This talk has moved to Spring 2016 — Sample Complexity of Episodic Fixed-Horizon Reinforcement Learning.