Source code for paref.pareto_reflection_sequences.multi_dimensional.grid_search
from paref.interfaces.moo_algorithms.blackbox_function import BlackboxFunction
from paref.interfaces.sequences_pareto_reflections.sequence_pareto_reflections import \
SequenceParetoReflections
from paref.pareto_reflection_sequences.multi_dimensional.find_edge_points_sequence import FindEdgePointsSequence
from paref.pareto_reflections.fill_gap import FillGap
[docs]
class GridSearch(SequenceParetoReflections):
"""Determine an equidistant grid of Pareto points
When to use
-----------
Use this sequence when you want to determine an equidistant grid of Pareto points, for example when you want to
get a solid approximation of the Pareto front taking into account based on the maximum number of
evaluations you granted.
.. note::
This sequence should be used *at the beginning* of the optimization process, i.e. before other Pareto reflection
are applied. This is because the sequence does not take into account the Pareto front of the evaluations and
might, therefore, return points that are already closet to or in the Pareto front.
What it does
------------
This sequence determines an equidistant grid of Pareto points including edge points for all dimensions.
Examples
--------
# TBA: add
"""
def __init__(self,
epsilon: float = 0.01
):
"""Specify some utopia point
Parameters
----------
epsilon : float default 0.01
epsilon for numerical stability
"""
self.epsilon = epsilon
self._edge_point_sequence = FindEdgePointsSequence()
self._close_gaps_sequence = None
def next(self, blackbox_function: BlackboxFunction) -> FillGap:
"""Return a :py:class:`fill gap <paref.pareto_reflections.fill_gap_2d.FillGap2D>` Pareto reflection
Parameters
----------
blackbox_function : BlackboxFunction
blackbox function to which sequence is applied
Returns
-------
FillGap
fill gap Pareto reflection corresponding to the greatest gap in Pareto front of the evaluations
"""
raise NotImplementedError('This method is not implemented yet.')
