ICLR 2022

2,337 papers

Top Research Directions

Research directions most represented by ICLR 2022 papers.

Bypassing the Computational Bottleneck of Adversarial Training 21
Sample Efficient Reinforcement Learning 19
The Mechanics of Benign Overfitting in High Dimensions 18
Clean Accuracy Degradation in Machine Learning Defenses 17
Overcoming the Homophily Assumption in Graph Neural Networks 13
Accelerating Sparse Neural Networks on Dense Hardware 13
Accelerating Neural Architecture Search Without Full Model Training 12
The Non-Linear Function Approximation Gap in Reinforcement Learning 12
The Mechanics of Backdoor Attacks in Neural Networks 12
Beyond the Lazy Training Regime of Neural Networks 11
The Infinite Width Approximation Gap in Deep Learning 10
Matrix-Free Curvature Estimation for Deep Learning 10
The Computational Overhead of Iterative Model Pruning 9
The Weisfeiler Lehman Expressivity Bottleneck 9
Why Theoretical FLOPs Fail to Predict Real Latency 9

Top Authors

AuthorPapers
Zhangyang Wang23
Sergey Levine21
Tianlong Chen15
Tongliang Liu14
Bo Han13
Chelsea Finn13
Gang Niu13
Jun Wang13
Yoshua Bengio13
Bernhard Schölkopf12
Joshua B. Tenenbaum12
J Zico Kolter12
Ping Li12
Tie-Yan Liu12
Dacheng Tao11
Stephan Günnemann11
Tom Goldstein11
Masashi Sugiyama10
Pieter Abbeel9
Pin-Yu Chen9

Impactful Papers

Papers from ICLR 2022 that shaped research clusters.

Foundational

Influential

Papers

Title
Limitations of Active Learning With Deep Transformer Language Models
Linear Convergence of SGD on Overparametrized Shallow Neural Networks
Linking Emergent and Natural Languages via Corpus Transfer
Lipschitz-constrained Unsupervised Skill Discovery
Local Feature Swapping for Generalization in Reinforcement Learning
Locality-Based Mini Batching for Graph Neural Networks
Localized Persistent Homologies for more Effective Deep Learning
Localized Randomized Smoothing for Collective Robustness Certification
Locally Invariant Explanations: Towards Causal Explanations through Local Invariant Learning
Local Patch AutoAugment with Multi-Agent Collaboration
Local Permutation Equivariance For Graph Neural Networks
Local Reweighting for Adversarial Training
Logarithmic landscape and power-law escape rate of SGD
Logarithmic Unbiased Quantization: Practical 4-bit Training in Deep Learning
Logical Activation Functions: Logit-space equivalents of Boolean Operators
Logit Attenuating Weight Normalization
Log-Polar Space Convolution
Long Document Summarization with Top-Down and Bottom-Up Representation Inference
Long Expressive Memory for Sequence Modeling
Looking Back on Learned Experiences For Class/task Incremental Learning
LoRA: Low-Rank Adaptation of Large Language Models
LORD: Lower-Dimensional Embedding of Log-Signature in Neural Rough Differential Equations
Losing Less: A Loss for Differentially Private Deep Learning
Lossless Compression with Probabilistic Circuits
Lossy Compression with Distribution Shift as Entropy Constrained Optimal Transport
Lossy Compression with Gaussian Diffusion
Lottery Image Prior
Lottery Tickets can have Structural Sparsity
Low-Budget Active Learning via Wasserstein Distance: An Integer Programming Approach
Low-Cost Algorithmic Recourse for Users With Uncertain Cost Functions
Low Entropy Deep Networks
Lower Bounds on the Robustness of Fixed Feature Extractors to Test-time Adversaries
Low-Precision Stochastic Gradient Langevin Dynamics
Low-rank Matrix Recovery with Unknown Correspondence
LSP : Acceleration and Regularization of Graph Neural Networks via Locality Sensitive Pruning of Graphs
M6-10T: A Sharing-Delinking Paradigm for Efficient Multi-Trillion Parameter Pretraining
MA2QL: A Minimalist Approach to Fully Decentralized Multi-Agent Reinforcement Learning
Machine Learning For Elliptic PDEs: Fast Rate Generalization Bound, Neural Scaling Law and Minimax Optimality
MaGNET: Uniform Sampling from Deep Generative Network Manifolds Without Retraining
Mako: Semi-supervised continual learning with minimal labeled data via data programming
MAML is a Noisy Contrastive Learner in Classification
Manifold Micro-Surgery with Linearly Nearly Euclidean Metrics
Map Induction: Compositional spatial submap learning for efficient exploration in novel environments
Mapping conditional distributions for domain adaptation under generalized target shift
Mapping Language Models to Grounded Conceptual Spaces
Marginal Tail-Adaptive Normalizing Flows
Mastering Spatial Graph Prediction of Road Networks
Mastering Visual Continuous Control: Improved Data-Augmented Reinforcement Learning
Match Prediction Using Learned History Embeddings
Max-Affine Spline Insights Into Deep Network Pruning

« prev Page 26/47 next »