ICML 2022

1,154 papers

Top Research Directions

Research directions most represented by ICML 2022 papers.

The Non-Linear Function Approximation Gap in Reinforcement Learning 23
Clean Accuracy Degradation in Machine Learning Defenses 22
Scaling Gaussian Processes Beyond Cubic Complexity 17
Provably Efficient Reinforcement Learning with Linear MDPs 15
The Function Approximation Gap in Reinforcement Learning Theory 15
Bypassing the Computational Bottleneck of Adversarial Training 14
Loss Landscape Flatness and the Generalization Gap 13
Mitigating Gradient Variance in Policy Optimization 13
Distribution Shift in Offline Reinforcement Learning 12
Scaling Bandit Learning to Large Action Spaces 10
Multi-Agent Learning Under Bandit Feedback 10
Horizon Free Bounds in Reinforcement Learning 10
Periodic Synchronization for Communication Efficient Distributed Learning 10
Optimistic Exploration for Efficient Reinforcement Learning 10
Matrix-Free Curvature Estimation for Deep Learning 10

Top Authors

AuthorPapers
Zhangyang Wang9
Zhaoran Wang9
Zhuoran Yang9
Bo Li7
Jun Zhu7
Shandian Zhe7
Tianbao Yang7
Bernhard Schölkopf6
Bo Han6
Manzil Zaheer6
Mihaela van der Schaar6
Nhat Ho6
Pranjal Awasthi6
Sergey Levine6
Tianlong Chen6
Yang Liu6
Andreas Krause5
David Woodruff5
Michael W. Mahoney5
Peter Richtárik5

Impactful Papers

Papers from ICML 2022 that shaped research clusters.

Foundational

Influential

Papers

Title
Online Learning with Knapsacks: the Best of Both Worlds
Online Nonsubmodular Minimization with Delayed Costs: From Full Information to Bandit Feedback
Only tails matter: Average-Case Universality and Robustness in the Convex Regime
On Measuring Causal Contributions via do-interventions
On Non-local Convergence Analysis of Deep Linear Networks
On Numerical Integration in Neural Ordinary Differential Equations
On the Adversarial Robustness of Causal Algorithmic Recourse
On the Convergence of Inexact Predictor-Corrector Methods for Linear Programming
On the Convergence of Local Stochastic Compositional Gradient Descent with Momentum
On the Convergence of the Shapley Value in Parametric Bayesian Learning Games
On the Difficulty of Defending Self-Supervised Learning against Model Extraction
On the Effects of Artificial Data Modification
On the Equivalence Between Temporal and Static Equivariant Graph Representations
On the Finite-Time Complexity and Practical Computation of Approximate Stationarity Concepts of Lipschitz Functions
On the Finite-Time Performance of the Knowledge Gradient Algorithm
On the Generalization Analysis of Adversarial Learning
On the Hidden Biases of Policy Mirror Ascent in Continuous Action Spaces
On the Impossibility of Learning to Cooperate with Adaptive Partner Strategies in Repeated Games
On the Learning of Non-Autoregressive Transformers
On the Optimization Landscape of Neural Collapse under MSE Loss: Global Optimality with Unconstrained Features
On the Practicality of Deterministic Epistemic Uncertainty
On the Robustness of CountSketch to Adaptive Inputs
On the Role of Discount Factor in Offline Reinforcement Learning
On the Sample Complexity of Learning Infinite-horizon Discounted Linear Kernel MDPs
On the Statistical Benefits of Curriculum Learning
On the Subspace Structure of Gradient-Based Meta-Learning
On the Surrogate Gap between Contrastive and Supervised Losses
On Transportation of Mini-batches: A Hierarchical Approach
On Well-posedness and Minimax Optimal Rates of Nonparametric Q-function Estimation in Off-policy Evaluation
Open-Sampling: Exploring Out-of-Distribution data for Re-balancing Long-tailed datasets
Optimal Algorithms for Mean Estimation under Local Differential Privacy
Optimal Algorithms for Stochastic Multi-Level Compositional Optimization
Optimal and Efficient Dynamic Regret Algorithms for Non-Stationary Dueling Bandits
Optimal Clipping and Magnitude-aware Differentiation for Improved Quantization-aware Training
Optimal Clustering with Noisy Queries via Multi-Armed Bandit
Optimal Estimation of Policy Gradient via Double Fitted Iteration
Optimally Controllable Perceptual Lossy Compression
Optimistic Linear Support and Successor Features as a Basis for Optimal Policy Transfer
Optimization-Derived Learning with Essential Convergence Analysis of Training and Hyper-training
Optimization-Induced Graph Implicit Nonlinear Diffusion
Optimizing Sequential Experimental Design with Deep Reinforcement Learning
Optimizing Tensor Network Contraction Using Reinforcement Learning
Orchestra: Unsupervised Federated Learning via Globally Consistent Clustering
Order Constraints in Optimal Transport
Out-of-Distribution Detection with Deep Nearest Neighbors
Overcoming Oscillations in Quantization-Aware Training
PAC-Bayesian Bounds on Rate-Efficient Classifiers
PACE: A Parallelizable Computation Encoder for Directed Acyclic Graphs
PAGE-PG: A Simple and Loopless Variance-Reduced Policy Gradient Method with Probabilistic Gradient Estimation
Pairwise Conditional Gradients without Swap Steps and Sparser Kernel Herding

« prev Page 16/24 next »