Improve initial values lsqcurvefit

I would like to improve the initial value guess x0 for lsqcurve fit. I need to be able to do it automatically without changing the values in the script. Is there a way to do this?

I need to be able to fit multiple curves so its not practical to change the initial values by hand. My code looks like

%%----------------EXAMPLE CODE
b0=[mean(data),data(1,1),0,1,1,1];      %<------WANT TO AUTOMATE THIS LINE TO IMPROVE FIT, SPECIFICALLY THE LAST 4 NUMBERS
fun_model=@(b,time)b(1)+b(2)*exp((b(3)*(1+b(4)...
              *(log(time)).^(-b(5)))).^(2));
fit=lsqcurvefit(fun_model,b0,time,data,lb,ub,'options')

%%----------------

Answer 1

[Alan Weiss]

If you have Global Optimization Toolbox, you can use MultiStart to generate multiple initial points randomly within finite bounds, as in this example.

It is also not hard to generate such points yourself:

 x0 = lb + rand(size(lb)).*(ub - lb);

Good luck,

Alan Weiss

MATLAB mathematical toolbox documentation

[Matthew]

So I have started trying to use Multistart and I get this error

 Error using CustomStartPointSet (line 78)
  A single input can be a CustomStartPointSet or a real matrix containing
at least one start point.

Not sure whats happening

[Alan Weiss]

It appears that you are trying to use an invalid custom start point set. Either show us how you are generating and using the custom start points, or do not use custom start points.

Alan Weiss

MATLAB mathematical toolbox documentation