Einladung zum Vortrag im Kolloquium Technische Kybernetik

  On the mean square error of randomized averaging algorithms

Dr. Paolo Frasca
Dipartimento di Matematica
Politecnico di Torino
 

Tuesday, 15. May 2012, 4:00 p.m.
IST-Seminar-Room 3.243 - Pfaffenwaldring 9 - Campus Stuttgart-Vaihingen

Abstract

By nature or by design randomness is inherent to several algorithms for network coordination. In this talk, I will consider randomized linear algorithms which are used to approximate average consensus. Among these algorithms, there are significant examples in which randomness induces a deviation between the average of the initial conditions and the actual convergence value. I will show that, under mild conditions on the distribution of the update matrices, the expected deviation has an upper bound which is inversely proportional to the size of the network. This result has been obtained in collaboration with J.M. Hendrickx.

Biographical Information

Paolo Frasca received the Ph.D. degree in Applied Mathematics from Politecnico di Torino, Italy, in 2009. His research interests include mathematical control theory and applications of probability and graph theory to distributed and networked systems.