Tag
This paper proposes the first FTRL-type algorithms for decentralized online convex optimization with compressed communication, achieving elegant theoretical guarantees and improved regret bounds compared to previous OGD-type methods.