AdaRL: What, Where, and How to Adapt in Transfer Reinforcement Learning
This work addresses the problem of sample-efficient adaptation in transfer reinforcement learning for practitioners, though it appears incremental as it builds on existing graphical representation methods.
The paper tackles the challenge of quickly adapting reinforcement learning policies to new environments by proposing AdaRL, a framework that uses graphical representations to identify and encode domain changes, enabling efficient adaptation with only a few samples and avoiding further policy optimization, achieving reliable performance in partially observable settings like Cartpole and Atari games.
One practical challenge in reinforcement learning (RL) is how to make quick adaptations when faced with new environments. In this paper, we propose a principled framework for adaptive RL, called \textit{AdaRL}, that adapts reliably and efficiently to changes across domains with a few samples from the target domain, even in partially observable environments. Specifically, we leverage a parsimonious graphical representation that characterizes structural relationships over variables in the RL system. Such graphical representations provide a compact way to encode what and where the changes across domains are, and furthermore inform us with a minimal set of changes that one has to consider for the purpose of policy adaptation. We show that by explicitly leveraging this compact representation to encode changes, we can efficiently adapt the policy to the target domain, in which only a few samples are needed and further policy optimization is avoided. We illustrate the efficacy of AdaRL through a series of experiments that vary factors in the observation, transition, and reward functions for Cartpole and Atari games.