Gerencsér Balázs
(Rényi Alfréd Matematikai Kutatóintézet, Budapest)

Multiple approaches for convergence rate approximation of ratio consensus

Absztrakt: Distributed average consensus, i.e. computing the average of inputs using only local communication along a network can be used by itself, e.g. for sensor fusion, or as a building block of an algorithm, e.g. for distributed optimization. We currently analyze average consensus schemes based on one such protocol, ratio consensus (see also push-sum), and our target is to understand their almost sure exponential convergence rate. We present multiple ways to address this problem. One by identifying the exact convergence rate, however based on asymptotic quantities. It is then possible to construct a computable upper bound for the i.i.d. setup. Finally, for certain subcases an alternative bound can be formulated with lower complexity - thus better scalability.
Joint work with L. Gerencsér, M. Kornyik.