Issue on n_parents using MOEAD

[MassimoValle]

Hi, I am need of a help.

Running MOEA/D with MixedVariableSampling() and with MixedVariableMating() I get the following error:
AttributeError: 'dict' object has no attribute 'n_parents'

It refers to this line in the moead.py class [line 103]:
P = self.selection.do(self.problem, pop, 1, self.mating.crossover.n_parents, neighbors=[self.neighbors[k]])

The error occurs at the end of the first iteration, when the algorithm is preparing to perform the mating.

I can't figure out where I am wrong in the implementation and what parameter I am not setting or I am setting but incorrectly.
This is my implementation of the code:

        problem = MyProblem(fast=True)

        partitions = 3
        ref_dirs = get_reference_directions("das-dennis", 2, n_partitions=partitions)


        algorithm = MOEAD(
            ref_dirs= ref_dirs,
            sampling=MixedVariableSampling(),
            mating=MixedVariableMating(eliminate_duplicates=MixedVariableDuplicateElimination()),
            n_neighbors=12,
            prob_neighbor_mating= 0.7
        )

        res = minimize(
            problem,
            algorithm,
            ('n_gen', 30),
            seed=1,
            save_history=True,
            verbose=False
        )

I am pretty sure the problem is defined correctly since with NSGA2, NSGA3 and GA it works correctly.

Anyone can help me?

[blankjul]

MOEAD does not support the optimization of mixed variables. You will have to use GA or NSGA2 as shown in the tutorial.
If you think MOEAD can be rewritten to make it work, please let me know or directly create a PR and I will have a look.

[MassimoValle]

Thank you for your reply.
May I ask from which version of pymoo the MOEA/D algorithm is no longer supported for mixed variable optimization?
This is because I have used this code in the past that used the optimization of mixed variables with MOEA/D and it worked correctly:

from random import Random
import numpy as np
from pymoo.core.problem import ElementwiseProblem
from pymoo.factory import get_sampling, get_crossover, get_mutation, get_reference_directions
from pymoo.operators.mixed_variable_operator import MixedVariableSampling, MixedVariableMutation, MixedVariableCrossover
from pymoo.factory import get_sampling, get_crossover, get_mutation
from pymoo.optimize import minimize
import sys
from pymoo.core.problem import starmap_parallelized_eval
from multiprocessing.pool import ThreadPool
from pymoo.algorithms.moo.moead import MOEAD, ParallelMOEAD
import random


mask = ["int", "real", "int", "real"]

sampling = MixedVariableSampling(mask, {
    "real": get_sampling("real_random"),
    "int": get_sampling("int_random"),
})

crossover = MixedVariableCrossover(mask, {
    "real": get_crossover("real_sbx"),
    "int": get_crossover("int_sbx")
})

mutation = MixedVariableMutation(mask, {
    "real": get_mutation("real_pm"),
    "int": get_mutation("int_pm")
})


class MyProblem(ElementwiseProblem):

    def __init__(self, fast = False, verbose = False):
        super().__init__(n_var= 4, n_obj = 2, n_constr = 0, xl = [1, 12.0, 2, 0.7], xu = [50, 30000.0, 11, 4.0])
        self.evaluationNumber = 0
        self.fast = fast
        self.verbose = verbose
        self.count = 0

    def _evaluate(self, x, out, *args, **kwargs):
        self.evaluationNumber = self.evaluationNumber + 1
	    # other stuffs


def main():

  problem = MyProblem(fast=fast, verbose=args.verbose)
  numb_obj = 2
  
  partitions = 3
  ref_dirs = get_reference_directions("das-dennis", numb_obj, n_partitions=partitions)
  algorithm = MOEAD(
      n_offsprings= None,
      sampling= sampling,
      crossover= crossover,
      mutation= mutation,
      ref_dirs= ref_dirs,
      n_neighbors=15,
      prob_neighbor_mating= 0.7
  )
  
  res = minimize(
              problem,
              algorithm,
              seed= None,
              save_history= True,
              verbose= True
          )

[blankjul]

ohh interesting that it has even worked before (I did not know this because I have never tested it with MOEAD).
My best guess would be 0.6.0 or 0.5.0. If this is a minor fix I am happy to check a PR and merge it quickly to make it work again.