Finite Sample Analysis of Minimax Offline Reinforcement Learning: Completeness, Fast Rates and First-Order Efficiency
This work provides theoretical guarantees and improved efficiency for off-policy evaluation in reinforcement learning, which is crucial for safely deploying RL agents without real-world interaction.
This paper theoretically characterizes off-policy evaluation (OPE) in reinforcement learning using function approximation for marginal importance weights and q-functions, estimated via minimax methods. It achieves fast convergence rates for weights and quality functions and presents the first finite-sample result with first-order efficiency in non-tabular environments.
We offer a theoretical characterization of off-policy evaluation (OPE) in reinforcement learning using function approximation for marginal importance weights and $q$-functions when these are estimated using recent minimax methods. Under various combinations of realizability and completeness assumptions, we show that the minimax approach enables us to achieve a fast rate of convergence for weights and quality functions, characterized by the critical inequality \citep{bartlett2005}. Based on this result, we analyze convergence rates for OPE. In particular, we introduce novel alternative completeness conditions under which OPE is feasible and we present the first finite-sample result with first-order efficiency in non-tabular environments, i.e., having the minimal coefficient in the leading term.