Distributed Stochastic Consensus Optimization Using Communication-Censoring Strategy

In this paper, a novel communication-efficient distributed stochastic algorithm (referred as CO-DSA) is proposed for solving large-scale consensus optimization problems. As compared to the existing relevant work where only sublinear convergence rate obtained strongly convex and smooth objective functions, CO-DSA achieves linear even in presence of an event-triggered based communication-censoring strategy. Moreover, by properly setting threshold function communication scheme, maintains same without communication. This means theoretically yields efficiency free. Numerical experiments verify theoretical findings also show excellent saving effect large networks.