FirstIndexLast

[help-gsl] Nelder and Mead (2008-09-19)

From: Mario dos Reis <mdosrei@..............>
Subject: Nelder and Mead
Date: Fri, 19 Sep 2008 11:22:43 +0100
To: regimbau@......
Cc: help-gsl@gnu.org

Did you optimize your code using Nelder and Mead? i.e.

gcc myprogram myprogram.c -O4 -DHAVE_INLINE=1 -DGSL_RANGE_CHECK=0 \
-lgslcblas -lgsl

Message: 3 Date: Wed, 17 Sep 2008 18:40:05 +0100 From: Brian Gough <bjg@gnu.org> Subject: Re: [Help-gsl] Nelder-Mead Simplex algorithm To: Tania Regimbau <regimbau@......> Cc: help-gsl@gnu.org Message-ID: <m3myi6llhm.wl%bjg@network-theory.co.uk> Content-Type: text/plain; charset=US-ASCII At Fri, 12 Sep 2008 20:15:10 +0200, Tania Regimbau wrote:
> > I used the implementation of the GSL Nelder Mead Simplex algorithm to
> > minimize a 3 variable function and it doesn't seem to perform very
> > well compared to the Matlab function fmeansearch based on the same
> > algorithm. Can someone tell me what the difference is between the two?
> > I thought I was using the same precision parameters but I may have
> > missed something.

I'm not aware of any major difference -- could you submit a bug report
to bug-gsl@gnu.org with an example program and the output compared
with MATLAB. Thanks.

-- Brian Gough