An Empirical Comparison of Abstraction in Models of Markov Decision Processes

Reinforcement learning studies the problem of solving sequential decision making problems. Model-based methods learn an effective policy in few actions by learning a model of the domain and simulating