Model-Based Reinforcement Learning with a Generative Model is Minimax Optimal.

Alekh Agarwal, Sham M. Kakade, Lin F. Yang

Browse the full COLT paper archive.