Expected shape for objective and constraint functions in multi-optimization problem using pymoo

Question

I'm trying to use pymoo for a NSGA 2 multiple optimization problem, but I get a Problem error for shape expect and what is provided. here is the code:

import numpy as np
from pymoo.algorithms.moo.nsga2 import NSGA2
from pymoo.core.problem import ElementwiseProblem
from pymoo.core.problem import Problem
from pymoo.optimize import minimize
from pymoo.visualization.scatter import Scatter

class TestProblem(ElementwiseProblem):

    def __init__(self):
        super().__init__(n_var=3,
                         n_obj=2,
                         n_ineq_const=2,
                         vtype=int)

    def _evaluate(self, TA, out, *args, **kwargs):
        f1 = 910 * TA[0] + 2280 * TA[1] + 5500 * TA[2]
        f2 = (200-TA[0])/TA[0] + (500-TA[1])/TA[1] + (100-TA[2])/TA[2]
        g1 = TA[2] + TA[0] - TA[1]
        g2 = TA[0] + TA[2] - 61

        out["F"] = [f1, -f2]
        out["G"] = np.column_stack([g1, g2])


problem = TestProblem()

algorithm = NSGA2(pop_size=100)

res = minimize(problem,
               algorithm,
               ("n_gen", 200),
               save_history=False,
               verbose=False,
               seed=1)

and the error I get is:

Exception: ('Problem Error: G can not be set, expected shape (100, 0) but provided (100, 1, 2)', ValueError('cannot reshape array of size 200 into shape (100,0)')) I would be thankful if you would help me with it.

I tried "np.column_stack" for a solution but the error remains.

Answer 1

1. You didn't set the number of inequality constraints properly, it should be:

n_ieq_constr=2

(Not n_ineq_const=2)

2. Next to this, add some lower and upper bounds for your design variables:

xl=np.array([-100, -100, -100]),
xu=np.array([100, 100, 100]),

So your full code becomes:

import numpy as np
from pymoo.algorithms.moo.nsga2 import NSGA2
from pymoo.core.problem import ElementwiseProblem
from pymoo.core.problem import Problem
from pymoo.optimize import minimize
from pymoo.visualization.scatter import Scatter

class TestProblem(ElementwiseProblem):

    def __init__(self):
        super().__init__(n_var=3,
                         n_obj=2,
                         n_ieq_constr=2,
                         xl=np.array([-100, -100, -100]),
                         xu=np.array([100, 100, 100]),
                         vtype=int)

    def _evaluate(self, TA, out, *args, **kwargs):
        f1 = 910 * TA[0] + 2280 * TA[1] + 5500 * TA[2]
        f2 = (200-TA[0])/TA[0] + (500-TA[1])/TA[1] + (100-TA[2])/TA[2]
        g1 = TA[2] + TA[0] - TA[1]
        g2 = TA[0] + TA[2] - 61

        out["F"] = [f1, -f2]
        out["G"] = [g1, g2]


problem = TestProblem()

algorithm = NSGA2(pop_size=100)

res = minimize(problem,
               algorithm,
               ("n_gen", 200),
               save_history=False,
               verbose=True,
               seed=1)

plot = Scatter()
plot.add(res.F, edgecolor="red", facecolor="none")
plot.show()