MixedVariableProblem with arrays of elements

[patschmidt2]

Hello,

I am trying to perform an optimization on a mixed variable problem that I would like to describe with something like this:

class MixedVariableProblem(ElementwiseProblem):
    def __init__(self, **kwargs):
        vars = {
            "array1": IntegerArray(len=3, bounds=([1,2,3],[5,6,7]))a,
           "array2": IntegerArray(len=5, bounds=([9, 10, 13, 16, 19],[20, 30, 40, 45, 60]))
        }
        super().__init__(vars=vars, n_obj=1, **kwargs)

However, it seems that currently only simple problems of e.g. only integers are supported. I tried to implement a simple IntegerArray class, but PyMoo internally seems to always expect to be the shape of the individuals to be of (n_offsprings, n_matings, n_var) for the crossover. Is there a way here how I could adapt n_var per variable? Currently, this seems to be a property of the problem and not the individuals.
I would like to have different variables here since the arrays represent different information. I could also try to just create one long array and optimize with, for example, NSGA2, but I want to test custom crossover and mutation operations here.

[blankjul]

There are two ways of dealing with this.

1. The one you are trying to implement to have your own data type. I would need to think a little more if this is possible without modiyfing some classes.
2. Probably the easier solution. Using the classical GA customazation approach and implementing all the operators. (see here: https://pymoo.org/customization/custom.html)

[patschmidt2]

Thanks for the hints! Regarding the second option: You mean I then declare n_var = len(array0) + len(array1) + ... and then do the interpretation of this single, large vector in the operators?