
Amy Zhang
TalkRL: The Reinforcement Learning Podcast
00:00
Understanding Block MDPs and Abstractions
This chapter explores block Markov Decision Processes (block MDPs) and their advantages in structured environments, focusing on latent state spaces and model irrelevance abstraction. It also examines state representations, bisimulation, and the significance of coarser abstractions in reinforcement learning, illustrated through practical examples.
Transcript
Play full episode