The Mirage of ActionDependent Baselines in Reinforcement Learning
[edit]
Proceedings of the 35th International Conference on Machine Learning, PMLR 80:50155024, 2018.
Abstract
Policy gradient methods are a widely used class of modelfree reinforcement learning algorithms where a statedependent baseline is used to reduce gradient estimator variance. Several recent papers extend the baseline to depend on both the state and action and suggest that this significantly reduces variance and improves sample efficiency without introducing bias into the gradient estimates. To better understand this development, we decompose the variance of the policy gradient estimator and numerically show that learned stateactiondependent baselines do not in fact reduce variance over a statedependent baseline in commonly tested benchmark domains. We confirm this unexpected result by reviewing the opensource code accompanying these prior papers, and show that subtle implementation decisions cause deviations from the methods presented in the papers and explain the source of the previously observed empirical gains. Furthermore, the variance decomposition highlights areas for improvement, which we demonstrate by illustrating a simple change to the typical value function parameterization that can significantly improve performance.
Related Material


