Provably Convergent Two-Timescale Off-Policy Actor-Critic with Function Approximation
Abstract
We present the first provably convergent two-timescale off-policy actor-critic algorithm (COF-PAC) with function approximation. Key to COF-PAC is the introduction of a new critic, the emphasis critic, which is trained via Gradient Emphasis Learning (GEM), a novel combination of the key ideas of Gradient Temporal Difference Learning and Emphatic Temporal Difference Learning. With the help of the emphasis critic and the canonical value function critic, we show convergence for COF-PAC, where the critics are linear and the actor can be nonlinear.
Cite
Text
Zhang et al. "Provably Convergent Two-Timescale Off-Policy Actor-Critic with Function Approximation." International Conference on Machine Learning, 2020.Markdown
[Zhang et al. "Provably Convergent Two-Timescale Off-Policy Actor-Critic with Function Approximation." International Conference on Machine Learning, 2020.](https://mlanthology.org/icml/2020/zhang2020icml-provably/)BibTeX
@inproceedings{zhang2020icml-provably,
title = {{Provably Convergent Two-Timescale Off-Policy Actor-Critic with Function Approximation}},
author = {Zhang, Shangtong and Liu, Bo and Yao, Hengshuai and Whiteson, Shimon},
booktitle = {International Conference on Machine Learning},
year = {2020},
pages = {11204-11213},
volume = {119},
url = {https://mlanthology.org/icml/2020/zhang2020icml-provably/}
}