Fig. 5From: Adversarial attack and defense in reinforcement learning-from AI security viewSamples for Dominant Adversarial Examples. For the first column is the original clean map for path finding. For columns on the right are the samples for Dominant Adversarial Examples generated by CDG algorithm proposed in this paper, and (a), (b), (c), (d) represent four different samples for dominant adversarial examplesBack to article page