Intrinsic Reward Reinforcement Learning