Contributions To Batch Mode Reinforcement Learning