HSL_MC73 : a fast multilevel Fiedler and profile reduction code

ePubs

The open archive for STFC research publications

Persistent URL	http://purl.org/net/epubs/work/54291
Record Status	Checked
Record Id	54291
Title	HSL_MC73 : a fast multilevel Fiedler and profile reduction code
Contributors	Y Hu (Woflram Research, Inc.), JA Scott (CCLRC Rutherford Appleton Lab.)
Abstract	In recent years, multilevel algorithms have been used for the efficient computation of the eigenvector corresponding to the smallest positive eigenvalue of the Laplacian matrix associated with a graph of a symmetric matrix (the Fiedler vector). Multilevel algorithms have also been proposed for computing profile-reducing orderings for sparse symmetric matrices. In this paper, these multilevel algorithms are described within a unified framework. This is then used in the design of a new Fortran 95 code HSL_MC73 that implements a multilevel algorithm for the computation of an approximate Fiedler vector as well as a number of multilevel profile-reducing algorithms. HSL_MC73 is used to compute spectral orderings for a class of undirected random graphs and its performance is compared with obtaining the Fiedler vector using a state-of-the-art sparse eigensolver.
Organisation	CCLRC , CSE , CSE-NAG
Keywords
Funding Information
Related Research Object(s):
Licence Information:
Language	English (EN)

Type	Details	URI(s)	Local file(s)	Year
Report	RAL Technical Reports RAl-TR-2003-036. 2003.		RAL-TR-2003-036.pdf	2003