This is the mail archive of the
gsl-discuss@sources.redhat.com
mailing list for the GSL project.
Re: minimization using the Brent algorithm (brent.c)
- From: Michael Na Li <lina at u dot washington dot edu>
- To: Z F <mail4me9999 at yahoo dot com>
- Cc: Fabrice Rossi <rossi at ufrmd dot dauphine dot fr>, Brian Gough <bjg at network-theory dot co dot uk>, Faheem Mitha <faheem at email dot unc dot edu>, gsl-discuss at sources dot redhat dot com
- Date: Mon, 17 Mar 2003 12:47:26 -0800
- Subject: Re: minimization using the Brent algorithm (brent.c)
- References: <20030317200852.54309.qmail@web20419.mail.yahoo.com>
On Mon, 17 Mar 2003, Z. F. stated:
>
> > Moreover, evaluation of F(x) is required for the Brent algorithm (or
> > the
> > golden section search), so there is no useless evaluation.
>
> The x, required for golden/Brent is not just any x, it is a particular
> place inside the interval (a;b). So the function call is useless.
I concur that requiring the user to specify x isn't very helpful. As in most
cases, I don't really have a good guess of x. So when I use gsl's Brent
algorithm, I do a couple of plain golden-section-like iterations to find the x
(and possibly shrink the initial interval during the process). I use my own
C++ wrapper class anyway so I don't have to change the implementation of
brent.c.
Michael