MATLAB: Fmincon – gradient causes trouble

fminconintegrationoptimization

Hi,
I have a rather complicated Optimization Problem. The function which I want to minimize consists of a big double integral. Computing the double integral is quite hard and takes time, but computing the gradient (which also consists of a big double integral) is even harder. I computed the gradient by hand and plugged it into fmincon using the interior-point and trust-region-reflective algorithm. It still gives me warnings that it was unable to compute the integral satisfying the tolerance parameter. Decreasing the parameter influenced the result too much.
Probably this is more an integration problem but since evaluating the function is possible, I wonder if there is a minimizer in MATLAB which does not need derivatives for a constrained, multivariable and non-linear problem.
Thanks in advance!

Best Answer

The genetic algorithm , ga(), in the Global Optimization Toolbox does not use derivatives.