A Reinforcement Learning Approach