Finding Structure in Reinforcement Learning