[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Help-gsl] converting old least-squares fit code to use gsl instead
From: |
Guido De Rosa |
Subject: |
Re: [Help-gsl] converting old least-squares fit code to use gsl instead - fails miserably |
Date: |
Wed, 4 Feb 2009 11:28:17 +0100 |
Martin,
try the simplex minimizers too. I found it more powerful for certain
functions.The only drawback is that you can't compute the covariance
matrix of the theoretical parameters.
Also, some experimentation may be required to find the optimal initial
step size.
I personally followed a twofold approach for a recent small program:
simplex first to find the minimum, then Levenberg-Marquardt to get
some statistics.
As far as it can help, here's my code (sorry, no autoconf: edit the
Makefile as you need....)
http://www.studenti.unina.it/~gcderosa/scanning-tunneling/dIdVsuperc.tar.gz
What matters for the current discussion is mainly in fit.c .
For the impatient, I also shared a screencast :-)
http://youtube.com/watch?v=LLLq7-0DivI
HTH,
Guido