Multi dimensional array in pymoo

[circlewithedges]

As I am new to coding and also to these type of optimization. I was trying to define a function using 2 dimension array. Can I do this using pymoo and what necessary changes i have to make in order to use multidimensional array. Reason for using multi dimensional array is that i have to solve large Optimization problem which consist Variables up to 4 indices (e.g. X_i_j_k_l) and i, j ,k, l can run up to any no. , is there any other way to implement this other than multi dimensional array. These i j k l can run up to different no.(e.g. i-(1,5), j- (1-3), k-(1-8), l - (1-14))

I have tried to write a code in 2d array but error is coming.

---

import numpy as np

from pymoo.core.problem import ElementwiseProblem

class MyProblem(ElementwiseProblem):

def __init__(self):
    super().__init__(n_var=4,
                     n_obj=2,
                     n_ieq_constr=1,
                     xl=np.array([[-20,-22],[-12,-15]]),
                     xu=np.array([[20,22],[12,15]]))
    
def _evaluate(self, x, out, *args, **kwargs):
    f1 = x[0][1]**2 + x[0][0]**3
    f2 = x[1][0]**2 + x[1][1]**3
    
    g1 = x[0][0] +x[0][1]**3 +x[1][0]**2 + x[1][1]**2 -200
   
    
    out["F"] = [-f1, -f2]
    out["G"] = [g1]

problem = MyProblem()

from pymoo.algorithms.moo.nsga2 import NSGA2
from pymoo.operators.crossover.sbx import SBX
from pymoo.operators.mutation.pm import PM
from pymoo.operators.sampling.rnd import FloatRandomSampling

algorithm = NSGA2(
pop_size=40,
n_offsprings=10,
sampling=FloatRandomSampling(),
crossover=SBX(prob=0.9, eta=15),
mutation=PM(eta=20),
eliminate_duplicates=True
)

from pymoo.termination import get_termination

termination = get_termination("n_gen", 40)

from pymoo.optimize import minimize

res = minimize(problem,
algorithm,
termination,
seed=1,
save_history=True,
verbose=True)

X = res.X
F = res.F

[blankjul]

Your variables always need to be 1-dimensional (this is the case for many optimization frameworks by the way).

You can flatten the array during optimization and then reshape it when evaluating. I have. provided an example for this a while ago for a 2-dimensional matrix (in theory this works for any dimension)

https://pymoo.org/case_studies/matrix_inversion.html

[tonysinghmss]

Currently I am trying to do POC on CVPR. I will try this approach.