ICML 2021

1,200 papers

Top Research Directions

Research directions most represented by ICML 2021 papers.

The Non-Linear Function Approximation Gap in Reinforcement Learning 26
Scaling Reinforcement Learning to High-Dimensional Visual Observations 19
Scaling Gaussian Processes Beyond Cubic Complexity 17
Clean Accuracy Degradation in Machine Learning Defenses 16
Bypassing the Computational Bottleneck of Adversarial Training 16
The Function Approximation Gap in Reinforcement Learning Theory 14
Exploration Strategies for Sparse Reward Reinforcement Learning 14
Horizon Free Bounds in Reinforcement Learning 13
Learning with Instance Dependent Label Noise 11
Variance Adaptive Confidence Bounds 11
Optimistic Exploration for Efficient Reinforcement Learning 11
Optimization and Generalization with Non-Smooth Activations 11
Accelerating Neural Architecture Search Without Full Model Training 11
Linearizing Deep Learning via the Neural Tangent Kernel 10
The Computational Overhead of Iterative Model Pruning 10

Top Authors

AuthorPapers
Masashi Sugiyama14
Sergey Levine12
Gang Niu11
Zhaoran Wang11
Zhuoran Yang11
Quanquan Gu8
Andreas Krause7
Bo Han7
Chelsea Finn7
Percy Liang7
Bernhard Schölkopf6
Csaba Szepesvari6
David Woodruff6
Michal Valko6
Pieter Abbeel6
Tongliang Liu6
Alex Dimakis5
Andrew Gordon Wilson5
Bo Li5
Gal Chechik5

Impactful Papers

Papers from ICML 2021 that shaped research clusters.

Foundational

Influential

Papers

Title
PODS: Policy Optimization via Differentiable Simulation
Pointwise Binary Classification with Pairwise Confidence Comparisons
Poisson-Randomised DirBN: Large Mutation is Needed in Dirichlet Belief Networks
Policy Analysis using Synthetic Controls in Continuous-Time
Policy Caches with Successor Features
Policy Gradient Bayesian Robust Optimization for Imitation Learning
Policy Information Capacity: Information-Theoretic Measure for Task Complexity in Deep Reinforcement Learning
Poolingformer: Long Document Modeling with Pooling Attention
PopSkipJump: Decision-Based Attack for Probabilistic Classifiers
Positive-Negative Momentum: Manipulating Stochastic Gradient Noise to Improve Generalization
Posterior Value Functions: Hindsight Baselines for Policy Gradient Methods
Post-selection inference with HSIC-Lasso
Practical and Private (Deep) Learning Without Sampling or Shuffling
Prediction-Centric Learning of Independent Cascade Dynamics from Partial Observations
Predict then Interpolate: A Simple Algorithm to Learn Stable Classifiers
Preferential Temporal Difference Learning
Principal Bit Analysis: Autoencoding with Schur-Concave Loss
Principal Component Hierarchy for Sparse Quadratic Programs
Principled Exploration via Optimistic Bootstrapping and Backward Induction
Principled Simplicial Neural Networks for Trajectory Prediction
Prior Image-Constrained Reconstruction using Style-Based Generative Models
Prioritized Level Replay
Privacy-Preserving Feature Selection with Secure Multiparty Computation
Privacy-Preserving Video Classification with Convolutional Neural Networks
Private Adaptive Gradient Methods for Convex Optimization
Private Alternating Least Squares: Practical Private Matrix Completion with Tighter Rates
Private Stochastic Convex Optimization: Optimal Rates in L1 Geometry
Probabilistic Programs with Stochastic Conditioning
Probabilistic Sequential Shrinking: A Best Arm Identification Algorithm for Stochastic Bandits with Corruptions
Problem Dependent View on Structured Thresholding Bandit Problems
Proceedings of the 38th International Conference on Machine Learning, ICML 2021, 18-24 July 2021, Virtual Event
ProGraML: A Graph-based Program Representation for Data Flow Analysis and Compiler Optimizations
Progressive-Scale Boundary Blackbox Attack via Projective Gradient Estimation
Projection Robust Wasserstein Barycenters
Projection techniques to update the truncated SVD of evolving matrices with applications
Provable Generalization of SGD-trained Neural Networks of Any Width in the Presence of Adversarial Label Noise
Provable Lipschitz Certification for Generative Models
Provable Meta-Learning of Linear Representations
Provable Robustness of Adversarial Training for Learning Halfspaces with Noise
Provably Correct Optimization and Exploration with Non-linear Policies
Provably Efficient Algorithms for Multi-Objective Competitive RL
Provably Efficient Fictitious Play Policy Optimization for Zero-Sum Markov Games with Structured Transitions
Provably Efficient Learning of Transferable Rewards
Provably Efficient Reinforcement Learning for Discounted MDPs with Feature Mapping
Provably End-to-end Label-noise Learning without Anchor Points
Provably Strict Generalisation Benefit for Equivariant Models
Proximal Causal Learning with Kernels: Two-Stage Estimation and Moment Restriction
PsiPhi-Learning: Reinforcement Learning with Demonstrations using Successor Features and Inverse Temporal Difference Learning
Pure Exploration and Regret Minimization in Matching Bandits
Putting the "Learning" into Learning-Augmented Algorithms for Frequency Estimation

« prev Page 18/24 next »