def train(self, x, y):
"""
training multiple estimators each for distinguishing a pair of classes.
Args:
X (numpy.ndarray): input points
y (numpy.ndarray): input labels
"""
self.estimators = []
self.classes = np.unique(y)
n_classes = self.classes.shape[0]
code_size = int(n_classes * self.code_size)
self.codebook = self.rand.random_sample((n_classes, code_size))
self.codebook[self.codebook > 0.5] = 1
self.codebook[self.codebook != 1] = 0
classes_index = dict((c, i) for i, c in enumerate(self.classes))
Y = np.array(
[self.codebook[classes_index[y[i]]] for i in range(x.shape[0])],
dtype=np.int)
logger.info("Require {} estimators.".format(Y.shape[1]))
for i in range(Y.shape[1]):
y_bit = Y[:, i]
unique_y = np.unique(y_bit)
if len(unique_y) == 1:
estimator = _ConstantPredictor()
estimator.fit(x, unique_y)
else:
if self.params is None:
estimator = self.estimator_cls()
else:
estimator = self.estimator_cls(*self.params)
estimator.fit(x, y_bit)
self.estimators.append(estimator)
def _fit_binary(estimator, X, y, classes=None):
"""Fit a single binary estimator."""
unique_y = np.unique(y)
if len(unique_y) == 1:
if classes is not None:
if y[0] == -1:
c = 0
else:
c = y[0]
warnings.warn("Label %s is present in all training examples." %
str(classes[c]))
estimator = _ConstantPredictor().fit(X, unique_y)
else:
estimator = clone(estimator)
estimator.fit(X, y)
return estimator
def _fit_binary(estimator, X, y, classes=None, sample_weight=None):
"""Fit a single binary estimator."""
unique_y = np.unique(y)
if len(unique_y) == 1:
if classes is not None:
if y[0] == -1:
c = 0
else:
c = y[0]
warnings.warn("Label %s is present in all training examples." %
str(classes[c]))
estimator = _ConstantPredictor().fit(X, unique_y)
else:
estimator = clone(estimator)
estimator.fit(X, y, sample_weight=None)
return estimator
def _fit_binary(estimator,
x,
y,
classes=None,
sample_weight=None,
eval_set=None,
eval_metric=None,
early_stopping_rounds=None,
verbose=True,
xgb_model=None,
sample_weight_eval_set=None,
callbacks=None):
"""Fit a single binary estimator.
This function is ported from sklearn multiclass module."""
unique_y = np.unique(y)
if len(unique_y) == 1:
if classes is not None:
if y[0] == -1:
c = 0
else:
c = y[0]
warnings.warn(
f"Label {str(classes[c])} is present in all training examples."
)
estimator = _ConstantPredictor().fit(x, unique_y)
else:
estimator = clone(estimator)
estimator.fit(x,
y,
sample_weight=sample_weight,
eval_set=eval_set,
eval_metric=eval_metric,
early_stopping_rounds=early_stopping_rounds,
verbose=verbose,
xgb_model=xgb_model,
sample_weight_eval_set=sample_weight_eval_set,
callbacks=callbacks)
return estimator