If the matrix is square we call the "solve" routine based on an LU
decomposition of the matrix (dgetrf and dgetrs). If LAPACK flags the
result as being doubtful with the info flag then we fall back on
"lssolve" that uses an SVD (dgelss).
According to Joerg Frochte <address@hidden> (on 01/26/04):
The result in octave is quite OK, but the result my own program
computes
is very unpleasing.
ans = [ -7.12e+11-3.70e-05-6.03e+13 1.82e-03 1.74e+00-1.70e+15
1.48e-01
2.79e+00 6.10e-02 ]
This has something to do with the fact that A is very close to be
singular.
octave:21> cond(A)
ans = 2.9251e+18
Nevertheless, the result of octave is more pleasing.
How does octave deal with such a situation?
I have the source-code but I am unable to find a point to start my
analysis,
because I do not know how octave is designed.
Could you tell me how octave works in this situation or give me a
file and
line where to start in the octave code?
Thanks very much,
Joerg Frochte
-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.
Octave's home on the web: http://www.octave.org
How to fund new projects: http://www.octave.org/funding.html
Subscription information: http://www.octave.org/archive.html
-------------------------------------------------------------
--
David Bateman address@hidden
Motorola CRM +33 1 69 35 48 04 (Ph)
Parc Les Algorithmes, Commune de St Aubin +33 1 69 35 77 01 (Fax)
91193 Gif-Sur-Yvette FRANCE
The information contained in this communication has been classified as:
[x] General Business Information
[ ] Motorola Internal Use Only
[ ] Motorola Confidential Proprietary
-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.
Octave's home on the web: http://www.octave.org
How to fund new projects: http://www.octave.org/funding.html
Subscription information: http://www.octave.org/archive.html
-------------------------------------------------------------