CEB Improves Model Robustness

Abstract

We demonstrate that the Conditional Entropy Bottleneck (CEB) can improve model robustness. CEB is an easy strategy to implement and works in tandem with data augmentation procedures. We report results of a large scale adversarial robustness study on CIFAR-10, as well as the ImageNet-C Common Corruptions Benchmark, ImageNet-A, and PGD attacks.

References in corpus (8)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Towards Deep Learning Models Resistant to Adversarial Attacks
• Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examples
• Adversarial Examples Are Not Bugs, They Are Features
• Benchmarking Neural Network Robustness to Common Corruptions and Perturbations
• The Conditional Entropy Bottleneck
• A Fourier Perspective on Model Robustness in Computer Vision
• Learnability for the Information Bottleneck