Learning Models For Reinforcement Learning