This is the mail archive of the
gsl-discuss@sources.redhat.com
mailing list for the GSL project.
Re: gsl-discuss Digest 6 Oct 2003 06:27:59 -0000 Issue 879
- From: Reinhold Bader <Reinhold dot Bader at lrz-muenchen dot de>
- To: wiedemann dot harald at litef dot de
- Cc: gsl-discuss at sources dot redhat dot com
- Date: Mon, 06 Oct 2003 10:02:07 +0200
- Subject: Re: gsl-discuss Digest 6 Oct 2003 06:27:59 -0000 Issue 879
- References: <1065421679.10920.ezmlm@sources.redhat.com>
gsl-discuss-digest-help@sources.redhat.com wrote:
Subject:
generalized eigenvalue problem
From:
Harald Wiedemann <wiedemann.harald@litef.de>
Date:
Mon, 6 Oct 2003 07:27:05 +0200
To:
gsl-discuss@sources.redhat.com
Hello everybody,
what is the easiest way to calculate the eigenvectors/eigenvalues
of the generalized eigenvalue problem
A x = alpha B x
A, B = real symmetric matrices
alpha = eigenvalue
x = eigenvector?
I'd recommend using the DSYGVD routine from LAPACK. The divide and
conquer algorithm may have trouble with numerical stability sometimes,
so you might want to check against the results from DSYGV if
there are problems.
Note that B needs to be positive definite in addition to being
symmetric in the above routines since the algorithm attempts
to calculate the matrix square root of B.
If you want to call from C, you might want to use the clapack package
(f2c'ed from Fortran).
LAPACK and clapack are available from http://www.netlib.org
Thanks in advance,
Harald Wiedemann
--
Dr. Reinhold Bader
Leibniz-Rechenzentrum, Abt. Hochleistungssysteme | Tel. +49 89 289 28825
Barerstr. 21, 80333 Muenchen | email Bader@lrz.de