Help with 'fminsearch'

Hi

I am fitting an exponential to my data using least squares, as in:

************************************************************

t = 0:7 ;
rel = [629 537 460 375 334 286 249 227];

fh = @(x,p) p(1) + p(2)*exp(-x./p(3))

errfh = @(p,x,y) sum((y(:)-fh(x(:),p)).^2)

p0 = [mean(rel) (max(rel)-min(rel)) (max(t) - min(t))/2];

P = fminsearch(errfh,p0,[],t,rel)

plot(t,rel,'bo',t,fh(t,P),'r-')

************************************************************

However, now I wish to increase MaxFunEvals. How can I do this, when I use fminsearch in this fashion?

Best, Niles.

[Sargondjani]

wow, i have never seen this: you supply the x (t) and y (rel) after the options. i can't find it in the documentation either

where did you learn this??

[Niles Martinsen]

I learned it here: http://www.mathworks.com/matlabcentral/newsreader/view_thread/169601

[Sargondjani]

hmmm... not much documentation there either, haha. it's interesting... i really wonder if this also works for other optimization algoritms. it would be very nice if you could pass the parameters after the optimset

[Oleg Komarov]

Usually oprimization routines all have a varargin input. Edit fminsearch or fminbnd.

[Walter Roberson]

In 2008, Jos sometimes made mistakes.

Answer 1

[Oleg Komarov]

P = fminsearch(errfh,p0,optimset('MaxFunEvals',1000),t,rel)

Check the documentation for optimset

[Niles Martinsen]

I tried this out, but it did not have the desired effect unfortunately.

[Oleg Komarov]

What do you mean it did not have the desired effect?

Try to reduce it to 10 or 50. Keep in mind it's a local solver.

Answer 2

[Walter Roberson]

You cannot increase MaxFunEvals when you use fminsearch in that fashion, as fminsearch takes a maximum of 3 inputs, not the 5 you are using. The function handle you provide to fminsearch must take exactly one argument. The initial guess follows the function handle, and the options structure (if provided) follows that.

http://www.mathworks.com/help/techdoc/ref/fminsearch.html

[Niles Martinsen]

Thanks, I'll try it out.

Best,

Niles.

[Oleg Komarov]

@Walter: I don't get it.

Isn't fminsearch treating x (t) and y (rel) as 'data' and optimizing wrt to p?

[Niles Martinsen]

Walter, actually when I think it over I am not sure I understand what you mean.

[Walter Roberson]

Guessing about what you are trying to pass to what:

P = fminsearch(@(p) errfh(p,t,rel), p0, optimset('MaxFunEvals',1000));