This is the mail archive of the
gsl-discuss@sources.redhat.com
mailing list for the GSL project.
Re: Conjugate Gradient Minimizer difficulties.
- From: Brian Gough <bjg at network-theory dot co dot uk>
- To: Britt Park <britt at sciencething dot org>
- Cc: <gsl-discuss at sources dot redhat dot com>
- Date: Wed, 29 May 2002 19:08:29 +0100 (BST)
- Subject: Re: Conjugate Gradient Minimizer difficulties.
- References: <B9198D3D.53A%britt@sciencething.org>
Britt Park writes:
> First apologies in advance if this turns out to be a stupid
> question/comment.
>
> I've been trying to use Fletcher/Reeves and Polak/Ribierre conjugate
> gradient minimization in my molecular mechanics code. Both work well up to
> about 100 to 200 iteration on a box of water molecules, and then seem to go
> into an endless loop. Steepest descent works well and has never choked.
> I've checked my derivatives numerically and they are correct. Any help (as
> to parameters or anything else) would be most appreciated.
>
Have you tried calling the gsl_multimin_..._restart() method? I
believe those routines lose accuracy over time need to be restarted.
regards
Brian