Title :
Products of stochastic matrices: Exact rate for convergence in probability for directed networks
Author :
Bajovic, Dragana ; Xavier, Joao ; Sinopoli, Bruno
Author_Institution :
Inst. for Syst. & Robot. (ISR, Inst. Super. Tecnico (IST), Lisbon, Portugal
Abstract :
We study the products Wk···W1 of random stochastic, not necessarily symmetric matrices. It is known that, under certain conditions, the product Wk · · · W1 converges almost surely (a.s.) to a random rank-one matrix; the latter is equivalent to |λ2(Wk · · · W1)| → 0 a.s., where λ2(·) is the second largest (in modulus) eigenvalue. In this paper, we show that the probability that |λ2(Wk · · · W1)| stays above ε ∈ (0,1] in the long run decays to zero exponentially fast ~ e-kI. Furthermore, we explicitly characterize the rate of this convergence I and show that it depends only on the underlying graphs that support the matrices Wk´s. Our results reveal that the rate I is essentially determined by the most likely way in which the union (over time) of the support graphs fails to form a directed tree.
Keywords :
directed graphs; eigenvalues and eigenfunctions; stochastic processes; tree searching; directed networks; directed tree; eigenvalue; graphs; probability; random rank one matrix; random stochastic matrices; symmetric matrices; Computational modeling; Convergence; Eigenvalues and eigenfunctions; Robot sensing systems; Stochastic processes; Symmetric matrices; Consensus; Convergence in probability; Directed networks; Exponential rate; Stochastic matrices;
Conference_Titel :
Telecommunications Forum (TELFOR), 2012 20th
Conference_Location :
Belgrade
Print_ISBN :
978-1-4673-2983-5
DOI :
10.1109/TELFOR.2012.6419349