Termination Criterion

[m-zakeri]

How can I understand with which condition(s) is my algorithm terminated when I use MultiObjectiveDefaultTermination (with customized values for arguments )?
In other words, I want to print the first criterion which causes the termination of my algorithm.
I can print the number of executed generations by res.algorithm.n_gen, but what about other criteria such as f_tol, cv_tol, etc.

[blankjul]

So far, the termination does not store why the decision has been made. The easiest way of achieving this is by inheriting from MultiObjectiveDefaultTermination and storing the decision in the object itself.
See the example below. Also, note that cv_tol is violated anytime for unconstrained problems to ensure the expression becomes false when f_tol is violated.

from pymoo.algorithms.moo.nsga2 import NSGA2
from pymoo.factory import get_problem
from pymoo.optimize import minimize
from pymoo.util.termination.default import MultiObjectiveDefaultTermination

problem = get_problem("zdt1")

algorithm = NSGA2(pop_size=100, eliminate_duplicates=True)

termination = MultiObjectiveDefaultTermination()


class MyMultiObjectiveDefaultTermination(MultiObjectiveDefaultTermination):

    def __init__(self, **kwargs) -> None:
        super().__init__(**kwargs)
        self.reasons = None

    def _decide(self, metrics):
        decisions = metrics[-1]
        cont = decisions["x_tol"] and (decisions["cv_tol"] or decisions["f_tol"])

        if not cont:
            self.reasons = [name for name in ["x_tol", "cv_tol", "f_tol"] if not decisions[name]]

        return cont


ret = minimize(problem,
               algorithm,
               termination=MyMultiObjectiveDefaultTermination(),
               seed=1,
               save_history=False,
               verbose=True)


print(ret.algorithm.termination.reasons)