Transfer In Reinforcement Learning Domains by Matthew Taylor