Are there any way to speed up the nonlcon satisfying for GA with intcon ? thanks

Hello, I have an optimization problem with few hundreds of variables, half of them as integers, I have nonlinear constraints, but I couldn't use fmincon or patternsearch as I have intcon, so I stuck with GA,

the nonlinear constraints takes a lot of time and generations to be satisfied (200 to 300 generations, and increases as the population size increases) giving decreasing non realistic values for fitness function, and then the fitness function value jumps to the realistic range and the minimization process takes place,

my question is: Are there any way to speed up the nonlcon satisfying ? thanks

Answer 1

[Matt J]

If you have a semi-accurate idea of where the solution is supposed to lie, you could set the Initial Population Range so that the initial population is bounded to some region that you think is a smart initial region to search,

http://www.mathworks.com/help/gads/genetic-algorithm-options.html#f14223

If you are very confident in those bounds, you should, of course, add them as constraints using the LB,UB input arguments to ga().

[Mamdouh]

thanks, Matt, for your reply I have a bad history with assigning such initial population in another problem, it retarded the convergence (although it was reasonable) or you can say that I hate guessing :), but actually, I did't try this with the current problem, but it looks like a good idea to fix my problem with some effort in proper guessing so I will/should try and see, thanks again, Regards :)