Inverse Reinforcement Learning via Matching of Optimality Profiles