I'd like to take this up. I'm working with genetic algorithms lately.

Sounds good, it will be a major addition to machinelearn.js. Would you mind giving me a brief idea on how you want to introduce GA in machinelearn.js? Will there be a single GeneticAlgorithm API with configurations?

I'd like to demonstrate the usage of the library in genetic algorithms. I was thinking of giving examples of how to use the library for genetic purposes. For the very beginning, I was thinking of introducing the concept of evolution, fitness , mutation, selection etc through an example of achieving a desired string given a random population. An API would be incredibly awesome, but since the algorithms such as selection etc are exclusive to the problem in question, I thought we'd let the end user explore.
First we can start by showing examples. I'll work on the API side by side to cover generalised functions what do you think?

I like the idea of initially introducing the smaller components of GA like evolution, fitness, mutation and selection as separate APIs.

I think a massive addition like GA should belong to its category in the library

^ We will create a new category called "Genetic Algorithms" or something and put all the associated APIs in there. We might create another layer of submenu so it can be structured like:

Genetic Algorithms
- Methods
-- Evolve
-- Train
-- Regularization
-- Mutation
-- Selection
-- Activation
-- Cost
-- Connection
-- Gating
-- Rate
- Architectures
-- NEAT

Then we can build some demos under
https://www.machinelearnjs.com/examples/
demonstration of solving a simple problem such as XOR problem using the APIs that you've introduced

Later we can introduce a fully fledged architecture like NEAT by using the APIs that you've created. It would be cool to build a game bot (running directly in the browser) using this =)

We can work on this feature together, happy to receive your inputs as well.

This is more structured. We will proceed the way you conceptualise it 😄
But it's gonna take some time. I'm still not sure how we can go about for generic methods of evolution and selection, since the genotype of the population will depend on the situation. Example: for a flappy bird ai, the genotype maybe a regression model that predicts if the bird should jump , whereas there are some other instances like the xor, it might be Anns.
How do we tackle the problem specific genotype configuration in our APIs?

Sounds good =)

How do we tackle the problem specific genotype configuration in our APIs?
I haven't thought too deeply about this yet. My initial proposal is to follow the structure of https://github.com/wagenaartje/neataptic

I also want to read more on genetic algorithms over the weekend, would you be up for a studying meeting over the internet this weekend?

Yes! We will have our conversations over telegram