Toward Practical Reinforcement Learning Algorithms: Classification Based Policy Iteration And Model-Based Learning