Cautious Actor-Critic

Lingwei Zhu, Toshinori Kitamura, Matsubara Takamitsu
Proceedings of The 13th Asian Conference on Machine Learning, PMLR 157:220-235, 2021.

Abstract

The oscillating performance of off-policy learning and persisting errors in the actor-critic(AC) setting call for algorithms that can conservatively learn to suit the stability-critical applications better. In this paper, we propose a novel off-policy AC algorithm cautious actor-critic (CAC). The name cautious comes from the doubly conservative nature that we exploit the classic policy interpolation from conservative policy iteration for the actor and the entropy-regularization of conservative value iteration for the critic. Our key observation is the entropy-regularized critic facilitates and simplifies the unwieldy interpolated actor update while still ensuring robust policy improvement. We compare CAC to state-of-the-art AC methods on a set of challenging continuous control problems and demonstrate thatCAC achieves comparable performance while significantly stabilizes learning.

Cite this Paper

BibTeX
@InProceedings{pmlr-v157-zhu21a, title = {Cautious Actor-Critic}, author = {Zhu, Lingwei and Kitamura, Toshinori and Takamitsu, Matsubara}, booktitle = {Proceedings of The 13th Asian Conference on Machine Learning}, pages = {220--235}, year = {2021}, editor = {Balasubramanian, Vineeth N. and Tsang, Ivor}, volume = {157}, series = {Proceedings of Machine Learning Research}, month = {17--19 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v157/zhu21a/zhu21a.pdf}, url = {https://proceedings.mlr.press/v157/zhu21a.html}, abstract = {The oscillating performance of off-policy learning and persisting errors in the actor-critic(AC) setting call for algorithms that can conservatively learn to suit the stability-critical applications better. In this paper, we propose a novel off-policy AC algorithm cautious actor-critic (CAC). The name cautious comes from the doubly conservative nature that we exploit the classic policy interpolation from conservative policy iteration for the actor and the entropy-regularization of conservative value iteration for the critic. Our key observation is the entropy-regularized critic facilitates and simplifies the unwieldy interpolated actor update while still ensuring robust policy improvement. We compare CAC to state-of-the-art AC methods on a set of challenging continuous control problems and demonstrate thatCAC achieves comparable performance while significantly stabilizes learning.} }
Endnote
%0 Conference Paper %T Cautious Actor-Critic %A Lingwei Zhu %A Toshinori Kitamura %A Matsubara Takamitsu %B Proceedings of The 13th Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Vineeth N. Balasubramanian %E Ivor Tsang %F pmlr-v157-zhu21a %I PMLR %P 220--235 %U https://proceedings.mlr.press/v157/zhu21a.html %V 157 %X The oscillating performance of off-policy learning and persisting errors in the actor-critic(AC) setting call for algorithms that can conservatively learn to suit the stability-critical applications better. In this paper, we propose a novel off-policy AC algorithm cautious actor-critic (CAC). The name cautious comes from the doubly conservative nature that we exploit the classic policy interpolation from conservative policy iteration for the actor and the entropy-regularization of conservative value iteration for the critic. Our key observation is the entropy-regularized critic facilitates and simplifies the unwieldy interpolated actor update while still ensuring robust policy improvement. We compare CAC to state-of-the-art AC methods on a set of challenging continuous control problems and demonstrate thatCAC achieves comparable performance while significantly stabilizes learning.
APA
Zhu, L., Kitamura, T. & Takamitsu, M.. (2021). Cautious Actor-Critic. Proceedings of The 13th Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 157:220-235 Available from https://proceedings.mlr.press/v157/zhu21a.html.

Related Material