Generalization Bounds for Adversarial Contrastive Learning

Xin Zou, Weiwei Liu

Deep networks are well-known to be fragile to adversarial attacks, and adversarial training is one of the most popular methods used to train a robust model. To take advantage of un…