Spectral Graph Theory via Higher Order Eigenvalues and Applications to the Analysis of Random Walks

Abstract

A basic goal in the field of spectral theory is to relate eigenvalues of matrices associated to a graph, namely the adjacency matrix, the Laplacian matrix or the random walk matrix, to the combinatorial properties of that graph. Classical results in this area mostly study the properties of first, second or the last eigenvalues of these matrices [2, 3, 4, 21]. In the last several years many of these results are extended and the bounds are improved using higher order eigenvalues. In this short monologue we overview several of these recent advances, and we describe one of the fundamental tools employed in these results, namely, the spectral embedding of graphs.