Stability and Generalization in Free Adversarial Training

arXiv:2404.08980v1 Announce Type: new
Abstract: While adversarial training methods have resulted in significant improvements in the deep neural nets' robustness against norm-bounded adversarial perturbations, their generalization performance from training samples to test data has been shown to be considerably worse than standard empirical risk minimization methods. Several recent studies seek to connect the generalization behavior of adversarially trained classifiers to various gradient-based min-max optimization algorithms used for their training. In this work, we study the generalization performance of adversarial training …

adversarial adversarial training arxiv cs.lg free stability stat.ml training type