Offline Policy Optimization with Variance Regularization

Riashat Islam, Samarth Sinha, Homanga Bharadhwaj, Samin Yeasar Arnob, Zhuoran Yang, Zhaoran Wang, Animesh Garg, Lihong Li, Doina Precup

Keywords: reinforcement learning, offline batch RL, off-policy, policy optimization, variance regularization

Abstract: Learning policies from fixed offline datasets is a key challenge to scale up reinforcement learning (RL) algorithms towards practical applications. This is often because off-policy RL algorithms suffer from distributional shift, due to mismatch between dataset and the target policy, leading to high variance and over-estimation of value functions. In this work, we propose variance regularization for offline RL algorithms, using stationary distribution corrections. We show that by using Fenchel duality, we can avoid double sampling issues for computing the gradient of the variance regularizer. The proposed algorithm for offline variance regularization can be used to augment any existing offline policy optimization algorithms. We show that the regularizer leads to a lower bound to the offline policy optimization objective, which can help avoid over-estimation errors, and explains the benefits of our approach across a range of continuous control domains when compared to existing algorithms.

One-sentence Summary: Variance regularization based on stationary state-action distribution corrections in offline policy optimization

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Reviewed Version (pdf): https://openreview.net/references/pdf?id=zXmylJQnhT