def _check_is_fitted(self):
"""Check if the estimator is fitted.
Raises:
NotFittedError: If the estimator is not fitted.
"""
sklearn_check_is_fitted(self, ['estimator_'])
def predict(self, X):
"""Predict the class labels for the provided data.
Args:
X (:class:`FDataGrid`): FDataGrid with the test samples.
Returns:
(np.array): y : array of shape [n_samples] or
[n_samples, n_outputs] with class labels for each data sample.
"""
sklearn_check_is_fitted(self, 'centroids_')
return self.classes_[self._pairwise_distance(
X, self.centroids_).argmin(axis=1)]
def predict(self, X):
"""Predict the class labels for the provided data.
Args:
X (:class:`FDataGrid`): FDataGrid with the test samples.
Returns:
y (np.array): array of shape [n_samples] with class labels
for each data sample.
"""
sklearn_check_is_fitted(self)
depths = [
distribution.predict(X)
for distribution in self.distributions_
]
return self.classes_[np.argmax(depths, axis=0)]
def check_is_fitted(estimator, attributes, msg=None, all_or_any=all):
"""Checks whether the net is initialized.
Note: This calls ``sklearn.utils.validation.check_is_fitted``
under the hood, using exactly the same arguments and logic. The
only difference is that this function has an adapted error message
and raises a ``skorch.exception.NotInitializedError`` instead of
an ``sklearn.exceptions.NotFittedError``.
"""
if msg is None:
msg = ("This %(name)s instance is not initialized yet. Call "
"'initialize' or 'fit' with appropriate arguments "
"before using this method.")
try:
sklearn_check_is_fitted(
estimator=estimator,
attributes=attributes,
msg=msg,
all_or_any=all_or_any,
)
except NotFittedError as e:
raise NotInitializedError(str(e))