Generalization

Benign Overfitting without Linearity: Neural Network Classifiers Trained by Gradient Descent for Noisy Linear Data COLT 2022 Eigenspace Restructuring: A Principle of Space and Frequency in Neural Networks COLT 2022 Evading the Simplicity Bias: Training a Diverse Set of Models Discovers Solutions With Superior OOD Generalization CVPR 2022 Exploiting Explainable Metrics for Augmented SGD CVPR 2022 Large Learning Rate Tames Homogeneity: Convergence and Balancing Effect ICLR 2022 Layer-wise Adaptive Graph Convolution Networks Using Generalized Pagerank ACML 2022 Optimization and Adaptive Generalization of Three layer Neural Networks ICLR 2022 Rate of Convergence of Polynomial Networks to Gaussian Processes COLT 2022 Reflash Dropout in Image Super-Resolution CVPR 2022 Scaling Laws for Neural Machine Translation ICLR 2022 Surrogate Gap Minimization Improves Sharpness-Aware Training ICLR 2022 What Makes Instruction Learning Hard? An Investigation and a New Challenge in a Synthetic Environment EMNLP 2022 A Diffusion Theory For Deep Learning Dynamics: Stochastic Gradient Descent Exponentially Favors Flat Minima ICLR 2021 A Law of Robustness for Two-Layers Neural Networks COLT 2021 A Priori Generalization Analysis of the Deep Ritz Method for Solving High Dimensional Elliptic Partial Differential Equations COLT 2021 Better Fine-Tuning by Reducing Representational Collapse ICLR 2021 Ensembling With a Fixed Parameter Budget: When Does It Help and Why? ACML 2021 Generalization in NLI: Ways (Not) To Go Beyond Simple Heuristics EMNLP 2021 HiddenCut: Simple Data Augmentation for Natural Language Understanding with Better Generalizability ACL 2021 Learning to Recombine and Resample Data For Compositional Generalization ICLR 2021 Norm-Based Generalisation Bounds for Deep Multi-Class Convolutional Neural Networks AAAI 2021 Overparameterisation and worst-case generalisation: friend or foe? ICLR 2021 Transient Non-stationarity and Generalisation in Deep Reinforcement Learning ICLR 2021 Computing the Testing Error Without a Testing Set CVPR 2020 High-Frequency Component Helps Explain the Generalization of Convolutional Neural Networks CVPR 2020