Closed bandundu closed 2 years ago
I have just found the time to work on the dynamic problem. For a sneak peak, see: http://develop.pymoo.org/algorithms/moo/dnsga2.html
The DF problems are available in the current development branch: https://github.com/anyoptimization/pymoo/blob/develop/pymoo/problems/dynamic/df.py I am currently working on the 0.6.0 release which will be available soon (hopefully).
The implementation is very similar to yours. I am searching for someone to test the implementation and perform a benchmark. Would this align with your research? Are you interested?
Hi! This would be excellent for me and is exactly where my research is heading. I'll gladly help wherever I can, I only might need guidance here and there.
As a first, how would one install the latest development branch?
With this i am getting following error
With this I am only getting
which does not contain the newest problems and algorithm.
Thank you very much!
I was able to install the latest development branch with
git clone https://github.com/anyoptimization/pymoo
cd pymoo
git checkout develop
pip install .
Have you found the time to check the DF
problems yet? I am also more than happy to have a meeting to talk about features for dynamic optimization in general.
Hello my optimization fellows,
for my thesis I'm currently implementing the CEC2018 DF Benchmark Suite in Pymoo
So far I defined following problem:
Now, one can test this problem with the following code. I am using the ask and tell interface as I also want to implement dynamic multi objective optimization algorithms that relocate the population after change in the environment happens.
Examining the output, one can notice that performance metrics are updated only every 30 generations, as the problems fitnessfunction changes every 30 generations.
At this point, my question.
Shouldn't the metrics still at least somewhat change every generation, as the fitness function of the DF1 Problem prefers other solutions? To me, it looks like NSGA doesn't follow the new Pareto-Front at all.
I also attached some visualizations to see my concern. visualisation.zip
Is this a bug or am I misunderstanding something and this is just how NSGA-II should work here?
I would be really thankful for any kind of help or clarification here, thanks!