Natural Policy Gradient for Average Reward Non-Stationary RL
Keywords: Non-Stationary Reinforcement Learning, Policy Gradient, Natural Actor-Critic
TL;DR: We present a dynamic regret analysis of the Natural Actor-Critic algorithm for infinite horizon average reward Non-Stationary Reinforcement Learning.
Abstract: We consider the problem of non-stationary reinforcement learning (RL) in the infinite-horizon average-reward setting. We model it by a Markov Decision Process with time-varying rewards and transition probabilities, with a variation budget of $\Delta_T$. Existing non-stationary RL algorithms focus on model-based and model-free value-based methods. Policy-based methods, however, despite their flexibility in practice, are not theoretically well understood in non-stationary RL. We propose and analyze the first model-free policy-based algorithm, Non-Stationary Natural Actor-Critic (NS-NAC), a policy gradient method with efficient exploration for change and a novel interpretation of learning rates as adapting factors. We present a dynamic regret of $\mathcal{\tilde{O}} (|\mathcal{S}|^{\frac{1}{2}}|\mathcal{A}|^{\frac{1}{2}}\Delta_T^{\frac{1}{9}}T^{\frac{8}{9}} )$, where $T$ is the time horizon, and $|\mathcal{S}|$, $|\mathcal{A}|$ are, respectively, the size of the state and action space. The regret analysis relies on adapting the Lyapunov function based analysis to dynamic environments and characterizing the effects of simultaneous updates in policy, value function estimate and environment.
Primary Area: reinforcement learning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 12405
Loading