1
$\begingroup$

I have a multi-objective optimization problem (NP-complete). To solve it, I've decided to use the genetic algorithm.

I have 3 "areas" of optimization, say parameters a, b and c, so cost function for my problem is C(F) = a + b + c -> min.

I want to somewhat modify the genetic algorithm. In the classic workflow, the initial population is created randomly. For my problem, I want to create a "smart" initial population. Since I have 3 parameters to be optimized, I wanna create 1/3 of the initial population optimized for parameter a, 1/3 optimized for parameter b, and 1/3 optimized for parameter c. Then, in the while loop just "mix" these good individuals. Is this a good way to go? Is this still considered the genetic algorithm? Is this a good modification?

P.S. I'm doing research for my master's project so I need to provide some novelty in my solution. Would this modification be considered fine for that?

Thanks in advance.

$\endgroup$

1 Answer 1

1
$\begingroup$

Yes, this is absolutely still considered a genetic algorithm. In fact, it is quite common to experiment with various strategies of obtaining an initial population. As you say, it is reasonable to start with a completely random population unless you have a reason to try something else.

Whether this is a good way to go is anybody's guess, really. It depends heavily on your problem and problem instances. Give it a try and see how it compares to a random initial population. Further, whether this is a modification that is considered fine, trust your advisor more than random people on the internet.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.