ICLR 2021 accepted papers on Learning theory and algorithms
83 of the 860 papers accepted at ICLR 2021 (9.7%) are on Learning theory and algorithms, compared with 7.9% at ICLR 2020. They fall into 9 topics.
See ICLR 2021 in the interactive map
Submitting to ICLR? Checklist →
Working on something in this area? Describe your idea in scime atlas to see the closest papers by meaning and how crowded the spot has become.
Topics and papers
Prediction and decision losses · 23 papers
- The Risks of Invariant Risk Minimization (88 citations)
- Learning explanations that are hard to vary (69 citations)
- Are Neural Rankers still Outperformed by Gradient Boosted Decision Trees? (41 citations)
- Uncertainty Sets for Image Classifiers using Conformal Prediction (31 citations)
- Randomized Ensembled Double Q-Learning: Learning Fast Without a Model (25 citations)
- Co-Mixup: Saliency Guided Joint Mixup with Supermodular Diversity (16 citations)
- Influence Functions in Deep Learning Are Fragile (14 citations)
- Shapley explainability on the data manifold (12 citations)
- Interpreting and Boosting Dropout from a Game-Theoretic View (7 citations)
- When Do Curricula Work? (7 citations)
- PAC Confidence Predictions for Deep Neural Network Classifiers (6 citations)
- A Learning Theoretic Perspective on Local Explainability (5 citations)
and 11 more in the interactive map.
Kernels and regression theory · 14 papers
- In Search of Lost Domain Generalization (79 citations)
- Optimal Regularization can Mitigate Double Descent (46 citations)
- A Wigner-Eckart Theorem for Group Equivariant Convolution Kernels (20 citations)
- Generalization bounds via distillation (13 citations)
- Effective Distributed Learning with Random Features: Improved Bounds and Algorithms (10 citations)
- Deep Equals Shallow for ReLU Networks in Kernel Regimes (9 citations)
- Deep Networks and the Multiple Manifold Problem (7 citations)
- Dataset Meta-Learning from Kernel Ridge-Regression (6 citations)
- Estimating Lipschitz constants of monotone deep equilibrium models (6 citations)
- On the Universality of the Double Descent Peak in Ridgeless Regression (5 citations)
- Meta-Learning with Neural Tangent Kernels (4 citations)
- Deep Neural Tangent Kernel and Laplace Kernel Have the Same RKHS (3 citations)
and 2 more in the interactive map.
Sample complexity · 13 papers
- Few-Shot Learning via Learning the Representation, Provably (53 citations)
- Why Are Convolutional Nets More Sample-Efficient than Fully-Connected Nets? (20 citations)
- Universal approximation power of deep residual neural networks via nonlinear control theory (19 citations)
- Empirical or Invariant Risk Minimization? A Sample Complexity Perspective (15 citations)
- Learning-based Support Estimation in Sublinear Time (7 citations)
- New Bounds For Distributed Mean Estimation and Variance Reduction (4 citations)
- Efficient Empowerment Estimation for Unsupervised Stabilization (2 citations)
- Estimating informativeness of samples with Smooth Unique Information (2 citations)
- Theoretical bounds on estimation error for meta-learning (2 citations)
- Improved Estimation of Concentration Under $\ell_p$-Norm Distance Metrics Using Half Spaces
- Robust Learning of Fixed-Structure Bayesian Networks in Nearly-Linear Time
- Sequential Density Ratio Estimation for Simultaneous Optimization of Speed and Accuracy
and 1 more in the interactive map.
Combinatorial and search optimization · 11 papers
- DC3: A learning method for optimization with hard constraints (68 citations)
- Fast and Complete: Enabling Complete Neural Network Verification with Rapid and Massively Parallel Incomplete Verifiers (65 citations)
- Deep symbolic regression: Recovering mathematical expressions from data via risk-seeking policy gradients (49 citations)
- Learning the Pareto Front with Hypernetworks (38 citations)
- Practical Massively Parallel Monte-Carlo Tree Search Applied to Molecular Design (9 citations)
- Representing Partial Programs with Blended Abstract Semantics (8 citations)
- Scaling the Convex Barrier with Active Sets (8 citations)
- Neural Learning of One-of-Many Solutions for Combinatorial Problems in Structured Output Spaces (4 citations)
- Teaching Temporal Logics to Neural Networks (4 citations)
- Molecule Optimization by Explainable Evolution (3 citations)
- Fidelity-based Deep Adiabatic Scheduling
Matrix and tensor methods · 6 papers
- Deciphering and Optimizing Multi-Task Learning: a Random Matrix Approach (2 citations)
- Faster Binary Embeddings for Preserving Euclidean Distances (1 citation)
- Learning a Latent Simplex in Input Sparsity Time (1 citation)
- On InstaHide, Phase Retrieval, and Sparse Matrix Factorization (1 citation)
- Local Search Algorithms for Rank-Constrained Convex Optimization
- Mixed-Features Vectors and Subspace Splitting
Fairness · 5 papers
- Tilted Empirical Risk Minimization (38 citations)
- FairBatch: Batch Selection for Model Fairness (37 citations)
- SenSeI: Sensitive Set Invariance for Enforcing Individual Fairness (18 citations)
- Fair Mixup: Fairness via Interpolation (10 citations)
- Individually Fair Rankings (5 citations)
Optimal transport · 4 papers
- Wasserstein-2 Generative Networks (17 citations)
- Continuous Wasserstein-2 Barycenter Estimation without Minimax Optimization (9 citations)
- Distributional Sliced-Wasserstein and Applications to Generative Modeling (9 citations)
- Neural Topic Model via Optimal Transport (5 citations)
Submodular and game algorithms · 4 papers
- EigenGame: PCA as a Nash Equilibrium (6 citations)
- Learning to Make Decisions via Submodular Regularization (5 citations)
- Auction Learning as a Two-Player Game (2 citations)
- Average-case Acceleration for Bilinear Games and Normal Matrices (1 citation)
