want to find $k$ centers such that (a) the objective is minimized, while (b) respecting the individual fairness constraint t...
[ "Clustering", "Fairness", "Approximation", "Unsupervised-Learning" ]
Better algorithms for individually fair lp-norm k-clustering problems with provable guarantees, using LP rounding.
9,679
2106.12150
title_snapshot
fpQojkIV5q8
DropGNN: Random Dropouts Increase the Expressiveness of Graph Neural Networks
https://openreview.net/forum?id=fpQojkIV5q8
[ "Pál András Papp", "Karolis Martinkus", "Lukas Faber", "Roger Wattenhofer" ]
Spotlight
null
This paper studies Dropout Graph Neural Networks (DropGNNs), a new approach that aims to overcome the limitations of standard GNN frameworks. In DropGNNs, we execute multiple runs of a GNN on the input graph, with some of the nodes randomly and independently dropped in each of these runs. Then, we combine the results o...
[ "graph neural networks", "GNN", "expressivity", "WL", "molecular graphs" ]
We devise a new GNN variant (DropGNN) with larger expressive power in both theory and practice
9,676
2111.06283
title_snapshot
jK9Hy4qJsB
Dynamic COVID risk assessment accounting for community virus exposure from a spatial-temporal transmission model
https://openreview.net/forum?id=jK9Hy4qJsB
[ "Yuan Chen", "Wenbo Fei", "Qinxia Wang", "Donglin Zeng", "Yuanjia Wang" ]
Poster
null
COVID-19 pandemic has caused unprecedented negative impacts on our society, including further exposing inequity and disparity in public health. To study the impact of socioeconomic factors on COVID transmission, we first propose a spatial-temporal model to examine the socioeconomic heterogeneity and spatial correlation...
[ "Spatial temporal modeling", "COVID-19 pandemic", "Health disparity", "Selection bias", "Gaussian process", "Disease transmission", "Latent process" ]
We propose a spatial-temporal model for COVID-19 transmission and a bias-corrected semiparametric individual risk assessment model for severe COVID outcomes to facilitate clinical management and public health decision making.
9,675
null
null
LKntyz8tp1S
Mean-based Best Arm Identification in Stochastic Bandits under Reward Contamination
https://openreview.net/forum?id=LKntyz8tp1S
[ "Arpan Mukherjee", "Ali Tajer", "Pin-Yu Chen", "Payel Das" ]
Poster
null
This paper investigates the problem of best arm identification in {\sl contaminated} stochastic multi-arm bandits. In this setting, the rewards obtained from any arm are replaced by samples from an adversarial model with probability $\varepsilon$. A fixed confidence (infinite-horizon) setting is considered, where the g...
[ "Best arm identification", "contaminated stochastic bandits" ]
null
9,665
2111.07458
title_snapshot
AVWROGUWpu
Topographic VAEs learn Equivariant Capsules
https://openreview.net/forum?id=AVWROGUWpu
[ "T. Anderson Keller", "Max Welling" ]
Poster
null
In this work we seek to bridge the concepts of topographic organization and equivariance in neural networks. To accomplish this, we introduce the Topographic VAE: a novel method for efficiently training deep generative models with topographically organized latent variables. We show that such a model indeed learns to or...
[ "Topographic Organization", "Unsupervised", "Equivariance", "Variational Inference", "Deep Generative Model", "Disentanglement" ]
We show how deep topographic generative models can be efficiently trained and subsequently leverged to learn equivariant sets of features from unsupervised sequences.
9,663
2109.01394
title_snapshot
YscYPF8bU13
A Central Limit Theorem for Differentially Private Query Answering
https://openreview.net/forum?id=YscYPF8bU13
[ "Jinshuo Dong", "Weijie J Su", "Linjun Zhang" ]
Spotlight
null
Perhaps the single most important use case for differential privacy is to privately answer numerical queries, which is usually achieved by adding noise to the answer vector. The central question is, therefore, to understand which noise distribution optimizes the privacy-accuracy trade-off, especially when the dimension...
[ "differential privacy", "central limit theorem", "uncertainty principle", "Cramer-Rao lower bound" ]
Adding even highly correlated noise to a high-dimensional query yields Gaussian Differential Privacy. Privacy-accuracy trade-off is related to Cramer-Rao lower bound.
9,649
2103.08721
title_snapshot
MySjw6CHPa4
Self-Supervised Learning of Event-Based Optical Flow with Spiking Neural Networks
https://openreview.net/forum?id=MySjw6CHPa4
[ "Jesse Hagenaars", "Federico Paredes Valles", "Guido De Croon" ]
Poster
null
The field of neuromorphic computing promises extremely low-power and low-latency sensing and processing. Challenges in transferring learning algorithms from traditional artificial neural networks (ANNs) to spiking neural networks (SNNs) have so far prevented their application to large-scale, complex regression tasks. F...
[ "spiking neural networks", "neuromorphic computing", "event camera", "self-supervised learning", "optical flow" ]
We present the first set of deep spiking neural networks to successfully solve the real-world large-scale problem of event-based optical flow estimation.
9,641
2106.01862
title_snapshot
zDtFO9vohmF
Kernel Functional Optimisation
https://openreview.net/forum?id=zDtFO9vohmF
[ "Arun Kumar Anjanapura Venkatesh", "Alistair Shilton", "Santu Rana", "Sunil Gupta", "Svetha Venkatesh" ]
Poster
null
Traditional methods for kernel selection rely on parametric kernel functions or a combination thereof and although the kernel hyperparameters are tuned, these methods often provide sub-optimal results due to the limitations induced by the parametric forms. In this paper, we propose a novel formulation for kernel select...
[ "Non-parametric kernels", "Kernel learning", "Bayesian functional optimisation", "Hyperkernels", "Kernel machines", "Gaussian Process", "Hyperparameter tuning", "Machine learning", "Support Vector Machines", "Reproducing Kernel Hilbert Spaces", "Black-box optimisation", "sample-efficient optim...
We propose a novel approach for the optimisation of kernel functionals using efficient Bayesian functional optimisation.
9,630
null
null
73OmmrCfSyy
Mind the Gap: Assessing Temporal Generalization in Neural Language Models
https://openreview.net/forum?id=73OmmrCfSyy
[ "Angeliki Lazaridou", "Adhiguna Kuncoro", "Elena Gribovskaya", "Devang Agrawal", "Adam Liska", "Tayfun Terzi", "Mai Gimenez", "Cyprien de Masson d'Autume", "Tomáš Kočiský", "Sebastian Ruder", "Dani Yogatama", "Kris Cao", "Susannah Young", "Phil Blunsom" ]
Spotlight
null
Our world is open-ended, non-stationary, and constantly evolving; thus what we talk about and how we talk about it change over time. This inherent dynamic nature of language contrasts with the current static language modelling paradigm, which trains and evaluates models on utterances from overlapping time periods. Desp...
[ "language modelling", "temporal splits", "model analysis" ]
We test and analyze temporal generalization capabilities of neural language models using time-stratified datasets.
9,628
2102.01951
title_snapshot
D-ti-5lgbG
COHESIV: Contrastive Object and Hand Embedding Segmentation In Video
https://openreview.net/forum?id=D-ti-5lgbG
[ "Dandan Shan", "Richard Ely Locke Higgins", "David Fouhey" ]
Poster
null
In this paper we learn to segment hands and hand-held objects from motion. Our system takes a single RGB image and hand location as input to segment the hand and hand-held object. For learning, we generate responsibility maps that show how well a hand's motion explains other pixels' motion in video. We use these respon...
[ "Hand Object Interaction", "Object Segmentation", "Contrastive Learning", "Attention", "Embeddings" ]
In this paper we learn to segment hands and hand-held objects using attention and contrastive-based learning.
9,627
null
null
Ib6VSrtZcu9
A Provably Efficient Model-Free Posterior Sampling Method for Episodic Reinforcement Learning
https://openreview.net/forum?id=Ib6VSrtZcu9
[ "Christoph Dann", "Mehryar Mohri", "Tong Zhang", "Julian Zimmert" ]
Poster
null
Thompson Sampling is one of the most effective methods for contextual bandits and has been generalized to posterior sampling for certain MDP settings. However, existing posterior sampling methods for reinforcement learning are limited by being model-based or lack worst-case theoretical guarantees beyond linear MDPs. Th...
[ "Reinforcement Learning", "Posterior Sampling", "Regret Bounds", "Theory" ]
null
9,626
2208.10904
title_snapshot
xJYek6zantM
Tighter Expected Generalization Error Bounds via Wasserstein Distance
https://openreview.net/forum?id=xJYek6zantM
[ "Borja Rodríguez Gálvez", "German Bassi", "Ragnar Thobaben", "Mikael Skoglund" ]
Poster
null
This work presents several expected generalization error bounds based on the Wasserstein distance. More specifically, it introduces full-dataset, single-letter, and random-subset bounds, and their analogous in the randomized subsample setting from Steinke and Zakynthinou [1]. Moreover, when the loss function is bounded...
[ "generalization error", "wasserstein distance" ]
New, tighter generalization error bounds based on the Wasserstein distance that recover from below previous bounds based on the relative entropy and generate several new ones.
9,621
2101.09315
title_snapshot
1oP1duoZxx
Boosting with Multiple Sources
https://openreview.net/forum?id=1oP1duoZxx
[ "Corinna Cortes", "Mehryar Mohri", "Dmitry Storcheus", "Ananda Theertha Suresh" ]
Poster
null
We study the problem of learning accurate ensemble predictors, in particular boosting, in the presence of multiple source domains. We show that the standard convex combination ensembles in general cannot succeed in this scenario and adopt instead a domain-weighted combination. We introduce and analyze a new bo...
[ "boosting", "multiple sources", "generalization bounds", "federated learning" ]
Novel multi-source boosting algorithm with domain-weighted combination of weak learners.
9,616
null
null
OLyhLK2eQP
Knowledge-inspired 3D Scene Graph Prediction in Point Cloud
https://openreview.net/forum?id=OLyhLK2eQP
[ "Shoulong Zhang", "Shuai Li", "Aimin Hao", "Hong Qin" ]
Poster
null
Prior knowledge integration helps identify semantic entities and their relationships in a graphical representation, however, its meaningful abstraction and intervention remain elusive. This paper advocates a knowledge-inspired 3D scene graph prediction method solely based on point clouds. At the mathematical modeling l...
[ "3D Scene Graph", "Knowledge Embedding", "Deep Learning" ]
null
9,603
null
null
28NikxkK6kJ
For high-dimensional hierarchical models, consider exchangeability of effects across covariates instead of across datasets
https://openreview.net/forum?id=28NikxkK6kJ
[ "Brian L. Trippe", "Hilary Finucane", "Tamara Broderick" ]
Poster
null
Hierarchical Bayesian methods enable information sharing across regression problems on multiple groups of data. While standard practice is to model regression parameters (effects) as (1) exchangeable across the groups and (2) correlated to differing degrees across covariates, we show that this approach exhibits poor st...
[ "Hierarchical Bayesian modeling", "linear models", "exchangeability", "frequentist properties" ]
For high-dimensional hierarchical linear models, modeling effects exchangeably across covariates instead of across groups yields better estimates.
9,602
2107.06428
title_snapshot
kGXlIEQgvC
CoFrNets: Interpretable Neural Architecture Inspired by Continued Fractions
https://openreview.net/forum?id=kGXlIEQgvC
[ "Isha Puri", "Amit Dhurandhar", "Tejaswini Pedapati", "Karthikeyan Shanmugam", "Dennis Wei", "Kush R. Varshney" ]
Poster
null
In recent years there has been a considerable amount of research on local post hoc explanations for neural networks. However, work on building interpretable neural architectures has been relatively sparse. In this paper, we present a novel neural architecture, CoFrNet, inspired by the form of continued fractions which ...
[ "neural architecture design", "explainable AI", "continued fractions" ]
Propose a new interpretable neural architecture inspired by continued fractions
9,592
2506.05586
title_snapshot
52XXcK8jY0J
Disentangling Identifiable Features from Noisy Data with Structured Nonlinear ICA
https://openreview.net/forum?id=52XXcK8jY0J
[ "Hermanni Hälvä", "Sylvain Le Corff", "Luc Lehéricy", "Jonathan So", "Yongjie Zhu", "Elisabeth Gassiat", "Aapo Hyvarinen" ]
Poster
null
We introduce a new general identifiable framework for principled disentanglement referred to as Structured Nonlinear Independent Component Analysis (SNICA). Our contribution is to extend the identifiability theory of deep generative models for a very broad class of structured models. While previous works have shown ide...
[ "nonlinear ICA", "ICA", "identifiable", "disentanglement", "representation learning", "deep generative models", "denoising" ]
New general identifiable framework for principled disentanglement using nonlinear ICA called Structured Nonlinear Independent Component Analysis (SNICA).
9,580
2106.09620
title_snapshot
0kCxbBQknN
Qu-ANTI-zation: Exploiting Quantization Artifacts for Achieving Adversarial Outcomes
https://openreview.net/forum?id=0kCxbBQknN
[ "Sanghyun Hong", "Michael-Andrei Panaitescu-Liess", "Yigitcan Kaya", "Tudor Dumitras" ]
Poster
null
Quantization is a popular technique that transforms the parameter representation of a neural network from floating-point numbers into lower-precision ones (e.g., 8-bit integers). It reduces the memory footprint and the computational cost at inference, facilitating the deployment of resource-hungry models. However, the ...
[ "Neural network quantization", "adversarial attack", "backdoor attack", "behavioral disparity" ]
We study the security vulnerability caused by exploiting the behavioral disparity that neural network quantization introduces to a model.
9,575
2110.13541
title_snapshot
CmI7NqBR4Ua
An Improved Analysis of Gradient Tracking for Decentralized Machine Learning
https://openreview.net/forum?id=CmI7NqBR4Ua
[ "Anastasia Koloskova", "Tao Lin", "Sebastian U Stich" ]
Poster
null
We consider decentralized machine learning over a network where the training data is distributed across $n$ agents, each of which can compute stochastic model updates on their local data. The agent's common goal is to find a model that minimizes the average of all local loss functions. While gradient tracking (GT) algo...
[ "Decentralized stochastic optimization", "Gradient tracking", "non-iid data", "decentralized optimization" ]
Improved analysis of Gradient Tracking with new proof technique.
9,547
2202.03836
title_snapshot
yGKklt8wyV
Graph Neural Networks with Local Graph Parameters
https://openreview.net/forum?id=yGKklt8wyV
[ "Pablo Barcelo", "Floris Geerts", "Juan L Reutter", "Maksimilian Ryschkov" ]
Poster
null
Various recent proposals increase the distinguishing power of Graph Neural Networks (GNNs) by propagating features between k-tuples of vertices. The distinguishing power of these “higher-order” GNNs is known to be bounded by the k-dimensional Weisfeiler-Leman (WL) test, yet their O(n^k) memory requirements limit their ...
[ "graph neural network", "GNN", "finite model theory", "homomorphism counts" ]
Adding hom counts of patterns to feature vectors of GNNs increases their power. We show this theoretically with a precise characterization and tools to recommend which patterns are best. We also show it empirically.
9,526
2106.06707
title_snapshot
njIekVo3wLP
Fast Routing under Uncertainty: Adaptive Learning in Congestion Games via Exponential Weights
https://openreview.net/forum?id=njIekVo3wLP
[ "Dong Quan Vu", "Kimon Antonakopoulos", "Panayotis Mertikopoulos" ]
Poster
null
We examine an adaptive learning framework for nonatomic congestion games where the players' cost functions may be subject to exogenous fluctuations (e.g., due to disturbances in the network, variations in the traffic going through a link). In this setting, the popular multiplicative/ exponential weights algorithm enjoy...
[ "game theory", "adaptive algorithms", "routing" ]
We consider the problem of learning the equilibrium flows in routing games under uncertainty and propose an adaptive algorithm achieving optimal convergence rates in both the stochastic setting and the static setting.
9,525
null
null
1yeYYtLqq7K
A flow-based latent state generative model of neural population responses to natural images
https://openreview.net/forum?id=1yeYYtLqq7K
[ "Mohammad Bashiri", "Edgar Y. Walker", "Konstantin-Klemens Lurz", "Akshay Kumar Jagadish", "Taliah Muhammad", "Zhiwei Ding", "Zhuokun Ding", "Andreas S. Tolias", "Fabian H. Sinz" ]
Spotlight
null
We present a joint deep neural system identification model for two major sources of neural variability: stimulus-driven and stimulus-conditioned fluctuations. To this end, we combine (1) state-of-the-art deep networks for stimulus-driven activity and (2) a flexible, normalizing flow-based generative model to capture th...
[ "mouse visual cortex", "neural system identification", "latent variable models", "normalizing flow", "generative models", "noise correlations" ]
We present a simple-to-train, yet flexible, flow-based generative model of neural population responses that successfully accounts for stimulus-driven responses and noise correlations.
9,524
null
null
h6EWbx5xTj7
Validating the Lottery Ticket Hypothesis with Inertial Manifold Theory
https://openreview.net/forum?id=h6EWbx5xTj7
[ "Zeru Zhang", "Jiayin Jin", "Zijie Zhang", "Yang Zhou", "Xin Zhao", "Jiaxiang Ren", "Ji Liu", "Lingfei Wu", "Ruoming Jin", "Dejing Dou" ]
Poster
null
Despite achieving remarkable efficiency, traditional network pruning techniques often follow manually-crafted heuristics to generate pruned sparse networks. Such heuristic pruning strategies are hard to guarantee that the pruned networks achieve test accuracy comparable to the original dense ones. Recent works have emp...
[ "Lottery Ticket Hypothesis", "neural network pruning", "dynamical systems", "inertial manifold", "theoretical evidence" ]
Theoretically verify the precondition and validity of the Lottery Ticket Hypothesis
9,522
null
null
ohfi44BZPC4
Three-dimensional spike localization and improved motion correction for Neuropixels recordings
https://openreview.net/forum?id=ohfi44BZPC4
[ "Julien Boussard", "Erdem Varol", "Hyun Dong Lee", "Nishchal Dethe", "Liam Paninski" ]
Poster
null
Neuropixels (NP) probes are dense linear multi-electrode arrays that have rapidly become essential tools for studying the electrophysiology of large neural populations. Unfortunately, a number of challenges remain in analyzing the large datasets output by these probes. Here we introduce several new methods for extra...
[ "Spike-sorting", "Neuropixels", "Localization", "Registration" ]
We propose a triangulation method to localize spikes in 3D in Neuropixels recordings, showing that this localization method leads to improved registration and clustering of the waveforms.
9,501
null
null
VjQw3v3FpJx
Robust and Decomposable Average Precision for Image Retrieval
https://openreview.net/forum?id=VjQw3v3FpJx
[ "Elias Ramzi", "Nicolas THOME", "Clément Rambour", "Nicolas Audebert", "Xavier Bitot" ]
Poster
null
In image retrieval, standard evaluation metrics rely on score ranking, e.g. average precision (AP). In this paper, we introduce a method for robust and decomposable average precision (ROADMAP) addressing two major challenges for end-to-end training of deep neural networks with AP: non-differentiability and non-decompos...
[ "Image Retrieval", "Deep Learning", "Average Precision", "Ranking", "Differentiability", "Decomposability" ]
In this paper we address the non-differentiability and non-decomposability of the average precision loss, yielding an efficient end-to-end training of deep models for image retrieval.
9,496
2110.01445
title_snapshot
2a96Bf7Qdrg
Learning One Representation to Optimize All Rewards
https://openreview.net/forum?id=2a96Bf7Qdrg
[ "Ahmed Touati", "Yann Ollivier" ]
Poster
null
We introduce the forward-backward (FB) representation of the dynamics of a reward-free Markov decision process. It provides explicit near-optimal policies for any reward specified a posteriori. During an unsupervised phase, we use reward-free interactions with the environment to learn two representations via off-the-sh...
[ "reward-free MDP", "successor states", "self-supervision", "provable representation learning" ]
We introduce a learnable "summary" of a reward-free MDP, from which provable near-optimal policies can be obtained for any reward function specified a posteriori, instantaneously without planning.
9,491
2103.07945
title_snapshot
d87PBvj7LA7
Understanding the Generalization Benefit of Model Invariance from a Data Perspective
https://openreview.net/forum?id=d87PBvj7LA7
[ "Sicheng Zhu", "Bang An", "Furong Huang" ]
Poster
null
Machine learning models that are developed to be invariant under certain types of data transformations have shown improved generalization in practice. However, a principled understanding of why invariance benefits generalization is limited. Given a dataset, there is often no principled way to select "suitable" data tra...
[ "Data Transformations", "Model Invariance", "Generalization", "Covering Number", "Model Complexity" ]
We understand the generalization benefit of model invariance by refining generalization bounds for invariant models based on sample-dependent properties of data transformations.
9,487
2111.05529
title_snapshot
DDoDN0BLLhb
A/B/n Testing with Control in the Presence of Subpopulations
https://openreview.net/forum?id=DDoDN0BLLhb
[ "Yoan Russac", "Christina Katsimerou", "Dennis Bohle", "Olivier Cappé", "Aurélien Garivier", "Wouter M Koolen" ]
Poster
null
Motivated by A/B/n testing applications, we consider a finite set of distributions (called \emph{arms}), one of which is treated as a \emph{control}. We assume that the population is stratified into homogeneous subpopulations. At every time step, a subpopulation is sampled and an arm is chosen: the resulting observatio...
[ "Pure exploration strategies", "subpopulations", "bandits" ]
null
9,479
2110.15573
title_snapshot
KPLf9FhwEqZ
Model Selection for Bayesian Autoencoders
https://openreview.net/forum?id=KPLf9FhwEqZ
[ "Ba-Hien Tran", "Simone Rossi", "Dimitrios Milios", "Pietro Michiardi", "Edwin V Bonilla", "Maurizio Filippone" ]
Poster
null
We develop a novel method for carrying out model selection for Bayesian autoencoders (BAEs) by means of prior hyper-parameter optimization. Inspired by the common practice of type-II maximum likelihood optimization and its equivalence to Kullback-Leibler divergence minimization, we propose to optimize the distributiona...
[ "Bayesian methods", "Probabilistic models", "Model selection", "Bayesian Autoencoders" ]
We develop a novel method for carrying out model selection for Bayesian autoencoders by means of prior hyper-parameter optimization.
9,476
2106.06245
title_snapshot
_0kknjKJ6BH
End-to-end reconstruction meets data-driven regularization for inverse problems
https://openreview.net/forum?id=_0kknjKJ6BH
[ "Subhadip Mukherjee", "Marcello Carioni", "Ozan Öktem", "Carola-Bibiane Schönlieb" ]
Poster
null
We propose a new approach for learning end-to-end reconstruction operators based on unpaired training data for ill-posed inverse problems. The proposed method combines the classical variational framework with iterative unrolling and essentially seeks to minimize a weighted combination of the expected distortion in the ...
[ "inverse problems", "data-driven regularization", "iterative unrolling", "adversarial learning" ]
This work combines the idea of iterative unrolling with data-driven regularization for solving inverse problems and establishes a link with optimal transport.
9,475
2106.03538
title_snapshot
tvDBe6K8L5o
Invertible Tabular GANs: Killing Two Birds with One Stone for Tabular Data Synthesis
https://openreview.net/forum?id=tvDBe6K8L5o
[ "JAEHOON LEE", "Jihyeon Hyeong", "Jinsung Jeon", "Noseong Park", "Jihoon Cho" ]
Poster
null
Tabular data synthesis has received wide attention in the literature. This is because available data is often limited, incomplete, or cannot be obtained easily, and data privacy is becoming increasingly important. In this work, we present a generalized GAN framework for tabular synthesis, which combines the adversarial...
[ "GAN", "Tabular Data Synthesis", "Fake Data" ]
Synthesize tabular data with GANs
9,447
2202.03636
title_judge
z5-chidgZU3
Risk Monotonicity in Statistical Learning
https://openreview.net/forum?id=z5-chidgZU3
[ "Zakaria Mhammedi" ]
Oral
null
Acquisition of data is a difficult task in many applications of machine learning, and it is only natural that one hopes and expects the population risk to decrease (better performance) monotonically with increasing data points. It turns out, somewhat surprisingly, that this is not the case even for the most standard al...
[ "Statistical Learning", "Risk Monotonicity", "Concentration Inequalities", "PAC-Bayesian Bounds" ]
null
9,442
2011.14126
title_snapshot
TRDAFiwDq8A
MagNet: A Neural Network for Directed Graphs
https://openreview.net/forum?id=TRDAFiwDq8A
[ "Xitong Zhang", "Yixuan He", "Nathan Brugnone", "Michael Perlmutter", "Matthew Hirn" ]
Poster
null
The prevalence of graph-based data has spurred the rapid development of graph neural networks (GNNs) and related machine learning algorithms. Yet, despite the many datasets naturally modeled as directed graphs, including citation, website, and traffic networks, the vast majority of this research focuses on undirected g...
[ "graph neural network", "directed graph", "magnetic Laplacian" ]
We describe how to build graph neural networks for directed graphs using complex Hermitian matrices.
9,426
2102.11391
title_snapshot
_OPHJ7nkZoC
Sample Complexity of Tree Search Configuration: Cutting Planes and Beyond
https://openreview.net/forum?id=_OPHJ7nkZoC
[ "Nina Balcan", "Siddharth Prasad", "Tuomas Sandholm", "Ellen Vitercik" ]
Spotlight
null
Cutting-plane methods have enabled remarkable successes in integer programming over the last few decades. State-of-the-art solvers integrate a myriad of cutting-plane techniques to speed up the underlying tree-search algorithm used to find optimal solutions. In this paper we provide sample complexity bounds for cut-sel...
[ "Automated algorithm configuration", "automated algorithm design", "data-driven algorithm design", "integer programming", "tree search", "branch-and-bound", "branch-and-cut", "cutting planes", "Chvátal-Gomory cuts", "machine learning theory", "sample complexity", "generalization guarantees" ]
null
9,410
2106.04033
title_snapshot
knKJgksd7kA
Interesting Object, Curious Agent: Learning Task-Agnostic Exploration
https://openreview.net/forum?id=knKJgksd7kA
[ "Simone Parisi", "Victoria Dean", "Deepak Pathak", "Abhinav Gupta" ]
Oral
null
Common approaches for task-agnostic exploration learn tabula-rasa --the agent assumes isolated environments and no prior knowledge or experience. However, in the real world, agents learn in many environments and always come with prior experiences as they explore new ones. Exploration is a lifelong process. In this pape...
[ "reinforcement learning", "exploration", "intrinsic motivation", "continual learning" ]
We propose a framework for combining agent-centric and environment-centric exploration. Key points: definition of a new intrinsic reward; exploration policy learning based on these rewards; transfer of the exploration policy to solve new tasks.
9,407
2111.13119
title_snapshot
59mdmZJV6IG
A Multi-Implicit Neural Representation for Fonts
https://openreview.net/forum?id=59mdmZJV6IG
[ "Pradyumna Reddy", "Zhifei Zhang", "Zhaowen Wang", "Matthew Fisher", "Hailin Jin", "Niloy Mitra" ]
Poster
null
Fonts are ubiquitous across documents and come in a variety of styles. They are either represented in a native vector format or rasterized to produce fixed resolution images. In the first case, the non-standard representation prevents benefiting from latest network architectures for neural representations; while, in t...
[ "Font Generation", "Generative Neural Networks", "Implicit Neural Networks" ]
null
9,406
2106.06866
title_snapshot
8SEJ8AT_6Dl
Minimax Optimal Quantile and Semi-Adversarial Regret via Root-Logarithmic Regularizers
https://openreview.net/forum?id=8SEJ8AT_6Dl
[ "Jeffrey Negrea", "Blair Bilodeau", "Nicolò Campolongo", "Francesco Orabona", "Daniel M. Roy" ]
Poster
null
Quantile (and, more generally, KL) regret bounds, such as those achieved by NormalHedge (Chaudhuri, Freund, and Hsu 2009) and its variants, relax the goal of competing against the best individual expert to only competing against a majority of experts on adversarial data. More recently, the semi-adversarial paradigm (Bi...
[ "online learning", "prediction with expert advice", "quantile regret", "semi-adversarial", "adaptive online learning", "FTRL", "regularized learning", "f-divergence", "aggregation", "generalized Bayes", "online variational inference" ]
null
9,404
2110.14804
title_snapshot
luCVRHASXC0
Model, sample, and epoch-wise descents: exact solution of gradient flow in the random feature model
https://openreview.net/forum?id=luCVRHASXC0
[ "Antoine Bodin", "Nicolas Macris" ]
Poster
null
Recent evidence has shown the existence of a so-called double-descent and even triple-descent behavior for the generalization error of deep-learning models. This important phenomenon commonly appears in implemented neural network architectures, and also seems to emerge in epoch-wise curves during the training process. ...
[ "Random Feature Model", "Gradient Flow", "Gradient Descent", "Double Descent", "Epoch-wise Double Descent", "Random Matrix", "Linear Pencil", "Cauchy Integrals", "High-dimensional Limits", "Stieltjes Transform" ]
null
9,402
2110.11805
title_snapshot
Y10GtvGEgR
On Plasticity, Invariance, and Mutually Frozen Weights in Sequential Task Learning
https://openreview.net/forum?id=Y10GtvGEgR
[ "Julian G. Zilly", "Alessandro Achille", "Andrea Censi", "Emilio Frazzoli" ]
Poster
null
Plastic neural networks have the ability to adapt to new tasks. However, in a continual learning setting, the configuration of parameters learned in previous tasks can severely reduce the adaptability to future tasks. In particular, we show that, when using weight decay, weights in successive layers of a deep network m...
[ "Transfer learning", "Deep learning", "Sparsity", "Invariance", "Sequential learning", "Critical learning periods", "Curriculum learning" ]
Training a deep network with weight decay can both improve invariance and generalization but also remove the ability to adapt to new learning tasks due to a type of sparsity we call mutually frozen weights.
9,400
null
null
e_yvNqkJKAW
Test-Time Classifier Adjustment Module for Model-Agnostic Domain Generalization
https://openreview.net/forum?id=e_yvNqkJKAW
[ "Yusuke Iwasawa", "Yutaka Matsuo" ]
Spotlight
null
This paper presents a new algorithm for domain generalization (DG), \textit{test-time template adjuster (T3A)}, aiming to robustify a model to unknown distribution shift. Unlike existing methods that focus on \textit{training phase}, our method focuses \textit{test phase}, i.e., correcting its prediction by itself duri...
[ "domain generalization", "test time adaptation", "prototypical classifier" ]
This paper presents a new algorithm for domain generalization (DG), \textit{test-time template adjuster (T3A)}, which correct its prediction by itself during test time.
9,393
null
null
ZAh31ihNaoF
De-randomizing MCMC dynamics with the diffusion Stein operator
https://openreview.net/forum?id=ZAh31ihNaoF
[ "Zheyang Shen", "Markus Heinonen", "Samuel Kaski" ]
Poster
null
Approximate Bayesian inference estimates descriptors of an intractable target distribution - in essence, an optimization problem within a family of distributions. For example, Langevin dynamics (LD) extracts asymptotically exact samples from a diffusion process because the time evolution of its marginal distributions c...
[ "Stein's method", "MCMC dynamics", "particle variational inference" ]
We develop deterministic particle variants for SGHMC family of diffusion-based MCMC sampling.
9,390
2110.03768
title_snapshot
6ZdqOpE_UVF
A universal probabilistic spike count model reveals ongoing modulation of neural variability
https://openreview.net/forum?id=6ZdqOpE_UVF
[ "David Liu", "Máté Lengyel" ]
Poster
null
Neural responses are variable: even under identical experimental conditions, single neuron and population responses typically differ from trial to trial and across time. Recent work has demonstrated that this variability has predictable structure, can be modulated by sensory input and behaviour, and bears critical sign...
[ "neural coding", "neural variability", "neural correlations", "Gaussian processes", "discrete distributions", "latent variable modelling" ]
We present a probabilistic model for neural spike counts that can capture arbitrary single neuron and joint statistics with their modulation by external covariates.
9,387
null
null
RQJWn82Xga2
Recurrent Bayesian Classifier Chains for Exact Multi-Label Classification
https://openreview.net/forum?id=RQJWn82Xga2
[ "Walter Gerych", "Thomas Hartvigsen", "Luke Buquicchio", "Emmanuel Agu", "Elke Rundensteiner" ]
Poster
null
Exact multi-label classification is the task of assigning each datapoint a set of class labels such that the assigned set exactly matches the ground truth. Optimizing for exact multi-label classification is important in domains where missing a single label can be especially costly, such as in object detection for auton...
[ "multilabel classification", "recurrent classifier chains", "classifier chains", "recurrent networks", "exact classification", "subset accuracy" ]
An improvement over Recurrent Classifier Chains for multi-label classification that incorporates Bayesian networks of class inter-dependencies.
9,386
null
null
FwVmM8Zol_8
Vector-valued Gaussian Processes on Riemannian Manifolds via Gauge Independent Projected Kernels
https://openreview.net/forum?id=FwVmM8Zol_8
[ "Michael John Hutchinson", "Alexander Terenin", "Viacheslav Borovitskiy", "So Takao", "Yee Whye Teh", "Marc Peter Deisenroth" ]
Poster
null
Gaussian processes are machine learning models capable of learning unknown functions in a way that represents uncertainty, thereby facilitating construction of optimal decision-making systems. Motivated by a desire to deploy Gaussian processes in novel areas of science, a rapidly-growing line of research has focused on...
[ "Gaussian processes", "Riemannian manifolds", "gauge independence" ]
Vector-valued Gaussian Processes on Riemannian Manifolds via Gauge Independant Projected Kernels
9,383
2110.14423
title_snapshot
ALvt7nXa2q
Overcoming Catastrophic Forgetting in Incremental Few-Shot Learning by Finding Flat Minima
https://openreview.net/forum?id=ALvt7nXa2q
[ "Guangyuan SHI", "Jiaxin Chen", "Wenlong Zhang", "Li-Ming Zhan", "Xiao-Ming Wu" ]
Spotlight
null
This paper considers incremental few-shot learning, which requires a model to continually recognize new categories with only a few examples provided. Our study shows that existing methods severely suffer from catastrophic forgetting, a well-known problem in incremental learning, which is aggravated due to data scarcity...
[ "incremental few-shot learning", "catastrophic forgetting", "flat minima" ]
We propose to overcome catastrophic forgetting in incremental few-shot learning by finding flat minima in the base training stage.
9,373
2111.01549
title_snapshot
OUH25e12YyH
Differentiable Unsupervised Feature Selection based on a Gated Laplacian
https://openreview.net/forum?id=OUH25e12YyH
[ "Ofir Lindenbaum", "Uri Shaham", "Erez Peterfreund", "Jonathan Svirsky", "Nicolas Aidan Casey", "Yuval Kluger" ]
Poster
null
Scientific observations may consist of a large number of variables (features). Selecting a subset of meaningful features is often crucial for identifying patterns hidden in the ambient space. In this paper, we present a method for unsupervised feature selection, and we demonstrate its advantage in clustering, a common ...
[ "Feature Selection", "Graph Laplacian", "Laplacian Score" ]
We present a method for unsupervised feature selection based on a gated Laplacian
9,361
2007.04728
title_snapshot