def calc_hypervol(ref=[], front=[], minimize=True, **kwargs):
"""Calculate hypervolume metrics of a Pareto set, given a reference point.
Parameters
----------
ref : {1D-array-like}, shape (n_objectives, )
The reference point.
front : {array-like, sparse matrix}, shape (n_optimals, n_objectives)
The Pareto optimals on which to calculate hypervolume.
Returns
-------
hypevol : scalar, the calculated hypervolume between the reference point
and the Pareto optimals.
"""
# return distance if single objective
if front.shape[1] == 1:
return np.subtract(front.ravel() - ref).sum()
elif front.shape[1] == 2:
_fs = front[ front[:, 0].argsort()[::-1] ]
return _calc_hypervol(ref=ref, front=_fs, minimize=minimize, **kwargs)
else:
pass
return hypervolume(front, ref)
def test_2d():
front = np.loadtxt(
os.path.join(os.path.dirname(__file__), 'fixtures', 'CF7.dat'))
ref = np.array([1.1, 1.1])
hv1 = wfg(front, ref)
hv2 = hv.hypervolume(front, ref)
assert abs(hv1 - hv2) <= 1e-12
def calculateHypervolume(d2Points, d1Reference):
"""Returns the Hypervolume Indicator (total area contained by the Pareto Front).
d2Points -- a 2D array of locations (each row is a location)
d1Reference -- a D-element list specifying the nadir point (e.g. [5, 5, 5, 5])
Returns a float for the Hypervolume Indicator value
"""
return hv.hypervolume(d2Points, d1Reference)
def my_nsga2(n,
nbgen,
evaluate,
ref_point=np.array([1, 1]),
IND_SIZE=5,
weights=(-1.0, -1.0)):
"""NSGA-2
NSGA-2
:param n: taille de la population
:param nbgen: nombre de generation
:param evaluate: la fonction d'évaluation
:param ref_point: le point de référence pour le calcul de l'hypervolume
:param IND_SIZE: la taille d'un individu
:param weights: les poids à utiliser pour la fitness (ici ce sera (-1.0,) pour une fonction à minimiser et (1.0,) pour une fonction à maximiser)
"""
creator.create("MaFitness", base.Fitness, weights=weights)
creator.create("Individual", list, fitness=creator.MaFitness)
toolbox = base.Toolbox()
paretofront = tools.ParetoFront()
# à compléter
# Pour récupérer l'hypervolume, nous nous contenterons de mettre les différentes aleur dans un vecteur s_hv qui sera renvoyé par la fonction.
pointset = [np.array(ind.fitness.getValues()) for ind in paretofront]
s_hv = [hv.hypervolume(pointset, ref_point)]
# Begin the generational process
for gen in range(1, nbgen):
if (gen % 10 == 0):
print("+", end="", flush=True)
else:
print(".", end="", flush=True)
# à completer
pointset = [np.array(ind.fitness.getValues()) for ind in paretofront]
s_hv.append(hv.hypervolume(pointset, ref_point))
return population, paretofront, s_hv
def compute_hvolume(obj: np.ndarray, ref: np.ndarray=None) -> float:
assert len(obj.shape) == 2
if ref is None:
ref = np.ones(obj.shape[1])
valid = np.all(np.less(obj, ref), axis=1)
obj = obj[valid]
obj = obj[non_dominated_selection(obj)]
hv = hypervolume(obj, ref)
return hv