initial point is a local minimum

Dear All,

I am trying to fit a set of experimental data to a model and obtain 2 parameters. I am using "lsqcurvefit" function. I used this function for three sets of experimental data and obtain answers but for the other sets I got the below error " Initial point is a local minimum.

Optimization completed because the size of the gradient at the initial point is less than the default value of the function tolerance"

as I tried to change the initial points but it does not have any effect and it always says that the initial points are local minimum.

Can anybody help me in this matter?

    function Y = myfun_D(x,F)
    Y = x(1)*exp(-x(2)*F);

    "x0=[5e-10,.6]; % Starting guess %
    [x,resnorm] = lsqcurvefit(@myfun_D,x0,F,D)

F and D are set of data.

Best Regards Gh

Answer 1

[Matt J]

Since there are only 2 unknowns, you should be able to make a surf plot of the resnorm function to see approximately where the minimum will be. You can then use that as your initial guess for lsqcurvefit.

I suspect that you are initializing nowhere near the correct value of x(2). If x(2)*F contain very large values, the exponential on the right hand side of

   Y = x(1)*exp(-x(2)*F);

will just underflow to zero and your resnorm function will look locally flat.