FindFit NMinimize and Nelder-Mead Optimization
- To: mathgroup at smc.vnet.net
- Subject: [mg80218] FindFit NMinimize and Nelder-Mead Optimization
- From: Lyle Gordon <lgordon at gmail.com>
- Date: Wed, 15 Aug 2007 04:12:37 -0400 (EDT)
I am running a non-linear curve fit using constrained global
optimization: FindFit and the NMinimize method NelderMead. After
running for a few minutes an error pops up saying that there was an
invalid comparison with some number, 0.064 for example. It appears
that for some reason the first of the variables (k) that I am curve-
fitting to determine has been set to a number. I'm not sure if this is
a bug in Mathematica or in my implementation. This error pops up a few
more times but the progress continues and the error is supressed but
the final curve fit does not seem to be ideal. Any suggestions would
be great.
Thanks very much,
Lyle Gordon
Orthopaedic Biomechanics Lab
Sunnybrook Health Sciences Centre
Toronto, Ontario.
Here is the sample of the code I am using, sorry for the \[Tau]1 and \
[Tau]2:
fit = FindFit[
fakedata, {functiontofit[k, c, \[Tau]1, \[Tau]2, treal],
0.1 > k > 0 && 100 > c > 0 && 0.1 > \[Tau]1 > 0 &&
0.1 > \[Tau]2 > 0 && \[Tau]2 > \[Tau]1}, {{k, 0.005}, {c,
20}, {\[Tau]1, 0.01}, {\[Tau]2, 0.02}}, treal,
EvaluationMonitor :> {Print["E: ", k, ", ", c, ", ", \[Tau]1,
", ", \[Tau]2]; Clear[k]},
Method -> {"NMinimize", Method -> "NelderMead"}]