Suggestion for using global optimization

I'm using global optimization toolbox (both simulated annealing and genetic algorithm) to perform an optimization. But the change happening in the objective function after each iteration is minuscule and the result that I'm getting is unsatisfactory. I must mention that the number of design variables is 66. Can you suggest particular changes that need to be done in the "option" structure of these tools to make them suitable for large number of design variables. Or is there something else that I'm missing out? I feel that the tool is unable to make significant alteration in the design variables in each iteration.

Answer 1

[Alan Weiss]

My best suggestion is to try using patternsearch instead of ga or simulannealbnd. patternsearch is easier to tune, and is almost always faster and more reliable than the other two solvers. For example, it is easy to set an initial mesh size in patternsearch, whereas simulannealbnd has no such tuning option, and it is more involved to set the initial range for ga.

Good luck,

Alan Weiss

MATLAB mathematical toolbox documentation