Title :
Convergence rate of a distributed algorithm for matrix scaling to doubly stochastic form
Author :
Dominguez-Garcia, Alejandro D. ; Hadjicostis, Christoforos N.
Author_Institution :
ECE Dept., Univ. of Illinois, Urbana, IL, USA
Abstract :
Motivated by matrix scaling applications and, more recently, distributed averaging previous work has considered settings where the interconnections between components in a distributed system are captured by a strongly connected directed graph (digraph) and each component aims to assign assigning weights on its outgoing edges (based on the weights on its incoming edges) so that the corresponding set of weights forms a doubly stochastic matrix. In particular, it has been shown that the system components can obtain a set of weights that form a doubly stochastic matrix via a variety of distributed algorithms. In this paper, we establish that the convergence rate of one such distributed algorithm is linear with rate between zero and one.
Keywords :
directed graphs; distributed algorithms; matrix algebra; stochastic processes; convergence rate; digraph; distributed algorithm convergence rate; doubly stochastic matrix; matrix scaling; strongly connected directed graph; Convergence; Distributed algorithms; Educational institutions; Eigenvalues and eigenfunctions; Electronic mail; Matrix decomposition; Vectors;
Conference_Titel :
Decision and Control (CDC), 2014 IEEE 53rd Annual Conference on
Conference_Location :
Los Angeles, CA
Print_ISBN :
978-1-4799-7746-8
DOI :
10.1109/CDC.2014.7039890