Fast Convergence for Consensus in Dynamic Networks

T.-H. Hubert Chan, Li Ning

Abstract: We study the convergence time required to achieve consensus in dynamic networks. In each time step, a node’s value is updated to some weighted average of its neighbors’ and its old values. We study the case when the underlying network is dynamic, and investigate different averaging models. Both our analysis and experiments show that dynamic networks exhibit fast convergence behavior, even under very mild connectivity assumptions.