paper

An Alon-Boppana Type Bound for Weighted Graphs and Lowerbounds for Spectral Sparsification

arXiv:1707.06364

Abstract

We prove the following Alon-Boppana type theorem for general (not necessarily regular) weighted graphs: if is an -node weighted undirected graph of average combinatorial degree (that is, has edges) and girth , and if are the eigenvalues of the (non-normalized) Laplacian of , then \[ \frac {λ_n}{λ_2} \geq 1 + \frac 4{\sqrt d} - O \left( \frac 1{d^{\frac 58} }\right) \] (The Alon-Boppana theorem implies that if is unweighted and -regular, then if the diameter is at least .) Our result implies a lower bound for spectral sparsifiers. A graph is a spectral -sparsifier of a graph if \[ L(G) \preceq L(H) \preceq (1+ε) L(G) \] where is the Laplacian matrix of and is the Laplacian matrix of . Batson, Spielman and Srivastava proved that for every there is an -sparsifier of average degree where and the edges of are a (weighted) subset of the edges of . Batson, Spielman and Srivastava also show that the bound on cannot be reduced below when is a clique; our Alon-Boppana-type result implies that cannot be reduced below when comes from a family of expanders of super-constant degree and super-constant girth. The method of Batson, Spielman and Srivastava proves a more general result, about sparsifying sums of rank-one matrices, and their method applies to an "online" setting. We show that for the online matrix setting the bound is tight, up to lower order terms.