Abstract

In this work, we build on recent advances in distributional reinforcement learning to give a generally applicable, flexible, and state-of-the-art dis tributional variant of DQN. We achieve this by using quantile regression to approximate the full quantile function for the state-action return distr bution. By reparameterizing a distribution over the sample space, this yields an implicitly defined return distribution and gives rise to a large class risk-sensitive policies. We demonstrate improved performance on the 57 Atari 2600 games in the ALE, and use our algorithm’s implicitly defined distributions to study the effects of risk-sensitiv policies in Atari games.