Wang, et al.. Distributed Bandit Learning: Near-optimal Regret with Efficient Communication. 29 May 2019.