Design for an Optimal Probe

Proceedings of the Twentieth International Conference on Machine Learning

Volume

Proceedings of the Twentieth International Conference on Machine Learning

Contents

Given a Markov decision process (MDP) with expressed prior uncertainties in the process transition probabilities, we consider the problem of computing a policy that optimizes expected total (finite-horizon) reward. Implicitly, such a policy would effectively resolve the "exploration-versus-exploitation tradeoff" faced, for example, by an agent that seeks to optimize total reinforcement obtained over the entire duration of its interaction with an uncertain world. A Bayesian formulation leads to an associated MDP defined over a set of generalized process "hyperstates" whose cardinality grows exponentiaily with the planning horizon. Here we retain the full Bayesian framework, but sidestep intractability by applying techniques from reinforcement learning theory. We apply our resulting actor-critic algorithm to a problem of "optimal probing," in which the task is to identify unknown transition probabilities of an MDP using online experience.