Efficient PAC Reinforcement Learning in Regular Decision Processes
This provides a theoretical foundation for efficient learning in structured non-Markovian environments, though it appears incremental as it extends PAC learning to a specific model class.
The paper tackles reinforcement learning in regular decision processes, a non-Markovian framework, by proving that a near-optimal policy can be PAC-learned in polynomial time based on minimal parameters that capture the process difficulty.
Recently regular decision processes have been proposed as a well-behaved form of non-Markov decision process. Regular decision processes are characterised by a transition function and a reward function that depend on the whole history, though regularly (as in regular languages). In practice both the transition and the reward functions can be seen as finite transducers. We study reinforcement learning in regular decision processes. Our main contribution is to show that a near-optimal policy can be PAC-learned in polynomial time in a set of parameters that describe the underlying decision process. We argue that the identified set of parameters is minimal and it reasonably captures the difficulty of a regular decision process.