AAAI Publications, Twentieth International Conference on Automated Planning and Scheduling

Font Size: 
When Policies Can Be Trusted: Analyzing a Criteria to Identify Optimal Policies in MDPs with Unknown Model Parameters
Emma Brunskill

Last modified: 2010-04-20


Computing a good policy in stochastic uncertain environments with unknown dynamics and reward model parameters is a challenging task. In a number of domains, ranging from space robotics to epilepsy management, it may be possible to have an initial training period when suboptimal performance is permitted. For such problems it is important to be able to identify when this training period is complete, and the computed policy can be used with high confidence in its future performance. A simple principled criteria for identifying when training has completed is when the error bounds on the value estimates of the current policy are sufficiently small that the optimal policy is fixed, with high probability. We present an upper bound on the amount of training data required to identify the optimal policy as a function of the unknown separation gap between the optimal and the next-best policy values. We illustrate with several small problems that by estimating this gap in an online manner, the number of training samples to provably reach optimality can be significantly lower than predicted offline using a Probably Approximately Correct framework that requires an input epsilon parameter.


reinforcement learning; markov decision process; policy identification; probably approximately correct

Full Text: PDF