Bellman-Shapley

New framework unifies bisimulation metrics for reinforcement learning

Researchers have developed a new framework for understanding bisimulation metrics in reinforcement learning, extending their application beyond discounted Markov decision processes (MDPs) to total- and average-reward se…