any tips on speeding up a ga optimization

I am running a ga optimization that takes over a week to run on a pc with 2 quad core processors (2.5ghz) with 32 gigs of ram. any general tips on speeding it up. I do a another pc with a single 3.5ghz quad core and 8 gigs of ram would this one be faster than the other one if I added more ram.

[Walter Roberson]

We do not have enough information on how your code is currently written or what your data sizes are. Your code might only be using one core most of the time at present.

Answer 1

[Alan Weiss]

My first recommendation would be to not use ga unless your problem has integer constraints and a nonlinear objective or constraints. Generally, for nonlinear objective and no integer constraint, if the problem is smooth then use fmincon, and if the problem is nonsmooth then use patternsearch as your first-choice solvers.

If you are trying to search for a global solution, then start the recommended solvers from a variety of initial points. For fmincon you can use MultiStart. If you have finite bounds on all components and usually you should), for patternsearch or fmincon you can start the solver multiple times manually from the points

 x0 = lb + rand(size(lb)).*(ub - lb);

Good luck,

Alan Weiss

MATLAB mathematical toolbox documentation