x = lsqcurvefi​t(fun,x0,x​data,ydata​) : which option allows me to change step size that varies x0?

Hi,

I'm using 'lsqcurvefit' to fit sets of parameters 'x0'. My problem is whatever value 'x0' that I start with, I always get a local minima that is very close to my initial 'x0'. I think one way to solve this problem is to change the step size that the program vary for x0 each iteration. Could anyone tell me which option allows me to change this?

Also, any other comments helps.

Thanks a lot!

Answer 1

[Matt J]

There is no parameter that lets you control the step-size.

If you are using finite differences to approximate the derivatives (the default), you could instead try modifying DiffMinChange, but I wouldn't count on it as a reliable solution. The best thing would be to try to understand why you have so many local minima. If you really do have so many, there is something pathological about your objective function.