Stochastic Abstract Policies: Generalizing Knowledge to Improve Reinforcement Learning | IEEE Journals & Magazine | IEEE Xplore