English

Regularization Matters in Policy Optimization

Machine Learning 2021-11-30 v5 Artificial Intelligence Machine Learning

Abstract

Deep Reinforcement Learning (Deep RL) has been receiving increasingly more attention thanks to its encouraging performance on a variety of control tasks. Yet, conventional regularization techniques in training neural networks (e.g., L2L_2 regularization, dropout) have been largely ignored in RL methods, possibly because agents are typically trained and evaluated in the same environment, and because the deep RL community focuses more on high-level algorithm designs. In this work, we present the first comprehensive study of regularization techniques with multiple policy optimization algorithms on continuous control tasks. Interestingly, we find conventional regularization techniques on the policy networks can often bring large improvement, especially on harder tasks. Our findings are shown to be robust against training hyperparameter variations. We also compare these techniques with the more widely used entropy regularization. In addition, we study regularizing different components and find that only regularizing the policy network is typically the best. We further analyze why regularization may help generalization in RL from four perspectives - sample complexity, reward distribution, weight norm, and noise robustness. We hope our study provides guidance for future practices in regularizing policy optimization algorithms. Our code is available at https://github.com/xuanlinli17/iclr2021_rlreg .

Keywords

Cite

@article{arxiv.1910.09191,
  title  = {Regularization Matters in Policy Optimization},
  author = {Zhuang Liu and Xuanlin Li and Bingyi Kang and Trevor Darrell},
  journal= {arXiv preprint arXiv:1910.09191},
  year   = {2021}
}

Comments

Published at ICLR 2021; please cite this paper's ICLR 2021 version at https://github.com/xuanlinli17/iclr2021_rlreg#citation or the arXiv version from "Export Bibtex Citation" on the right, instead of the "2019 OpenReview" version in Google Scholar. Thanks!