Improved Network Robustness with Adversary Critic

Alexander Matyasko, Lap-Pui Chau

Ideally, what confuses neural network should be confusing to humans. However, recent experiments have shown that small, imperceptible perturbations can change the network predictio…