← Search

Pengyue Hou

2 accepted papers

2023

Improving Adversarial Robustness with Self-Paced Hard-Class Pair Reweighting

AAAI 2023technical

Deep Neural Networks are vulnerable to adversarial attacks. Among many defense strategies, adversarial training with untargeted attacks is one of the most effective methods. Theoretically, adversarial perturbation in untargeted attacks can be added along arbitrary directions and the predicted labels…