Meta Planning as Inference

I am currently researching on the theoretical formulation in the intersection of Reinforcment Learning & Meta Learning using an Probabilistic & Latent Inference perspective with my team and advising from Professor Sicun Gao from UCSD's Computer Science & Engineering Department and Professor Ying Nian Wu from UCLA Statistics & Data Science Department. We believe that probability itself carries certain "structure" that can help us systematically understand different environment and solve the multi-planning tasks.

Artistic representation of RL as inference
Artistic representation of RL as inference (credit to GenSpark for generation)
Original Codebase On-going Codebase