The Effectiveness of Random Forgetting for Robust Generalization

arXiv:2402.11733v1 Announce Type: new
Abstract: Deep neural networks are susceptible to adversarial attacks, which can compromise their performance and accuracy. Adversarial Training (AT) has emerged as a popular approach for protecting neural networks against such attacks. However, a key challenge of AT is robust overfitting, where the network's robust performance on test data deteriorates with further training, thus hindering generalization. Motivated by the concept of active forgetting in the brain, we introduce a novel learning paradigm called "Forget to Mitigate …

abstract accuracy adversarial adversarial attacks adversarial training arxiv attacks challenge cs.ai cs.cv cs.lg key network networks neural networks overfitting performance popular random robust test training type