[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Bug-gsl] Minor bug in SVD jacobi
From: |
Jari Häkkinen |
Subject: |
[Bug-gsl] Minor bug in SVD jacobi |
Date: |
Tue, 01 Feb 2005 01:01:53 +0100 |
User-agent: |
Mozilla Thunderbird 1.0 (Macintosh/20041206) |
Hi.
I am submitting a bug report followed by two suggestions for the SVD
part of the GSL manual.
i) I found a minor error in the gsl_linalg_SV_decomp_jacobi routine. A
mix up of matrices has been implemented in the check that the the number
of columns is smaller than the number of rows in the matrix to become
decomposed. The code says (line 441 latest CVS):
if (Q->size1 < Q->size2)
{
/* FIXME: only implemented M>=N case so far */
GSL_ERROR ("svd of MxN matrix, M<N, is not implemented",
GSL_EUNIMPL);
}
else if (Q->size1 != A->size2)
...
The Q's in the first line above should be A's.
ii) It would be nice if the GSL manual would clearly state that the SVD
algorithms are implemented for rows>=columns case only.
iii) The documentation for the function gsl_linalg_SV_decomp_mod should
clearly state the size of the vector 'work'. As it is now you have to
deduce it from the gsl_linalg_SV_decomp documentation or read the source
code.
Cheers,
Jari
- [Bug-gsl] Minor bug in SVD jacobi,
Jari Häkkinen <=