Value Based Reinforcement Learning