Offline-Online Reinforcement Learning for Linear Mixture MDPs
For RL practitioners dealing with environment shift, this work provides a principled method to safely combine offline and online data, with theoretical guarantees.
The paper studies offline-online RL in linear mixture MDPs under environment shift, proposing an algorithm that adaptively leverages offline data. It provably improves over purely online learning when offline data are informative and matches online-only performance otherwise, with regret bounds and matching lower bounds.
We study offline-online reinforcement learning in linear mixture Markov decision processes (MDPs) under environment shift. In the offline phase, data are collected by an unknown behavior policy and may come from a mismatched environment, while in the online phase the learner interacts with the target environment. We propose an algorithm that adaptively leverages offline data. When the offline data are informative, either due to sufficient coverage or small environment shift, the algorithm provably improves over purely online learning. When the offline data are uninformative, it safely ignores them and matches the online-only performance. We establish regret upper bounds that explicitly characterize when offline data are beneficial, together with nearly matching lower bounds. Numerical experiments further corroborate our theoretical findings.