CEB Improves Model Robustness
arXiv:2002.05380v132 citations
AI Analysis
This addresses robustness issues in machine learning models, but appears incremental as it builds on existing methods like data augmentation.
The paper tackled improving model robustness by applying the Conditional Entropy Bottleneck (CEB) strategy, reporting results on adversarial robustness studies across datasets like CIFAR-10, ImageNet-C, ImageNet-A, and PGD attacks.
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.