MVA-2021 / reinforcement_learning / hw2_approximate_rl / fig