It seems that you are already aware of a solution i.e. gene values desired so then why optimization is required. Please share more details about your problem.

One possible way is to reduce the probability of mutation. Elite chromosomes can never be replaced so there is no need to manipulate any chromosome. Furthermore, introducing diversity in population by checking probability on each gene is good for better exploration.

You can even make a hybrid version by guiding the optimization provided a favorable fitness is achieved.

Thanks

1. Select & keep a certain percentage of elite solutions from the sorted population in each generation.

2. Following the suggestion of Rohail Gulbaz try to play with the mutation probability (maybe between 0.01 to 0.1 (max)).

3. You may also try to vary the crossover point and type.

In general, the worse genes should be automatically eliminated if you are using an elitist-GA version. So, there should not be any need of deleting additional genes from the population pool.

Again, if your application does not explicitly require the global optima, you should be satisfied with what you have got. However, for further improvement, one way is to go for hybridization as suggested by Rohail Gulbaz

It would be great if you could provide further information about your GA-based method. But in general, if you know which characteristics yields undesirable results, you can exclude those gens. There is also another approach to this problem. You can let the undesirable solutions to be generated and just sort them and have them at the bottom of your list. Then, you can define a fitness function that allows selection of parents from the top generated solutions. Furthermore, keeping the mal solution maybe beneficial, as your design objectives may change and those mal solutions become a desirable one. Also, keeping the mal solution is always informative for the designer/user.

Rohail Gulbaz

Thanks Rohail for answering my question. You're right about your skepticism of needing an optimization for such a problem.

Here's the full image, I have a problem where I need to optimize multi objectives. The "optimal" objectives can't be found simply due to the chromosome's high dimensions. Furthermore, my problem can't be divided into sub-problems since the genes that build up the whole solution is 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.

Hope that I have succeeded in explaining the idea.

I think, I will go for Hybridization. Thank you again

You can try Bacterial Evolutionary Algorithm. That algorithm does not destroy the good genes.

@Bushr welcome

All the genes play an important rolein in the optimization process, The likely good genes contribute information for improving the qulity of new ones and the likely bad genes could help to avoid non factible solutions and to re-direct the search to new areas of the domain. So, I think that all the genes must to participate in the solution search process.