Algorithm close to a default simple GA?

[seyeint]

Cheers, first of all I have to say this project not only seems very intelligently engineered, but also has the correct amount of abstraction for all types of users with various use cases that were for sure lacking in previous EC accelerators.

I've done research on ECs and developed a type of genetic algorithm and I am naturally interested in translating into to into this framework.
I've read the paper and I've skimmed through some of the algos and functions that are already implemented but I would like to ask if there are any default genetic algorithm implementation so I can build upon that my own custom algo with some sort of guidance while I am not an expert with the framework. If there's no GA, do you have any opinion about which implemented algo would give me a good starting point leading to my own implementation of my GA?

Thank you.

[BillHuang2001]

Thank you for your insightful feedback on our project.

Currently, there isn't a specific SimpleGA implementation available. Previously, we had one, but it was removed due to concerns regarding potential misinterpretation. It's important to note that "GA" represents a class of algorithms, and the performance can vary significantly depending on the specific operators employed.

So instead of providing a SimpleGA, we propose to offer a simple abstraction for all GA algorithms. This abstraction will allow users to define a series of operators to construct their own genetic algorithms. (similar to the approach of torch.nn.Sequential for building neural network blocks). We are expecting this to land in the next major release (about 1-3 weeks later).

If you want to build your own GA now, we provide comprehensive guidance on how to craft custom algorithms within EvoX. You can access this resource, including an ExampleGA implementation, and our step-by-step guide on translating a typical GA implementation into our framework, at Custom algorithms and problems in EvoX.

Thank you.

[seyeint]

Thanks for your quick answer. And yes that example of the typical GA implementation helps, I didn't really see it before you pointed it out. Depending on the mechanics one applies to their own algorithm (that might involve logic that reduces the ability of this framework to fully explore its efficiencies), I'm guessing one can at least expect the basic Tpop and Tfit parallelized calculation enhancement, which is already great.