Large Networks, Perturbation of Block Structures

This chapter applies the results of the previous chapters for the spectral clustering of large networks. Networks are modeled both by edge‐weighted graphs or contingency tables, and usually subject to random errors due to their evolving and flexible nature. Asymptotic properties of SD and SVD of the...

Full description

Saved in:
Bibliographic Details
Published inSpectral Clustering and Biclustering pp. 96 - 160
Main Author Bolla, Marianna
Format Book Chapter
LanguageEnglish
Published United Kingdom Wiley 2013
John Wiley & Sons, Ltd
Subjects
block‐structure
contingency tables
edge‐weighted graphs
large networks
MATHEMATICS
perturbation
Wigner‐type error matrix
Online AccessGet full text

Cover

More Information
Summary:This chapter applies the results of the previous chapters for the spectral clustering of large networks. Networks are modeled both by edge‐weighted graphs or contingency tables, and usually subject to random errors due to their evolving and flexible nature. Asymptotic properties of SD and SVD of the involved matrices are discussed when not only the number of graph vertices or the number of rows and columns of the contingency table tends to infinity, but the cluster sizes also grow proportionally with them. Mostly, perturbation results for the SD and SVD of blown‐up matrices burdened with a Wigner‐type error matrix are investigated. Conversely, given a weight‐matrix or rectangular array of nonnegative entries, we are looking for the underlying block‐structure. The chapter shows that under general circumstances, clusters of vertices of a large graph, and simultaneously those of the rows and columns of a large contingency table, are identified with high probability.
ISBN:9781118344927
1118344928
DOI:10.1002/9781118650684.ch03