arXiv++ Combinatorics

Browse math.CO papers from arXiv

Sharp edge, vertex, and mixed Cheeger type inequalities for finite Markov kernels

Published: 2006-06-07 | Updated: 2007-05-05
Comments: v1: original version (>20 pages) v2: v1 was far too verbose; this pares it to under 10 pages
Journal: Electronic Communications in Probability, vol. 12, pp. 377-389, 2007.

Abstract

We show how the evolving set methodology of Morris and Peres can be used to show Cheeger inequalities for bounding the spectral gap of a finite Markov kernel. This leads to sharp versions of several previous Cheeger inequalities, including ones involving edge-expansion, vertex-expansion, and mixtures of both. A bound on the smallest eigenvalue also follows.

BibTeX

Loading...