Offline Meta-Reinforcement Learning with Advantage Weighting

          Eric Mitchell 1 Rafael Rafailov 1 Xue Bin Peng 2 Sergey Levine 2 Chelsea Finn 1

             Abstract                  of reinforcement learning algorithms, when the goal is to
                                   ultimately learn many tasks. Meta-RL algorithms exploit
    This paper introduces the offline meta-            shared structure among tasks during meta-training, amor-
    reinforcement learning (offline meta-RL)    ...