Multi-agent Markov Entanglement

Shuze Chen, Tianyi Peng

Value decomposition has long been a fundamental technique in multi-agent dynamic programming and reinforcement learning (RL). Specifically, the value function of a global state $(s…