Observed Adversaries in Deep Reinforcement Learning

Eugene Lim, Harold Soh

In this work, we point out the problem of observed adversaries for deep policies. Specifically, recent work has shown that deep reinforcement learning is susceptible to adversarial…