Create a separate class for the args dict to be used in the GeneticAlgorithm

[gnypit]

What the args dict should look like:

args = {
        'genome': (g1, g2, ...),
        'selection': [(s11, s12, ...), ..., (sN1, sN2, ...)],
        'crossover': [(c11, c12, ...), ..., (cM1, cM2, ...)]
}

Where:

1. g1, g2, etc., are args for the genome_generator func;
2. s11, s12, etc., are args for the 1st selection operator passed in the selection_operators list of func and sN1, sN2, etc., are args of the Nth selection operator;
3. c11, c12, etc., are args for the 1st crossover operator passed in the crossover_operators list of func and cM1, cM2, etc., are args of the Mth crossover operator.

Let's create a class for that to enforce this format inside the GeneticAlgorithm.

[monokatarina]

Hi @gnypit!
I’d love to contribute to this issue (#44). Before I start, could you clarify a few details?

Validation Scope:

Should the class validate all argument types (e.g., check if genome is always a tuple)?

Or just enforce the structure (keys/format)?

Additional Methods:

Would you like helper methods (e.g., to_dict(), from_dict())?

Or keep it minimal (just store/validate args)?

Error Handling:

How strict should validation be? (e.g., raise ValueError vs. log warnings?)

[gnypit]

Hello @monokatarina !

I'm happy you want to work on this issue. The most important thing is to make sure that whatever type of arguments is passed, they are under proper keys. This way in the GeneticAlgorithm class an attribute called args can be safely used. Genome could be either a dict or a list - to be completly honest, I just haven't decided yet. But please keep in mind that in this issue the 'genome' key refers to the argument for the generator of genome for multiple chromosomes (Members of Generations).

When you look at PyGAD, you always provide a genome space to choose genes from. They are chosen with a uniform probability distribution. But in the case of simulations of the quantum internet (e.g. the BB84 protocol of communication between just 2 peaople) you have parameters of different data types, so the order in which they are given matters. I haven't yet figured out how to ensure this flexibility and yet make the code universal, for people to use in different cases. Do you have any ideas?

But when it comes to values stored in the args attribute inside the GeneticAlgorithm I think we can 'fix' data type of arguments - for the genome generator as a tuple and for selection/crossover operators as either a list of tuples or a tuple, including one element.

The general idea for the package is that a User may either choose a predefined genome generatoer for their chromosomes and predefined operators for selection and crossover, or define their own in a manner structured by the package, which the algorithm will be able to handle. This is linked to the issue #24

Having that in mind, it would be good to e.g. raise ValueError/KeyError, even if the overall structure could change in the future.

Did I answer your questions? How else may I help? I've just merged docstrings for the genetic_algorithm.py and will be working on implementing multiprocessing in another issue #20 to experiment a little bit with making the GeneticAlgorithm class able to use multiple cores in parallel to create and evaluate rival generations, I'll create a separate branch for that.