|
From: | Rafael Gonzalez |
Subject: | Re: [ESA-SoC] Generalized eigenvalue |
Date: | Wed, 15 Apr 2015 16:44:50 -0500 |
The matrix
B = [ 3 -2 -.9 2*eps -2 4 1 -eps -eps/4 eps/2 -1 0 -.5 -.5 .1 1 ];
has elements on the order of roundoff error. It is an example for which the nobalance option is necessary to compute the eigenvectors correctly. Try the statements
[VB,DB] = eig(B) B*VB - VB*DB [VN,DN] = eig(B,'nobalance') B*VN - VN*DN
---
Case | Routine |
Real symmetric A | DSYEV |
Real nonsymmetric A: | |
With preliminary balance step | DGEEV (with SCLFAC = 2 instead of 8 in DGEBAL) |
d = eig(A,'nobalance') | DGEHRD, DHSEQR |
[V,D] = eig(A,'nobalance') | DGEHRD, DORGHR, DHSEQR, DTREVC |
Hermitian A | ZHEEV |
Non-Hermitian A: | |
With preliminary balance step | ZGEEV (with SCLFAC = 2 instead of 8 in ZGEBAL) |
d = eig(A,'nobalance') | ZGEHRD, ZHSEQR |
[V,D] = eig(A,'nobalance') | ZGEHRD, ZUNGHR, ZHSEQR, ZTREVC |
Real symmetric A, symmetric positive definite B. |
DSYGV |
Special case: eig(A,B,'qz') for real A, B (same as real nonsymmetric A, real general B) |
DGGEV |
Real nonsymmetric A, real general B | DGGEV |
Complex Hermitian A, Hermitian positive definite B. |
ZHEGV |
Special case: eig(A,B,'qz') for complex A or B (same as complex non-Hermitian A, complex B) |
ZGGEV |
Complex non-Hermitian A, complex B | ZGGEV |
Hi,
I saw the project ideas list of ESA Summer of Code, and saw the generalized
eigenvalue problem, however I didn't actually get, what should be done with
it. I read this through:
http://octave.1599824.n4.nabble.com/General-eigenvalue-problem-proposal-td4651990.html
but so far I saw, the generalized eigenvector problem is done, just
balancing isn't in the eig() function. Am I right or maybe I've
misunderstood something? Please, could you explain me, is this the main
focus of this project, or what else should be done with eig()?
Regards,
András Mihálykó
--
View this message in context: http://octave.1599824.n4.nabble.com/ESA-SoC-Generalized-eigenvalue-tp4669870.html
Sent from the Octave - Maintainers mailing list archive at Nabble.com.
[Prev in Thread] | Current Thread | [Next in Thread] |