Is it right to exclude (freeze) some genes from genetic operations?

Grasshopper Developer
1

Dear All,

I have successfully applied NSGA-II on a multi-objective optimization problem.

While I was observing the results during the optimization process, I have noticed that some variables (genes) have reached very good values that matches my objectives, while others haven’t. However during the optimization, the good genes are being frequently changed and swapped to undesired results (due to genetic operations; mutation and crossover) until the algorithm reaches a good local optima, or hits a condition.

My question is this:

Can I exclude some genes from the optimization process since they have already reached my condition sand finally combine them with the remained genes in the final chromosome?
@johnharding

2

That’s called Elitism. You basically put aside a proportion of the population (say 5%) before any crossover or mutation takes place. Not sure what library you’re using, but it should include the option.

3

Thanks for answering John,
Actually, Elitism is not exactly what I am trying to do. My idea is to exclude some parts of the chromosome/solution (i.e. genes inside the chromosome that build up the whole solution).

4

Oh, I see. Never heard of that, apologies. Usually the phenotype is assessed as a whole whenever I’ve worked with GAs myself, rather than it even being possible to extrapolate parts of a chromosome like that and judge them to be good or bad compared to other parts of the chromosome. Maybe someone on here can help you though! Otherwise, if you really can split your chromosome like that could you not split your problem into sub-problems that you co-evolve?

5

That’s right. Optimizing sub-chromosomes is equivalent to optimizing sub-problems in case genes are not related to each other. However, in some cases, inter-related genes can be optimized better if they can evolve together and then split into different sub-problems.
Thank you again

1 Like
6

Ok, then you need some sort of sub-evaluation scheme implemented I guess before the chromosomes are combined. Interesting. Nope, never heard of such a thing - a few hours on google awaits…

1 Like