Stochastic Graph Neural Network-Based Value Decomposition for Multi-Agent Reinforcement Learning in Urban Traffic Control | IEEE Conference Publication | IEEE Xplore