Gentle Local Robustness implies Generalization

Abstract

Robustness and generalization ability of machine learning models are of utmost importance in various application domains. There is a wide interest in efficient ways to analyze those properties. One important direction is to analyze connection between those two properties. Prior theories suggest that a robust learning algorithm can produce trained models with a high generalization ability. However, we show in this work that the existing error bounds are vacuous for the Bayes optimal classifier which is the best among all measurable classifiers for a classification problem with overlapping classes. Those bounds cannot converge to the true error of this ideal classifier. This is undesirable, surprizing, and never known before. We then present a class of novel bounds, which are model-dependent and provably tighter than the existing robustness-based ones. Unlike prior ones, our bounds are guaranteed to converge to the true error of the best classifier, as the number of samples increases. We further provide an extensive experiment and find that two of our bounds are often non-vacuous for a large class of deep neural networks, pretrained from ImageNet.

To appear in Machine Learning journal

References in corpus (16)

• Towards Deep Learning Models Resistant to Adversarial Attacks
• Understanding deep learning requires rethinking generalization
• A PAC-Bayesian Approach to Spectrally-Normalized Margin Bounds for Neural Networks
• Robust Large Margin Deep Neural Networks
• Spectrally-normalized margin bounds for neural networks
• Non-Vacuous Generalization Bounds at the ImageNet Scale: A PAC-Bayesian Compression Approach
• Robustness and Generalization for Metric Learning
• Stronger generalization bounds for deep nets via a compression approach
• High probability generalization bounds for uniformly stable algorithms with nearly optimal rate
• Fine-Grained Analysis of Stability and Generalization for Stochastic Gradient Descent
• PAC-Bayes Generalisation Bounds for Heavy-Tailed Losses through Supermartingales
• Robustness Implies Generalization via Data-Dependent Generalization Bounds
• Instance-Dependent Generalization Bounds via Optimal Transport
• Non-Vacuous Generalisation Bounds for Shallow Neural Networks
• Tighter PAC-Bayes Generalisation Bounds by Leveraging Example Difficulty
• Some Fundamental Aspects about Lipschitz Continuity of Neural Networks