Policy DistillationPolicies for complex visual tasks have been successfully learned with deep reinforcement learning, using an approach called deep Q-networks (DQN), but relatively large (task-specific) networks and...https://arxiv.org/abs/1511.06295
