Model-based Adversarial Meta-Reinforcement Learning

Meta-reinforcement learning and multi-task reinforcement learning aim to improve the sample efficiency by leveraging the shared structure within a family of tasks.