Federated Q-Learning with Reference-Advantage Decomposition: Almost Optimal Regret and Logarithmic Communication Cost.

Zhong Zheng, Haochen Zhang, Lingzhou Xue

Browse the full ICLR paper archive.