Squeeze Training for Adversarial Robustness. (arXiv:2205.11156v2 [cs.LG] UPDATED)

The vulnerability of deep neural networks (DNNs) to adversarial examples has
attracted great attention in the machine learning community. The problem is
related to non-flatness and non-smoothness of normally obtained loss
landscapes. Training augmented with adversarial examples (a.k.a., adversarial
training) is considered as an effective remedy. In this paper, we highlight
that some collaborative examples, nearly perceptually indistinguishable from
both adversarial and benign examples yet show extremely lower prediction loss,
can be utilized to enhance adversarial training. A novel method …

adversarial attention community great loss machine machine learning networks neural networks non prediction problem remedy robustness training vulnerability