Optimization of Options for Genetic Algorithm

I want to find the optimal configuration of options for my application of Genetic Algorithm.

I have successfully used GA for my application but now I want to find the optimal combination of options to improve my final objective function value.

For example: Population size: 200 or 100 or 50

Fitness Scaling Options: fitscalingrank' or 'fitscalingprop' or 'fitscalingtop'

The combination of what options would give me the best result. In short I want to optimize the parameters for GA.

If it is only number like population size I think it would be easier but since we have different selection options, fitness scalin options etc. I have no idea how to do it.

If anyone has an idea please let me know :)

Thanks!

[Moritz Wegner]

I have the same problem and would be really glad if anybody knew a way of doing that. Have you found out anything since Vishnu?

Answer 1

[Umang Pandey]

Hi Vishnu,

There are no single optimal parametric settings that can be universally applied to all sets of optimization problems. The nature of the optimization problem at hand, number of objectives, characteristics of objective function, nature of the constraints, and application (in terms of exploration v/s exploitation preference) all these factors determine the optimal population size, mutation and crossover function, scaling options, etc.

It is advisable to refer to the following documentation after you have identified the nature of the optimization problem you want to solve to be able to determine the optimal parameter settings:

https://www.mathworks.com/help/gads/genetic-algorithm-options.html

The documentation lists the nature of the problem and constraints each parametric setting is suitable for.

If you are unsure about the optimization problem, try using two or three combination of settings which you feel will be best suited after going through the documentation and finalize upon and further modify the settings for the one giving the best convergence.