Learning Models For Reinforcement Learning From Human