ICLR 2022 accepted papers on Learning theory and algorithms
93 of the 1,095 papers accepted at ICLR 2022 (8.5%) are on Learning theory and algorithms, compared with 9.7% at ICLR 2021. They fall into 9 topics.
See ICLR 2022 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
Kernels and regression theory · 22 papers
- Neural Collapse Under MSE Loss: Proximity to and Dynamics on the Central Path (19 citations)
- PAC-Bayes Information Bottleneck (11 citations)
- Machine Learning For Elliptic PDEs: Fast Rate Generalization Bound, Neural Scaling Law and Minimax Optimality (10 citations)
- Approximation and Learning with Deep Convolutional Models: a Kernel Perspective (5 citations)
- Generalized Kernel Thinning (5 citations)
- The Role of Permutation Invariance in Linear Mode Connectivity of Neural Networks (3 citations)
- Generalization Through the Lens of Leave-One-Out Error (2 citations)
- How many degrees of freedom do we need to train deep networks: a loss landscape perspective (2 citations)
- On the Generalization of Models Trained with SGD: Information-Theoretic Bounds and Implications (2 citations)
- A Generalized Weighted Optimization Method for Computational Learning and Inversion (1 citation)
- Learning Curves for Gaussian Process Regression with Power-Law Priors and Targets (1 citation)
- Neural Networks as Kernel Learners: The Silent Alignment Effect (1 citation)
and 10 more in the interactive map.
Prediction and decision losses · 18 papers
- Critical Points in Quantum Generative Models (11 citations)
- Learning Optimal Conformal Classifiers (11 citations)
- How Did the Model Change? Efficiently Assessing Machine Learning API Shifts (8 citations)
- Bridging the Gap: Providing Post-Hoc Symbolic Explanations for Sequential Decision-Making Problems with Inscrutable Representations (7 citations)
- Sound and Complete Neural Network Repair with Minimality and Locality Guarantees (7 citations)
- Boosting the Certified Robustness of L-infinity Distance Nets (6 citations)
- GradSign: Model Performance Inference with Theoretical Insights (5 citations)
- PAC Prediction Sets Under Covariate Shift (5 citations)
- Focus on the Common Good: Group Distributional Robustness Follows (3 citations)
- Efficient and Differentiable Conformal Prediction with General Function Classes (1 citation)
- A Neural Tangent Kernel Perspective of Infinite Tree Ensembles
- Axiomatic Explanations for Visual Search, Retrieval, and Similarity Learning
and 6 more in the interactive map.
Combinatorial and search optimization · 14 papers
- Efficient Active Search for Combinatorial Optimization Problems (27 citations)
- Complete Verification via Multi-Neuron Relaxation Guided Branch-and-Bound (23 citations)
- Pareto Set Learning for Neural Multi-Objective Combinatorial Optimization (22 citations)
- Neural Stochastic Dual Dynamic Programming (9 citations)
- CrossBeam: Learning to Search in Bottom-Up Program Synthesis (6 citations)
- Maximum n-times Coverage for Vaccine Design (3 citations)
- Multi-objective Optimization by Learning Space Partition (3 citations)
- Safe Neurosymbolic Learning with Differentiable Symbolic Execution (3 citations)
- Universal Approximation Under Constraints is Possible with Transformers (2 citations)
- Constraining Linear-chain CRFs to Regular Languages (1 citation)
- Low-Budget Active Learning via Wasserstein Distance: An Integer Programming Approach (1 citation)
- Enabling Arbitrary Translation Objectives with Adaptive Tree Search
and 2 more in the interactive map.
Sample complexity · 13 papers
- Sqrt(d) Dimension Dependence of Langevin Monte Carlo (4 citations)
- Towards Empirical Sandwich Bounds on the Rate-Distortion Function (4 citations)
- On the Optimal Memorization Power of ReLU Neural Networks (3 citations)
- A Theory of Tournament Representations (2 citations)
- Online Facility Location with Predictions (2 citations)
- PER-ETD: A Polynomially Efficient Emphatic Temporal Difference Learning Method (2 citations)
- Minimax Optimality (Probably) Doesn't Imply Distribution Learning for GANs (1 citation)
- Triangle and Four Cycle Counting with Predictions in Graph Streams (1 citation)
- Distribution Compression in Near-Linear Time
- Certified Robustness for Deep Equilibrium Models via Interval Bound Propagation
- Learnability of convolutional neural networks for infinite dimensional input via mixed and anisotropic smoothness
- Neural Network Approximation based on Hausdorff distance of Tropical Zonotopes
and 1 more in the interactive map.
Matrix and tensor methods · 11 papers
- Gradient Step Denoiser for convergent Plug-and-Play (18 citations)
- A generalization of the randomized singular value decomposition (6 citations)
- Generative Principal Component Analysis (6 citations)
- Fast Regression for Structured Inputs (1 citation)
- Learning Representation from Neural Fisher Kernel with Low-rank Approximation (1 citation)
- A Comparison of Hamming Errors of Representative Variable Selection Methods
- Implicit Bias of Projected Subgradient Method Gives Provable Robust Recovery of Subspaces of Unknown Codimension
- Scalable Sampling for Nonsymmetric Determinantal Point Processes
- Unrolling PALM for Sparse Semi-Blind Source Separation
- Multi-Mode Deep Matrix and Tensor Factorization
- Provable Learning-based Algorithm For Sparse Recovery
Optimal transport · 7 papers
- Augmented Sliced Wasserstein Distances (4 citations)
- Fast topological clustering with Wasserstein distance (3 citations)
- A fast and accurate splitting method for optimal transport: analysis and implementation (1 citation)
- Learning meta-features for AutoML (1 citation)
- Monotonic Differentiable Sorting Networks
- LOSSY COMPRESSION WITH DISTRIBUTION SHIFT AS ENTROPY CONSTRAINED OPTIMAL TRANSPORT
- Optimal Transport for Long-Tailed Recognition with Learnable Cost Matrix
