A parallel algorithm for the diagonalization of symmetric matrices

Author

Cernuschi-Frias, Bruno ; Lew, Sergio E. ; Lez, Hema N J Gonza ; Pfefferman, Jonas D.

Author_Institution

Fac. de Ingenieria, Buenos Aires Univ., Argentina

Volume

5

fYear

2000

fDate

2000

Firstpage

81

Abstract

A parallel algorithm for the diagonalization of symmetric matrices is presented. The Givens-Jacobi rotator method is extended and modified to solve the eigensystem problem of symmetric matrices in a full parallel way. The algorithm solves the diagonalization of symmetric matrices in approximately N “parallel” iterations for large N, while the Givens-Jacobi algorithm requires 3N to 5N “parallel” iterations. A proof of the convergence for small rotation angles is presented. Preliminary simulations done with arbitrary randomly generated symmetric matrices sustain the efficiency of the algorithm

Keywords

convergence of numerical methods; eigenvalues and eigenfunctions; mathematics computing; matrix algebra; parallel algorithms; Givens-Jacobi rotator method; convergence; diagonalization; eigensystem problem; parallel algorithm; symmetric matrices; Application software; Automatic control; Computer vision; Convergence; Covariance matrix; Image coding; Iterative algorithms; Jacobian matrices; Parallel algorithms; Symmetric matrices;

fLanguage

English

Publisher

ieee

Conference_Titel

Circuits and Systems, 2000. Proceedings. ISCAS 2000 Geneva. The 2000 IEEE International Symposium on

Conference_Location

Geneva

Print_ISBN

0-7803-5482-6

Type

conf

DOI

10.1109/ISCAS.2000.857368

Filename

857368