On Principled Entropy Exploration in Policy Optimization
Abstract
In this paper, we investigate Exploratory Conservative Policy Optimization (ECPO), a policy optimization strategy that improves exploration behavior while assuring monotonic progress in a principled objective. ECPO conducts maximum entropy exploration within a mirror descent framework, but updates policies using reversed KL projection. This formulation bypasses undesirable mode seeking behavior and avoids premature convergence to sub-optimal policies, while still supporting strong theoretical properties such as guaranteed policy improvement. Experimental evaluations demonstrate that the proposed method significantly improves practical exploration and surpasses the empirical performance of state-of-the art policy optimization methods in a set of benchmark tasks.
Cite
Text
Mei et al. "On Principled Entropy Exploration in Policy Optimization." International Joint Conference on Artificial Intelligence, 2019. doi:10.24963/IJCAI.2019/434Markdown
[Mei et al. "On Principled Entropy Exploration in Policy Optimization." International Joint Conference on Artificial Intelligence, 2019.](https://mlanthology.org/ijcai/2019/mei2019ijcai-principled/) doi:10.24963/IJCAI.2019/434BibTeX
@inproceedings{mei2019ijcai-principled,
title = {{On Principled Entropy Exploration in Policy Optimization}},
author = {Mei, Jincheng and Xiao, Chenjun and Huang, Ruitong and Schuurmans, Dale and Müller, Martin},
booktitle = {International Joint Conference on Artificial Intelligence},
year = {2019},
pages = {3130-3136},
doi = {10.24963/IJCAI.2019/434},
url = {https://mlanthology.org/ijcai/2019/mei2019ijcai-principled/}
}