Sample Complexity of Variance-reduced Distributionally Robust Q-learning
This addresses the challenge of distributional shifts in RL for theoretical and applied settings, with incremental improvements in sample complexity analysis.
The paper tackles the problem of learning robust policies in reinforcement learning under distributional shifts by proposing distributionally robust Q-learning algorithms, achieving a minimax sample complexity upper bound of ̃O(|S||A|(1-γ)^{-4}ε^{-2}) that is independent of ambiguity size.
Dynamic decision-making under distributional shifts is of fundamental interest in theory and applications of reinforcement learning: The distribution of the environment in which the data is collected can differ from that of the environment in which the model is deployed. This paper presents two novel model-free algorithms, namely the distributionally robust Q-learning and its variance-reduced counterpart, that can effectively learn a robust policy despite distributional shifts. These algorithms are designed to efficiently approximate the $q$-function of an infinite-horizon $γ$-discounted robust Markov decision process with Kullback-Leibler ambiguity set to an entry-wise $ε$-degree of precision. Further, the variance-reduced distributionally robust Q-learning combines the synchronous Q-learning with variance-reduction techniques to enhance its performance. Consequently, we establish that it attains a minimax sample complexity upper bound of $\tilde O(|\mathbf{S}||\mathbf{A}|(1-γ)^{-4}ε^{-2})$, where $\mathbf{S}$ and $\mathbf{A}$ denote the state and action spaces. This is the first complexity result that is independent of the ambiguity size $δ$, thereby providing new complexity theoretic insights. Additionally, a series of numerical experiments confirm the theoretical findings and the efficiency of the algorithms in handling distributional shifts.