ICLR 2020 accepted papers on Learning theory and algorithms
54 of the 687 papers accepted at ICLR 2020 (7.9%) are on Learning theory and algorithms, compared with 7.4% at ICLR 2019. They fall into 9 topics.
See ICLR 2020 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 · 14 papers
- Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds (274 citations)
- Distributionally Robust Neural Networks (107 citations)
- Empirical Studies on the Properties of Linear Regions in Deep Neural Networks (26 citations)
- Towards Hierarchical Importance Attribution: Explaining Compositional Semantics for Neural Sequence Models (25 citations)
- Learning Space Partitions for Nearest Neighbor Search (24 citations)
- PAC Confidence Sets for Deep Neural Networks via Calibrated Prediction (21 citations)
- NeurQuRI: Neural Question Requirement Inspector for Answerability Prediction in Machine Reading Comprehension (19 citations)
- Phase Transitions for the Information Bottleneck in Representation Learning (19 citations)
- Pure and Spurious Critical Points: a Geometric Study of Linear Networks (9 citations)
- Oblique Decision Trees from Derivatives of ReLU Networks (8 citations)
- Extreme Classification via Adversarial Softmax Approximation (7 citations)
- An Inductive Bias for Distances: Neural Nets that Respect the Triangle Inequality (4 citations)
and 2 more in the interactive map.
Kernels and regression theory · 12 papers
- Fantastic Generalization Measures and Where to Find Them (147 citations)
- A Constructive Prediction of the Generalization Error Across Scales (70 citations)
- Ridge Regression: Structure, Cross-Validation, and Sketching (23 citations)
- Compression based bound for non-compressed network: unified generalization error analysis of large compressible deep neural network (22 citations)
- Finite Depth and Width Corrections to the Neural Tangent Kernel (21 citations)
- Understanding Generalization in Recurrent Neural Networks (17 citations)
- Truth or backpropaganda? An empirical investigation of deep learning theory (16 citations)
- Harnessing the Power of Infinitely Wide Deep Nets on Small-data Tasks (13 citations)
- Neural tangent kernels, transportation mappings, and universal approximation (10 citations)
- Beyond Linearization: On Quadratic and Higher-Order Approximation of Wide Neural Networks (9 citations)
- Improved Sample Complexities for Deep Neural Networks and Robust Classification via an All-Layer Margin (9 citations)
- Kernel of CycleGAN as a principal homogeneous space (1 citation)
Sample complexity · 9 papers
- Maxmin Q-learning: Controlling the Estimation Bias of Q-learning (36 citations)
- Lipschitz constant estimation of Neural Networks via sparse polynomial optimization (25 citations)
- Effect of Activation Functions on the Training of Overparametrized Neural Nets (12 citations)
- Learning-Augmented Data Stream Algorithms (9 citations)
- A closer look at the approximation capabilities of neural networks (8 citations)
- Geometric Insights into the Convergence of Nonlinear TD Learning (8 citations)
- On Universal Equivariant Set Networks (5 citations)
- Learning to Guide Random Search (4 citations)
- Stable Rank Normalization for Improved Generalization in Neural Networks and GANs (2 citations)
Combinatorial and search optimization · 7 papers
- A Learning-based Iterative Method for Solving Vehicle Routing Problems (123 citations)
- Augmenting Genetic Algorithms with Deep Neural Networks for Exploring the Chemical Space (111 citations)
- Differentiation of Blackbox Combinatorial Solvers (52 citations)
- Chameleon: Adaptive Code Optimization for Expedited Deep Neural Network Compilation (17 citations)
- Learning to Represent Programs with Property Signatures (17 citations)
- Infinite-Horizon Differentiable Model Predictive Control (5 citations)
- Watch the Unobserved: A Simple Approach to Parallelizing Monte Carlo Tree Search (4 citations)
Matrix and tensor methods · 5 papers
- Tensor Decompositions for Temporal Knowledge Base Completion (131 citations)
- Sparse Coding with Gated Learned ISTA (19 citations)
- Understanding l4-based Dictionary Learning: Interpretation, Stability, and Robustness (5 citations)
- Short and Sparse Deconvolution --- A Geometric Approach (3 citations)
- Span Recovery for Deep Neural Networks with Applications to Input Obfuscation
Fairness · 3 papers
- Training individually fair ML models with sensitive subspace robustness (51 citations)
- Conditional Learning of Fair Representations (17 citations)
- Rényi Fair Inference (14 citations)
Clustering · 2 papers
- Learning to Link (9 citations)
- Spectral Embedding of Regularized Block Models (2 citations)
Submodular and game algorithms · 1 paper
- A Generalized Training Approach for Multiagent Learning (22 citations)
Optimal transport · 1 paper
- Learning transport cost from subset correspondence (4 citations)
