Regret Bounds for Reinforcement Learning and other Results of Computational Learning Theory