Representation Learning and Exploration in Reinforcement Learning