Action-depedent Control Variates for Policy Optimization via Stein's Identity
This work addresses a key bottleneck in reinforcement learning for practitioners by enhancing sample efficiency, though it is incremental as it builds on existing control variate methods.
The paper tackles the large variance issue in policy gradient estimation, which hinders sample efficiency, by proposing a control variate method based on Stein's identity that introduces action-dependent baseline functions, resulting in significant improvements in sample efficiency for state-of-the-art policy gradient approaches.
Policy gradient methods have achieved remarkable successes in solving challenging reinforcement learning problems. However, it still often suffers from the large variance issue on policy gradient estimation, which leads to poor sample efficiency during training. In this work, we propose a control variate method to effectively reduce variance for policy gradient methods. Motivated by the Stein's identity, our method extends the previous control variate methods used in REINFORCE and advantage actor-critic by introducing more general action-dependent baseline functions. Empirical studies show that our method significantly improves the sample efficiency of the state-of-the-art policy gradient approaches.