Spectral analysis in bipartite biregular graphs and community detection

Abstract

This thesis concerns to spectral gap of random regular graphs and consists of two main con- tributions. First, we prove that almost all bipartite biregular graphs are almost Ramanujan by providing a tight upper bound for the non trivial eigenvalues of its adjacency operator, proving Alon's Conjecture for this family of graphs. Secondly, we use a spectral algorithm to recover hidden communities in a random network model we call regular stochastic block model. We rely on a technique introduced recently by Massoullie, which we develop here for random regular graphs.