How to print out the non dominant solution after each iteration while using Genetic Algorithm or NSGA II?

[ghost]

I am interested in monitoring the performance of the optimization algorithm. Hence, after every few iterations, I would like to print out the nondominant solutions from the population. How can that be achieved if I am using Genetic Algorithms or NSGA II?

Any help is much appreciated!

[dhadka]

You should be able to get the population from the algorithm, which is a list of solutions, and print them out. Similar to how you would read the result.

[ghost]

Yes I know how to do that. However, after algorithm.run(10,000) for example, I would like to get the non-dominant solution after every 10th iteration. At the end I would like to plot the list of solutions out on a graph. The x-axis should reflect 1 non dominant solution after every 10th iteration, while the y-axis should reflect the fitness of the non dominant solutions. Therefore, I would like to have something as:

for i in range(10000):
    algorithm.run(1)
    if i%10 == 0:
        print(nondominant(algorithm.population))

I know that this is not how to do it, but I hope my idea is clearer now. Thank you

[dhadka]

Ah, I see what you're saying, thanks for clarifying. So there are a few things you can do...

1. Call algorithm.step() instead of algorithm.run(). algorithm.step() performs one iteration of the algorithm.

2. The run() method also takes a callback argument that is invoked each iteration. Something like this should work:

def print_nondominant(algorithm):
    if algorithm.nfe % 100 == 0: # this assumes 100 is a multiple of your population size
        print(nondominant(algorithm.population))

algorithm.run(10000, callback=print_nondominant)

[ghost]

Thank you @dhadka for your answer. That was helpful and your solution works.

[Jmiguel17]

Hi,

I was wondering if instead of saving the non-dominated solutions in each generation (nondominant(algorithm.population)) if there is a way to save the archive that saves the cumulative non-dominated solutions up to the current generation?

Do all algorithms create an Archive with the non-dominated solutions? If there isn't, I could create a callback method, that would create my own Archive with the cumulative updated non-dominated solutions in each generation. However, maybe this is already implemented and just I would be slowing down the search as I would be performing a non-dom sort at the end of each generation.

beforehand thanks for your help

[github-actions[bot]]

This issue is stale and will be closed soon. If you feel this issue is still relevant, please comment to keep it active. Please also consider working on a fix and submitting a PR.