Adversarially Guided Actor-Critic

Despite definite success in deep reinforcement learning problems, actor-critic algorithms are still confronted with sample inefficiency in complex environments. Many of those methods consider an actor and a critic whose respective losses are obtained using different motivations and approaches. In this paper, we introduce a third protagonist, the adversary, whose task is to match the action probability distribution of the actor. While the adversary minimizes the KL-divergence between its action distribution and the actor policy, the actor maximizes the log-probability difference between its action and that of the adversary in combination with maximizing expected rewards. This novel objective stimulates the actor to follow strategies that could not have been correctly predicted from previous trajectories, making its behavior innovative in tasks where the reward is extremely rare. Our experimental analysis shows that the resulting Adversarially Guided Actor-Critic (AGAC) algorithm leads to more exhaustive exploration. Notably, AGAC outperforms current state-of-the-art methods on a set of various hard-exploration and procedurally-generated tasks.

• Machine Intelligence