def __init__(self, path=None):
super().__init__(path)
ones = np.ones(self.config.dimension)
self._dist_initial = np.sqrt(np.log(1 / self.config.initial_value) / 4)
self._x0 = self._dist_initial * ones / np.sqrt(self.config.dimension)
self._max_value = 1.0
self._domain = ContinuousDomain(-ones, ones)
def __init__(self, path=None):
super().__init__(path)
self._max_value = self.config.s
theta = np.random.normal(size=self.config.dimension)
self._theta = theta / np.linalg.norm(theta) * self.config.s
self._domain = ContinuousDomain(-2 * np.ones(self.config.dimension),
2 * np.ones(self.config.dimension))
def __init__(self, path=None):
super().__init__(path)
self._max_value = 1.03162842
d = self.config.dimension
if d <= 2:
raise Exception(
"Need dimension at least 3 to create embedded version of Camelback"
)
self._x0 = np.array([0.5, 0.2] + [0.] * (d - 2))
self._domain = ContinuousDomain(np.array([-2, -1] + [-1] * (d - 2)),
np.array([2, 1] + [1] * (d - 2)))
def test_norm_denorm_same(self):
self.init()
for i in range(20):
self.X = (np.random.random((self.dim,)) - 0.5) * 2.
assert (self.X <= 1.0).all()
assert (self.X >= -1.0).all()
domain = ContinuousDomain(np.asarray([-5] * self.dim), np.asarray([2] * self.dim))
X_norm = rembo_algorithm.normalize(self.X, domain)
X_denorm = rembo_algorithm.denormalize(X_norm, domain)
assert np.isclose(self.X, X_denorm).all(), ("Not quite the same values after norm+denorm! ", (self.X, X_denorm))
def __init__(self, path=None):
super().__init__(path)
self._x0 = np.array([0., 0.])
self._max_value = 1.
self._domain = ContinuousDomain(np.array([-0.5, -0.5]),
np.array([0.5, 0.5]))
def __init__(self, path=None):
super().__init__(path)
self._x0 = np.array([0.5, 0.2])
self._x0 = np.array([-0.12977758051079197, 0.2632096107305229])
self._max_value = 1.03162842
self._domain = ContinuousDomain(np.array([-2, -1]), np.array([2, 1]))
def __init__(self, path=None):
super().__init__(path)
self._x = np.array([1.])
self._max_value = 1.0026469
self._domain = ContinuousDomain(np.array([-1]), np.array([2]))
def __init__(self, path=None):
super().__init__(path)
self._x = np.array([0.1])
self._max_value = 1.1 # at 0.5
self._domain = ContinuousDomain(np.array([0.]), np.array([1]))
def __init__(self, path=None):
super().__init__(path)
self._x = np.array([15])
self._max_value = 1.25375424 # determined using scipy.minimze
self._domain = ContinuousDomain(np.array([-20]), np.array([20]))
def __init__(self, path=None):
super().__init__(path)
ones = np.ones(self.config.dimension)
self._x0 = 0.5 * ones / np.sqrt(self.config.dimension)
self._max_value = 1.0
self._domain = ContinuousDomain(-ones, ones)