fminbnd not adjusting largest initial guess.

Hello, I am using fmincon to fit some parameters in a pdepe solver. I have noticed that if I set an initial guess for one parameter to be many orders of magnitude larger than others, it is not adjusted by fmincon.

For example, I start with the conditions [1e-3,1e-3,4e6] I will get back adjusted values for first two but always get back 4e6 for the final.

Any ideas?

Answer 1

[Alan Weiss]

You might want to rescale your problem. This is not always easy, but it is often worthwhile. Or use some of the suggestions in Optimizing a Simulation or ODE.

Good luck,

Alan Weiss

MATLAB mathematical toolbox documentation