A probabilistic analysis of bias optimality in unichain Markov decision processes

This paper focuses on bias optimality in unichain, finite state, and action -space Markov decision processes, Using relative value functions, we presen t new methods for evaluating optimal bias. This leads to a probabilistic an alysis which transforms the original reward problem into a minimum average cost problem. The result is an explanation of how and why bias implicitly d iscounts future rewards.