Design of Experiments for Reinforcement Learning