How does GA create crossover/mutation with constraints ?

Hello. I am currently doing an optimization where polynomial (4 parameters) have to fulfil a constraint of the type (max(poly)-Value) < 0.

To do so, I use the genetic algorithm from matlab.

I have a question though, how is it possible that the ga command uses crossover/mutation functions, but still fulfil that condition ? What will the code exactly do if this condition is not fulfilled ? Skip the fitness evaluation ?

Thank you for your help

Answer 1

[Alan Weiss]

The nonlinear constraint algorithms are described here. In brief, the algorithm allows nonlinear constraints to be violated, and in subsequent iterations attempts to obtain feasible points.

Alan Weiss

MATLAB mathematical toolbox documentation

Answer 2

[Boris Huljak]

Thanks, I read your link and it is kind of still blur. When you say they trie to obtain feasible, is there a simplified explaination on how they do so ? I did noticied my solution are similar to each other, but not exactly the same, so I suspect there is some kind of variation of each of the value to be optimized ?

Anyway, thank you for the infos !

EDIT : I would like to ask something practical. So far, my optimization does converge, but all my population are pretty close to each other (most likely a local minimum), is it possible to change an option so that my population is a bit more scattered ? Something like : "If a member does not fulfill the constraint, it is deleted and a new member (generated by my initial population function) is created and replace him" ?

Thank you !