Meta Automatic Curriculum Learning

Rémy Portelas, Clément Romac, Katja Hofmann +1

A major challenge in the Deep RL (DRL) community is to train agents able to generalize their control policy over situations never seen in training. Training on diverse tasks has be…