On the Global Convergence of Risk-Averse Policy Gradient Methods with Expected Conditional Risk Measures

Xian Yu, Lei Ying
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:40425-40451, 2023.

Abstract

Risk-sensitive reinforcement learning (RL) has become a popular tool to control the risk of uncertain outcomes and ensure reliable performance in various sequential decision-making problems. While policy gradient methods have been developed for risk-sensitive RL, it remains unclear if these methods enjoy the same global convergence guarantees as in the risk-neutral case. In this paper, we consider a class of dynamic time-consistent risk measures, called Expected Conditional Risk Measures (ECRMs), and derive policy gradient updates for ECRM-based objective functions. Under both constrained direct parameterization and unconstrained softmax parameterization, we provide global convergence and iteration complexities of the corresponding risk-averse policy gradient algorithms. We further test risk-averse variants of REINFORCE and actor-critic algorithms to demonstrate the efficacy of our method and the importance of risk control.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-yu23j, title = {On the Global Convergence of Risk-Averse Policy Gradient Methods with Expected Conditional Risk Measures}, author = {Yu, Xian and Ying, Lei}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {40425--40451}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/yu23j/yu23j.pdf}, url = {https://proceedings.mlr.press/v202/yu23j.html}, abstract = {Risk-sensitive reinforcement learning (RL) has become a popular tool to control the risk of uncertain outcomes and ensure reliable performance in various sequential decision-making problems. While policy gradient methods have been developed for risk-sensitive RL, it remains unclear if these methods enjoy the same global convergence guarantees as in the risk-neutral case. In this paper, we consider a class of dynamic time-consistent risk measures, called Expected Conditional Risk Measures (ECRMs), and derive policy gradient updates for ECRM-based objective functions. Under both constrained direct parameterization and unconstrained softmax parameterization, we provide global convergence and iteration complexities of the corresponding risk-averse policy gradient algorithms. We further test risk-averse variants of REINFORCE and actor-critic algorithms to demonstrate the efficacy of our method and the importance of risk control.} }
Endnote
%0 Conference Paper %T On the Global Convergence of Risk-Averse Policy Gradient Methods with Expected Conditional Risk Measures %A Xian Yu %A Lei Ying %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-yu23j %I PMLR %P 40425--40451 %U https://proceedings.mlr.press/v202/yu23j.html %V 202 %X Risk-sensitive reinforcement learning (RL) has become a popular tool to control the risk of uncertain outcomes and ensure reliable performance in various sequential decision-making problems. While policy gradient methods have been developed for risk-sensitive RL, it remains unclear if these methods enjoy the same global convergence guarantees as in the risk-neutral case. In this paper, we consider a class of dynamic time-consistent risk measures, called Expected Conditional Risk Measures (ECRMs), and derive policy gradient updates for ECRM-based objective functions. Under both constrained direct parameterization and unconstrained softmax parameterization, we provide global convergence and iteration complexities of the corresponding risk-averse policy gradient algorithms. We further test risk-averse variants of REINFORCE and actor-critic algorithms to demonstrate the efficacy of our method and the importance of risk control.
APA
Yu, X. & Ying, L.. (2023). On the Global Convergence of Risk-Averse Policy Gradient Methods with Expected Conditional Risk Measures. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:40425-40451 Available from https://proceedings.mlr.press/v202/yu23j.html.

Related Material