NeurIPS camready: MDP Homomorphic Networks Posted on November 5, 2020 | by frans In this work we show how symmetries that can occur in MDPs can be exploited for more efficient deep reinforcement learning.