lsqnonlin- multivalued problem

Hi ,

I am finding some problems with the function lsnoleq, it is multivalued and I am doing a project in which I have to compare transmission and reflection data with theoretically obtained transmission and reflection from generated permittivity values. I use lsnoleq to compare the generated permittivity values with the experimentally obtained data and find the value that least differs from the theoretical and experimental values and fix that permittivity value as the obtained permittivity.

The problem with the lsnoleq is that it gives me different values for permittivity for different starting points and different lower and upper bounds.

The lsnoleq also seem to oscillate between the two upper and lower bounds.

 options = optimset('MaxFunEvals', 5000, 'TolX', 1e-20, 'TolFun', 1e-20, 'MaxIter', 5000, 'LargeScale', 'on');

    ep_new =lsqnonlin(@t_th_0_2lay,[2 0], [1.5 0],[2.3 .2],options);

t_th_0_2lay is the function which calculates the transmission and reflection data from the generated values of permittivity between 1.5 and 2.3 for the real part and 0 and .2 for the imaginary part.

for example of we give the starting point as [1 0] it comes closer to 1 if we give [2 0] it goes near 2.

is there a better function to calculate the minimum (with real and imaginary inputs of course!)

I tried to use fminsearch but it has some problems with the input being imaginary . please help me with this!

Thomas