This is the mail archive of the
gsl-discuss@sourceware.org
mailing list for the GSL project.
Re: Toeplitz solver
- From: Z F <mail4me9999 at yahoo dot com>
- To: Tommy Nordgren <tommy dot nordgren at comhem dot se>
- Cc: gsl-discuss at sources dot redhat dot com
- Date: Thu, 1 Nov 2007 18:40:32 -0700 (PDT)
- Subject: Re: Toeplitz solver
--- Tommy Nordgren <tommy.nordgren@comhem.se> wrote:
>
> On 1 nov 2007, at 06.06, Z F wrote:
>
> > Hello everybody,
> >
> > Could someone suggest a (comprehendible) reference on Toeplitz
> > equations
> > solver for a general Toeplitz matrix. I am interested in fast
> methods
> > :)
> >
> > Numerical recepies has a mistake in the ALGORITHM, and in the code,
> of
> > course. (The third edition has this old bug also). It addresses
> vector
> > indecies which are out of range. (look at eq 2.8.23 and 2.824,
> terms
> > with R_{m+1} and R_{-m-1} are out of range of the vector size)
> >
> > And as to myself, I can not seem to trace were the problem happens.
> >
> > Thanks a lot,
> >
> > ZF
> >
> Numerical Recipes adds an offset to pointers returned from
> allocation routines,
> so that the useful lower index can be other than zero.
THanks for the comment, I knew about that.
I also know what the problem was. The formulae are correct, but the
description of the iteration process is wrong. I have changed it and
it is going.. (If you want to know, NR says first find G and H and then
find X, where is one has to first find X and then G and H. See the
section on that in NR)
The question remains of there is a reference for a clear description
of N log N method to solve Toeplitz equations.
Thanks..
ZF
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com