Re: fminserch does not converge for every variable

yaoli <yaoli54@xxxxxxxxx> wrote in message
<7287242.1214850166976.JavaMail.jakarta@xxxxxxxxxxxxxxxxxxxxxx>...
Thanks for the reply, Scott.

My tusk is genetic mapping. So of the 17 parameters, 12 of them are
actually 4 parameter for 3 different genotypes, therefore, the number of
parameters can not be reduced.

The thing is, my colleagues have been using the same method applied to
similar questions with about the same number of parameters, and fminsearch
has always been doing a great job for them. So I'm wondering, is this because
that my function is a little bit too weird.

Anyway, I think the sensitivity analysis is a good advise to start with...

If your colleagues are using a Nelder/Meade simplex
to solve problems with 17 parameters, they are probably
kidding themselves that it has converged to a valid
solution, certainly so in general. In addition, they are
wasting huge amounts of time, since these polytope
methods are ridiculously slow for that many parameters.
But hey, it is their time they are wasting.

Next, are you telling us that four of your parameters
are discrete parameters? This is certainly an improper
use of this tool, which is not designed to solve discrete
problems.

John
.

Relevant Pages

  • Re: Useful maths links
    ... colleagues and not primary then I would prefer to know before I start ... wasting my time looking to see what you and others consider to be ... you just lost me altogether. ... discard. ...
    (uk.education.teachers)
  • Re: Useful maths links
    ... colleagues and not primary then I would prefer to know before I start ... wasting my time looking to see what you and others consider to be ... When you get to the site it says that it is aimed at Secondary level. ... Ellie ...
    (uk.education.teachers)