Source code for paref.pareto_reflections.find_edge_points

import numpy as np

from paref.interfaces.moo_algorithms.blackbox_function import BlackboxFunction
from paref.pareto_reflections.minimize_g import MinGParetoReflection




class FindEdgePoints(MinGParetoReflection):
    """Find the edge points of the Pareto front

    .. warning::

        This Pareto reflection assumes that there exist edge points

    When to use
    -----------
    This Pareto reflection should be used if the edge points of the Pareto front are searched.

    .. note::

        In two dimensions, the edge points of the Pareto correspond to the Pareto points which are
        minimal in one component, i.e. to the one Pareto points.

    What it does
    ------------
    The Pareto points of this map are the ones which minimize the weighted sum where one component
    is given much smaller weight than the others.

    Mathematical formula
    --------------------

    .. math::
        p(x) = \sum_{i=1,...,n,i\\neq j}\\epsilon x_{i}+ x_j

    where :math:`j` is the component to which the edge corresponds.
    """

    def __init__(self,
                 dimension: int,
                 blackbox_function: BlackboxFunction, ):
        """

        Parameters
        ----------
        dimension :
            component to which the edge corresponds
        blackbox_function :
            blackbox function to which this reflection is applied
        """
        self.dimension = dimension
        super().__init__(blackbox_function=blackbox_function)

    @property
    def g(self):
        return lambda x: np.sum(x) - x[self.dimension]