Maximum number of function evaluations has been exceeded - increase MaxFunEvals option.

Dear Matlab users: I've got a following problem: I try to find parameters of this function:

by simplex method (see the following code)

format compact
format long
xdata = [0.000,0.200,0.400,0.600,0.800,1.000,1.200,1.400,1.600,1.800,2.000,2.200,2.400,2.600,2.800,3.000,3.200,3.400,3.600,3.800,4.000,4.200,4.400,4.600,4.800,5.000];
ydata = [0.007,0.041,0.165,0.449,0.816,0.982,0.741,0.212,-0.362,-0.808,-0.975,-0.774,-0.290,0.290,0.775,0.982,0.849,0.527,0.237,0.077,0.018,0.003,0.000,0.000,0.000,0.000];

%Function to calculate the sum of residuals for our a given parameters

fun= @(p) sum(ydata - (-1*(p(1)-p(2)*(((xdata)-p(3))).^4).*(exp(-1*p(4)*((xdata)-p(3)).^2))).^2);
%starting guess for our parameters 

%starting guess
pguess = [1.0,12.0,1.0,2.0];

%optimise
[p,fminres] = fminsearch(fun,pguess)

Unfortunately it doesn't work very well because this type of message appears:

Exiting: Maximum number of function evaluations has been exceeded
         - increase MaxFunEvals option.
         Current function value: -Inf

I know that parameters, should be cca a= 1.0; b=12; c=2.4 and r=1.02 (in code, abcr parameters are p(1)...(p4)).

I need to use simplex method for iteration, so I try to fix this problem, but I am totaly newbie in matlab and don't know how to fix it.

Does anybody have any suggestion??

Thanks

Paul

Answer 1

[Matt J]

Using the optimset command you can change MaxFunEvals and other algorithm parameters of fminsearch. You can also use optimoptions if you have the Optimization Toolbox.

[Pavol Namer]

Thank you for your answer,

So you mean:

 [p,fminres] = fminsearch(fun,pguess,optimset'MaxFunEvals',1000)

And for

optimoptions,

which method for simplex algoritm do I need to use? Thank you

[Alan Weiss]

You cannot use optimoptions for fminsearch, you have to use optimset. For example,

 options = optimset('MaxFunEvals',1000);
 [p,fminres] = fminsearch(fun,pguess,options)

For more information, consult the documentation on setting options.

Alan Weiss

MATLAB mathematical toolbox documentation

Answer 2

[mura0087]

I have this same problem! Even if I change MaxFunEvals to a very high number (1e30) I still get an Inf value out of fminsearch.

[Walter Roberson]

The initial points tested all generated inf, so fminsearch() does not know which direction to look in.

You should give an initial point that has a finite value associated.

[mura0087]

I had been giving the initial points as rand(2,1) but that's clearly not cutting it.

I am using fminsearch to minimize the error function of an implicit equation to my dataset. Some more information on what I did for that- is here. The fit works fine on data that I generated using the implicit function (sanity check), I can obtain values for the coefficients, but will NOT fit my experimental data.

I have not ruled out that my experimental data does not fit, or requires some sort of transformation. I do know, based on messing around with generating data with different coefficients, that the fminsearch is fairly sensitive to the start points I give it. Sometimes I need to give it values close to where I know the coefficients are.

Problem is, for the experimental data, I expect one of the coeffs to be between 0 and 1 (can make a reasonable guess here), but the other coeff needs be above 0 (too many options for me to get a good guess).

I am going to revisit the model to see if I can get a better understanding of what my coeffs should be, and set better guesses. Or loop over many, many guesses

[Walter Roberson]

In the File Exchange, John D'Errico has a bounded version of fminsearch.

[mura0087]

awesome, thank you!