Wave equation based algorithm for distributed eigenvector computation

We propose a novel distributed algorithm to compute eigenvectors and eigenvalues of the graph Laplacian matrix L. We prove that, by propagating waves through the graph, a local fast Fourier transform yields the local component of every eigenvector of L. For large graphs, the proposed algorithm is or...

Full description

Saved in:
Bibliographic Details
Main Authors: Sahai, Tuhin, Speranzon, Alberto, Banaszuk, A
Format: Conference Proceeding
Language:English
Subjects:
Clustering algorithms
Convergence
Eigenvalues and eigenfunctions
Equations
Laplace equations
Markov processes
Propagation
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We propose a novel distributed algorithm to compute eigenvectors and eigenvalues of the graph Laplacian matrix L. We prove that, by propagating waves through the graph, a local fast Fourier transform yields the local component of every eigenvector of L. For large graphs, the proposed algorithm is orders of magnitude faster than random walk based approaches. We prove the equivalence of the proposed algorithm to eigenvector computation and derive convergence rates. We also demonstrate its utility on a distributed estimation example.
ISSN:0191-2216
DOI:10.1109/CDC.2010.5717290