Learning with a Strong Adversary

Ruitong Huang, Bing Xu, Dale Schuurmans +1

The robustness of neural networks to intended perturbations has recently attracted significant attention. In this paper, we propose a new method, \emph{learning with a strong adver…