[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Bug-gsl] linalg/bidiag.c
From: |
Brian Gough |
Subject: |
Re: [Bug-gsl] linalg/bidiag.c |
Date: |
Fri, 30 May 2003 15:35:26 +0100 |
Mario Pernici writes:
> As the example shows, the orthogonal matrix U obtained from
> gsl_linalg_bidiag_unpack is a M x N matrix (linalg/bidiag.c line
> 169) satisfying U^T U = I, but not U U^T = I, which is different
> from the M x M orthogonal householder matrix U obtained from
> sl_linalg_bidiag_decomp; V is the same in both cases; in both cases
> one gets a bidiagonal matrix, but they differ (B has an extra row
> of zeroes).
> In both cases B = U^T A V and A = U B V^T . The
> gsl_linalg_bidiag_unpack results seem to be fine, but maybe it
> could be mentioned in the manual that the two U's and B's are
> different in the two functions.
Thanks -- now mentioned in the manual.
The compact MxN form of the matrix is used for efficiency (important
for M>>N).
Brian