Re: [Help-gsl] Re: QR Decomposition for a matrix with complex values

[Jonny Taylor]

OK, it turns out the problem was my confusion over exactly how the  
complex householder transformations work. With that sorted, I can get  
it to work.

The comment in householdercomplex.c:
"To compute H'v, pass the conjugate of tau as the first argument to  
gsl_linalg_householder_hm() rather than tau itself"
was the bit I hadn't appreciated. Shouldn't this be mentioned  
somewhere in the manual? It's only by reading the source that you  
discover this pretty crucial point.

With that taken into account, I've got it working. Source file is  
attached in case it is of use to anybody. I have not included  
equivalents of some of the functions in qr.c because I did not test  
them since I don't need them. I have no idea whether it's more  
"correct" to store tau or its conjugate in the tau vector; I have  
chosen to store the conjugate.

Thanks for your help
Jonny

qr_complex.cpp
Description: Binary data

On 28 Jan 2008, at 21:07, Brian Gough wrote:

> At Thu, 24 Jan 2008 14:33:44 +0000,
> Jonny Taylor wrote:
> > I have a copy of Golub & van Loan (2nd edition) here, and as far as I
> > can see it only talks about real QR decomposition. Did you have a
> > part in mind where it talks specifically about complex decomposition?
>
> As far as I remember it just says that it's the same algorithm as the
> real case, so it doesn't describe it more than that.