Learning Task Belief Similarity with Latent Dynamics for Meta-Reinforcement Learning

Menglong Zhang, Fuyuan Qian

Meta-reinforcement learning requires utilizing prior task distribution information obtained during exploration to rapidly adapt to unknown tasks. The efficiency of an agent's explo…