Tree simplification and the 'plateaux' phenomenon of graph Laplacian eigenvalues

TL;DR

This paper introduces a tree simplification method to reduce computational costs in analyzing graph Laplacian eigenvalues and investigates the emergence of eigenvalue plateaux in simplified trees and general graphs.

Contribution

It presents a novel tree simplification procedure that preserves topological information and explains the occurrence of eigenvalue plateaux in simplified trees and broader graphs.

Findings

Eigenvalue plateaux are observed in simplified trees representing retinal dendritic structures.

The paper identifies conditions under which eigenvalue plateaux occur in general graphs.

Specific eigenvalues and their multiplicity bounds are characterized.

Abstract

We developed a procedure of reducing the number of vertices and edges of a given tree, which we call the "tree simplification procedure," without changing its topological information. Our motivation for developing this procedure was to reduce computational costs of graph Laplacian eigenvalues of such trees. When we applied this procedure to a set of trees representing dendritic structures of retinal ganglion cells of a mouse and computed their graph Laplacian eigenvalues, we observed two "plateaux" (i.e., two sets of multiple eigenvalues) in the eigenvalue distribution of each such simplified tree. In this article, after describing our tree simplification procedure, we analyze why such eigenvalue plateaux occur in a simplified tree, and explain such plateaux can occur in a more general graph if it satisfies a certain condition, identify these two eigenvalues specifically as well as the lower bound to their multiplicity.