def fit(self, X, y, sample_weight=None):
assert isinstance(y, np.ndarray)
assert len(y.shape) == 1, "only one-dimensional target is supported"
self.labels = np.unique(y)
self.labels.sort()
for l0, l1 in self.iterate_labels():
if self.use_all_data:
l0_vs_v1 = np.array(y >= l1, dtype='int')
new_x = X
else:
l0_vs_v1 = np.zeros(y.shape) - 1.0
l0_vs_v1[y == l0] = 0
l0_vs_v1[y == l1] = 1
ii = logical_or(y == l0, y == l1)
new_x = X[ii]
e = clone(self.estimator)
try:
e.fit(new_x, l0_vs_v1, sample_weight=sample_weight)
except TypeError:
print("%s doesn't support `sample_weight`. Ignoring it.")
e.fit(new_x, l0_vs_v1)
self.estimators[(l0, l1)] = e
def mask_or(m1, m2):
"""Logical or of the mask candidates m1 and m2, treating nomask as false.
Result may equal m1 or m2 if the other is nomask.
"""
if m1 is nomask:
return make_mask(m2)
if m2 is nomask:
return make_mask(m1)
if m1 is m2 and is_mask(m1):
return m1
return make_mask(umath.logical_or(m1, m2))
def __call__(self, x):
"Execute the call behavior."
return umath.logical_or(umath.greater(x, self.y2), umath.less(x, self.y1))
