def learn_one(self, X, y, *, weight=1.0, tree=None):
"""Update the node with the provided instance.
Parameters
----------
X: numpy.ndarray of length equal to the number of features.
Instance attributes for updating the node.
y: numpy.ndarray of length equal to the number of targets.
Instance targets.
weight: float
Instance weight.
tree: HoeffdingTreeRegressor
Regression Hoeffding Tree to update.
"""
self.update_stats(y, weight)
self.update_attribute_observers(X, y, weight, tree)
if self.perceptron_weights is None:
# Creates matrix of perceptron random weights
_, rows = get_dimensions(y)
_, cols = get_dimensions(X)
self.perceptron_weights = self._random_state.uniform(
-1.0, 1.0, (rows, cols + 1))
self._normalize_perceptron_weights()
if tree.learning_ratio_const:
learning_ratio = tree.learning_ratio_perceptron
else:
learning_ratio = tree.learning_ratio_perceptron / (
1 + self.stats[0] * tree.learning_ratio_decay)
for i in range(int(weight)):
self._update_weights(X, y, learning_ratio, tree)
def learn_from_instance(self, X, y, weight, rht):
"""Update the node with the provided instance.
Parameters
----------
X: numpy.ndarray of length equal to the number of features.
Instance attributes for updating the node.
y: numpy.ndarray of length equal to the number of targets.
Instance targets.
weight: float
Instance weight.
rht: HoeffdingTreeRegressor
Regression Hoeffding Tree to update.
"""
if self.perceptron_weight is None:
self.perceptron_weight = {}
# Creates matrix of perceptron random weights
_, rows = get_dimensions(y)
_, cols = get_dimensions(X)
self.perceptron_weight[0] = \
self.random_state.uniform(-1.0, 1.0, (rows, cols + 1))
# Cascade Stacking
self.perceptron_weight[1] = \
self.random_state.uniform(-1.0, 1.0, (rows, rows + 1))
self.normalize_perceptron_weights()
try:
self._observed_class_distribution[0] += weight
except KeyError:
self._observed_class_distribution[0] = weight
if rht.learning_ratio_const:
learning_ratio = rht.learning_ratio_perceptron
else:
learning_ratio = rht.learning_ratio_perceptron / \
(1 + self._observed_class_distribution[0] *
rht.learning_ratio_decay)
try:
self._observed_class_distribution[1] += weight * y
self._observed_class_distribution[2] += weight * y * y
except KeyError:
self._observed_class_distribution[1] = weight * y
self._observed_class_distribution[2] = weight * y * y
for i in range(int(weight)):
self.update_weights(X, y, learning_ratio, rht)
for i, x in enumerate(X):
try:
obs = self._attribute_observers[i]
except KeyError:
# Creates targets observers, if not already defined
if rht.nominal_attributes is not None and i in rht.nominal_attributes:
obs = NominalAttributeRegressionObserver()
else:
obs = NumericAttributeRegressionObserverMultiTarget()
self._attribute_observers[i] = obs
obs.observe_attribute_class(x, y, weight)
def update_buffer_content(self, X, y, timestamp, uid):
self.members['uids'].append(uid)
if self.members['X'] is None:
self.members['X'] = np.zeros((0, get_dimensions(X)[1]))
self.members['y'] = np.zeros((0, get_dimensions(y)[1]))
self.earliest_timestamp = timestamp
self.members['X'] = np.vstack((self.members['X'], X))
self.members['y'] = np.vstack((self.members['y'], y))
self.members['timestamps'].append(timestamp)
self.size += 1
self.latest_timestamp = timestamp
def update_weights(self, X, y, learning_ratio, rht):
"""Update the perceptron weights
Parameters
----------
X: numpy.ndarray of length equal to the number of features.
Instance attributes for updating the node.
y: numpy.ndarray of length equal to the number of targets.
Targets values.
learning_ratio: float
perceptron learning ratio
rht: HoeffdingTreeRegressor
Regression Hoeffding Tree to update.
"""
normalized_sample = rht.normalize_sample(X)
normalized_base_pred = self._predict_base(normalized_sample)
_, n_features = get_dimensions(X)
_, n_targets = get_dimensions(y)
normalized_target_value = rht.normalize_target_value(y)
self.perceptron_weight[0] += learning_ratio * \
(normalized_target_value - normalized_base_pred)[:, None] @ \
normalized_sample[None, :]
# Add bias term
normalized_base_pred = np.append(normalized_base_pred, 1.0)
normalized_meta_pred = self._predict_meta(normalized_base_pred)
self.perceptron_weight[1] += learning_ratio * \
(normalized_target_value - normalized_meta_pred)[:, None] @ \
normalized_base_pred[None, :]
self.normalize_perceptron_weights()
# Update faded errors for the predictors
# The considered errors are normalized, since they are based on
# mean centered and sd scaled values
self.fMAE_M = 0.95 * self.fMAE_M + np.absolute(
normalized_target_value -
rht.normalize_target_value(self._observed_class_distribution[1] /
self._observed_class_distribution[0]))
# Ignore added bias term in the comparison
self.fMAE_P = 0.95 * self.fMAE_P + np.absolute(
normalized_target_value - normalized_base_pred[:-1])
self.fMAE_SP = 0.95 * self.fMAE_SP + np.absolute(
normalized_target_value - normalized_meta_pred)
def add_sample(self, X, y, arrival_time):
if not self._is_initialized:
self._n_features = get_dimensions(X)[1]
self._n_targets = get_dimensions(y)[1]
self.configure()
if self._n_features != get_dimensions(X)[1]:
raise ValueError("Inconsistent number of features in X: {}, previously observed {}.".
format(get_dimensions(X)[1], self._n_features))
if not self.overlap_windows:
self._add_sample_no_overlap(X, y, arrival_time)
else:
self._add_sample_overlap(X, y, arrival_time)
def partial_fit(self, X, y, sample_weight=None):
"""Incrementally trains the model. Train samples (instances) are
composed of X attributes and their corresponding targets y.
Tasks performed before training:
* Verify instance weight. if not provided, uniform weights (1.0) are
assumed.
* If more than one instance is passed, loop through X and pass
instances one at a time.
* Update weight seen by model.
Training tasks:
* If the tree is empty, create a leaf node as the root.
* If the tree is already initialized, find the corresponding leaf for
the instance and update the leaf node statistics.
* If growth is allowed and the number of instances that the leaf has
observed between split attempts exceed the grace period then attempt
to split.
Parameters
----------
X: numpy.ndarray of shape (n_samples, n_features)
Instance attributes.
y: numpy.ndarray of shape (n_samples, n_targets)
Target values.
sample_weight: float or array-like
Samples weight. If not provided, uniform weights are assumed.
"""
if y is not None:
# Set the number of targets once
if not self._n_targets_set:
_, self._n_targets = get_dimensions(y)
self._n_targets_set = True
row_cnt, _ = get_dimensions(X)
if sample_weight is None:
sample_weight = np.ones(row_cnt)
if row_cnt != len(sample_weight):
raise ValueError(
'Inconsistent number of instances ({}) and weights ({}).'.format(
row_cnt, len(sample_weight)
)
)
for i in range(row_cnt):
if sample_weight[i] != 0.0:
self._train_weight_seen_by_model += sample_weight[i]
self._partial_fit(X[i], y[i], sample_weight[i])
def predict(self, X):
""" Predict classes for the passed data.
Parameters
----------
X : numpy.ndarray of shape (n_samples, n_features)
The set of data samples to predict the class labels for.
Returns
-------
A numpy.ndarray with all the predictions for the samples in X.
Notes
-----
The predict function will average the predictions from all its learners
to find the most likely prediction for the sample matrix X.
"""
r, c = get_dimensions(X)
proba = self.predict_proba(X)
predictions = []
if proba is None:
return None
for i in range(r):
predictions.append(np.argmax(proba[i]))
return np.asarray(predictions)
def predict_proba(self, X):
""" Estimate the probability of X belonging to each class-labels.
Parameters
----------
X : numpy.ndarray of shape (n_samples, n_features)
Samples one wants to predict the class probabilities for.
Returns
-------
A numpy.ndarray of shape (n_samples, n_labels), in which each outer
entry is associated with the X entry of the same index. And where the
list in index [i] contains len(self.target_values) elements, each of
which represents the probability that the i-th sample of X belongs to
a certain class-label.
"""
n_samples, n_features = get_dimensions(X)
y_proba = []
if self.ensemble is None:
self._init_ensemble(n_features=n_features)
return np.zeros(n_samples)
for i in range(n_samples):
y_proba.append(self._predict_proba(np.asarray([X[i]])))
return np.asarray(y_proba)
def partial_fit(self, X, y, classes=None, sample_weight=None):
""" Partially (incrementally) fit the model.
Parameters
----------
X : numpy.ndarray of shape (n_samples, n_features)
The features to train the model.
y: numpy.ndarray of shape (n_samples)
An array-like with the class labels of all samples in X.
classes: numpy.ndarray, list, optional (default=None)
Array with all possible/known class labels. This is an optional parameter, except
for the first partial_fit call where it is compulsory.
sample_weight: numpy.ndarray of shape (n_samples), optional (default=None)
Samples weight. If not provided, uniform weights are assumed.
"""
if self.classes is None and classes is not None:
self.classes = classes
if sample_weight is None:
weight = 1.0
else:
weight = sample_weight
if y is not None:
row_cnt, _ = get_dimensions(X)
weight = check_weights(weight, expand_length=row_cnt)
for iterator in range(row_cnt):
if weight[iterator] != 0.0:
self._partial_fit(X[iterator], y[iterator], self.classes,
weight[iterator])
return self
def predict_proba(self, X):
"""Predicts probabilities of all label of the instance(s).
Parameters
----------
X: numpy.ndarray of shape (n_samples, n_features)
Samples for which we want to predict the labels.
Returns
-------
numpy.array
Predicted the probabilities of all the labels for all instances in X.
"""
r, _ = get_dimensions(X)
predictions = []
for i in range(r):
votes = copy.deepcopy(self.get_votes_for_instance(X[i]))
if votes == {}:
# Tree is empty, all classes equal, default to zero
predictions.append([0])
else:
if sum(votes.values()) != 0:
votes = normalize_values_in_dict(votes, inplace=False)
if self.classes is not None:
y_proba = np.zeros(int(max(self.classes)) + 1)
else:
y_proba = np.zeros(int(max(votes.keys())) + 1)
for key, value in votes.items():
y_proba[int(key)] = value
predictions.append(y_proba)
return np.array(predictions)
def _init_ensemble(self, X):
self._set_max_features(get_dimensions(X)[1])
self.ensemble = [
ARFBaseLearner(
index_original=i,
classifier=ARFHoeffdingTreeClassifier(
max_byte_size=self.max_byte_size,
memory_estimate_period=self.memory_estimate_period,
grace_period=self.grace_period,
split_criterion=self.split_criterion,
split_confidence=self.split_confidence,
tie_threshold=self.tie_threshold,
binary_split=self.binary_split,
stop_mem_management=self.stop_mem_management,
remove_poor_atts=self.remove_poor_atts,
no_preprune=self.no_preprune,
leaf_prediction=self.leaf_prediction,
nb_threshold=self.nb_threshold,
nominal_attributes=self.nominal_attributes,
max_features=self.max_features,
random_state=self.random_state),
instances_seen=self.instances_seen,
drift_detection_method=self.drift_detection_method,
warning_detection_method=self.warning_detection_method,
is_background_learner=False) for i in range(self.n_estimators)
]
def predict_proba(self, X, max_score):
""" Predicts probabilities of all label of the X instance(s)
Parameters
----------
X: numpy.ndarray of shape (n_samples, n_features)
Samples for which we want to predict the labels.
max_score: float
The maximum score of an instance could have in the tree.
Returns
-------
numpy.array
Predicted the probabilities of all the labels for all instances in X.
"""
r, _ = get_dimensions(X)
predictions = []
for i in range(r):
votes = copy.deepcopy(self.get_votes_for_instance(X[i], max_score))
y_proba = np.zeros(int(max(votes.keys())) + 1)
for key, value in votes.items():
y_proba[int(key)] = value
predictions.append(y_proba)
return np.array(predictions)
def normalize_sample(self, X):
"""Normalize the features in order to have the same influence during the
process of training.
Parameters
----------
X: np.array
features.
Returns
-------
np.array:
normalized samples
"""
if self.examples_seen <= 1:
_, c = get_dimensions(X)
return np.zeros((c + 1), dtype=np.float64)
mean = self.sum_of_attribute_values / self.examples_seen
variance = (self.sum_of_attribute_squares -
(self.sum_of_attribute_values**2) / self.examples_seen) / (
self.examples_seen - 1)
sd = np.sqrt(variance,
out=np.zeros_like(variance),
where=variance >= 0.0)
normalized_sample = np.zeros(X.shape[0] + 1, dtype=np.float64)
np.divide(X - mean, sd, where=sd != 0, out=normalized_sample[:-1])
# Augments sample with the bias input signal (or y intercept for
# each target)
normalized_sample[-1] = 1.0
return normalized_sample
def partial_fit(self, X, y, classes=None, sample_weight=None):
""" Partially (incrementally) fit the model.
Parameters
----------
X : numpy.ndarray of shape (n_samples, n_features)
The features to train the model.
y: numpy.ndarray of shape (n_samples)
An array-like with the labels of all samples in X.
classes: numpy.ndarray, optional (default=None)
Array with all possible/known classes. Usage varies depending on
the learning method.
sample_weight: numpy.ndarray of shape (n_samples), optional (default=None)
Samples weight. If not provided, uniform weights are assumed.
Usage varies depending on the learning method.
Returns
-------
self
"""
r, c = get_dimensions(X)
if self._STMSamples is None:
self._STMSamples = np.empty(shape=(0, c))
self._LTMSamples = np.empty(shape=(0, c))
for i in range(r):
self._partial_fit(X[i, :], y[i])
return self
def partial_fit(self, X, y, sample_weight=None):
""" Partially (incrementally) fit the model.
Parameters
----------
X: numpy.ndarray of shape (n_samples, n_features)
The data upon which the algorithm will create its model.
y: numpy.ndarray of shape (n_samples)
An array-like containing the target values for all
samples in X.
sample_weight: Not used.
Returns
-------
KNNRegressor
self
Notes
-----
For the K-Nearest Neighbors regressor, fitting the model is the
equivalent of inserting the newer samples in the observed window,
and if the size_limit is reached, removing older results.
"""
r, c = get_dimensions(X)
for i in range(r):
self.data_window.add_sample(X=X[i], y=y[i])
return self
def predict(self, X):
"""
Predict the class label for sample X
Parameters
----------
X: numpy.ndarray
An array of shape (n_samples, n_features) with the samples to
predict the class label for.
Returns
-------
numpy.ndarray
A 1D array of shape (, n_samples), containing the
predicted class labels for all instances in X.
"""
if self._ensemble:
if self.update_strategy == self._REPLACE_STRATEGY:
trees_in_ensemble = sum(i is not None for i in self._ensemble)
else: # self.update_strategy == self._PUSH_STRATEGY
trees_in_ensemble = len(self._ensemble)
if trees_in_ensemble > 0:
d_test = xgb.DMatrix(X)
for i in range(trees_in_ensemble - 1):
margins = self._ensemble[i].predict(d_test,
output_margin=True)
d_test.set_base_margin(margin=margins)
predicted = self._ensemble[trees_in_ensemble -
1].predict(d_test)
return np.array(predicted > 0.5).astype(int)
# Ensemble is empty, return default values (0)
return np.zeros(get_dimensions(X)[0])
def _partial_fit(self, X, y):
if self._first_run:
self._X_buffer = np.array([]).reshape(0, get_dimensions(X)[1])
self._y_buffer = np.array([])
self._first_run = False
self._X_buffer = np.concatenate((self._X_buffer, X))
self._y_buffer = np.concatenate((self._y_buffer, y))
while self._X_buffer.shape[0] >= self.window_size:
self._train_on_mini_batch(X=self._X_buffer[0:self.window_size, :],
y=self._y_buffer[0:self.window_size])
delete_idx = [i for i in range(self.window_size)]
self._X_buffer = np.delete(self._X_buffer, delete_idx, axis=0)
self._y_buffer = np.delete(self._y_buffer, delete_idx, axis=0)
# Check window size and adjust it if necessary
self._adjust_window_size()
# Support for concept drift
if self.detect_drift:
correctly_classifies = self.predict(X) == y
# Check for warning
self._drift_detector.add_element(int(not correctly_classifies))
# Check if there was a change
if self._drift_detector.detected_change():
# Reset window size
self._reset_window_size()
if self.update_strategy == self._REPLACE_STRATEGY:
self._model_idx = 0
def predict(self, X):
"""Predicts the target value using mean class or the perceptron.
Parameters
----------
X: numpy.ndarray of shape (n_samples, n_features)
Samples for which we want to predict the labels.
Returns
-------
numpy.ndarray
Predicted target values.
"""
predictions = []
if self.samples_seen > 0 and self._tree_root is not None:
r, _ = get_dimensions(X)
for i in range(r):
node = self._tree_root.filter_instance_to_leaf(X[i], None,
-1).node
if node.is_leaf():
predictions.append(node.predict_one(X[i], tree=self))
else:
# The instance sorting ended up in a Split Node, since no branch was found
# for some of the instance's features. Use the mean prediction in this case
predictions.append(node.stats[1] / node.stats[0])
else:
# Model is empty
predictions.append(0.0)
return np.asarray(predictions)
def learn_from_instance(self, X, y, weight, ht):
"""Update the node with the provided instance.
Parameters
----------
X: numpy.ndarray of length equal to the number of features.
Instance attributes for updating the node.
y: int
Instance class.
weight: float
Instance weight.
ht: HoeffdingTreeClassifier
Hoeffding Tree to update.
"""
try:
self._observed_class_distribution[y] += weight
except KeyError:
self._observed_class_distribution[y] = weight
self._observed_class_distribution = dict(
sorted(self._observed_class_distribution.items()))
if self.list_attributes.size == 0:
self.list_attributes = self._sample_features(get_dimensions(X)[1])
for i in self.list_attributes:
try:
obs = self._attribute_observers[i]
except KeyError:
if ht.nominal_attributes is not None and i in ht.nominal_attributes:
obs = NominalAttributeClassObserver()
else:
obs = NumericAttributeClassObserverGaussian()
self._attribute_observers[i] = obs
obs.observe_attribute_class(X[i], int(y), weight)
def predict_proba(self, X):
""" Estimates the probability of each sample in X belonging to each of the class-labels.
Parameters
----------
X : Numpy.ndarray of shape (n_samples, n_features)
The matrix of samples one wants to predict the class probabilities for.
Returns
-------
A numpy.ndarray of shape (n_samples, n_labels), in which each outer entry is associated
with the X entry of the same index. And where the list in index [i] contains
len(self.target_values) elements, each of which represents the probability that
the i-th sample of X belongs to a certain class-label.
"""
predictions = deque()
r, _ = get_dimensions(X)
if self._observed_class_distribution == {}:
# Model is empty, all classes equal, default to zero
return np.zeros((r, 1))
else:
for i in range(r):
votes = do_naive_bayes_prediction(X[i], self._observed_class_distribution,
self._attribute_observers)
sum_values = sum(votes.values())
if self._classes is not None:
y_proba = np.zeros(int(max(self._classes)) + 1)
else:
y_proba = np.zeros(int(max(votes.keys())) + 1)
for key, value in votes.items():
y_proba[int(key)] = value / sum_values if sum_values != 0 else 0.0
predictions.append(y_proba)
return np.array(predictions)
def partial_fit(self, X, y, classes=None, sample_weight=None):
""" Partially (incrementally) fit the model.
Parameters
----------
X : numpy.ndarray of shape (n_samples, n_features)
The features to train the model.
y: numpy.ndarray of shape (n_samples)
An array-like with the class labels of all samples in X.
classes: None
Not used by this method.
sample_weight: None
Not used by this method.
Returns
-------
self
"""
row_cnt, _ = X.shape
if self.samples_seen == 0:
self._random_state = check_random_state(self.random_state)
self.n_features = get_dimensions(X)[1]
self.build_trees()
for i in range(row_cnt):
self._partial_fit(X[i], y[i])
return self
def partial_fit(self, X, y, classes=None, sample_weight=None):
""" Partially (incrementally) fit the model.
Parameters
----------
X : numpy.ndarray of shape (n_samples, n_features)
The features to train the model.
y: numpy.ndarray of shape (n_samples)
An array-like with the class labels of all samples in X.
classes: numpy.ndarray, optional (default=None)
No used.
sample_weight: numpy.ndarray of shape (n_samples), optional \
(default=None)
Samples weight. If not provided, uniform weights are assumed.
Usage varies depending on the learning method.
Returns
-------
self
"""
n_rows, n_cols = get_dimensions(X)
if sample_weight is None:
sample_weight = np.ones(n_rows)
for i in range(n_rows):
self._partial_fit(np.asarray([X[i]]), np.asarray([y[i]]),
classes=classes, sample_weight=np.asarray([sample_weight[i]]))
return self
def predict(self, X):
""" Predict classes for the passed data.
Parameters
----------
X : numpy.ndarray of shape (n_samples, n_features)
The set of data samples to predict the class labels for.
Returns
-------
A numpy.ndarray with all the predictions for the samples in X.
"""
r, _ = get_dimensions(X)
predictions = []
for i in range(r):
y_proba = self.predict_proba(X)
if y_proba is None:
# Ensemble is empty, all classes equal, default to zero
predictions.append(0)
else:
# if prediction of this instance being anomaly is greater than the threshold defined,
# then this instance is classified as an anomaly.
if y_proba[0][1] > self.anomaly_threshold:
predictions.append(1)
else:
predictions.append(0)
return np.asarray(predictions)
def predict(self, X):
""" predict
The predict function will average the predictions from all its learners
to find the most likely prediction for the sample matrix X.
Parameters
----------
X: Numpy.ndarray of shape (n_samples, n_features)
A matrix of the samples we want to predict.
Returns
-------
numpy.ndarray
A numpy.ndarray with the label prediction for all the samples in X.
"""
r, c = get_dimensions(X)
proba = self.predict_proba(X)
predictions = []
if proba is None:
return None
for i in range(r):
predictions.append(np.argmax(proba[i]))
return np.asarray(predictions)
def _init_ensemble(self, X):
self._set_max_features(get_dimensions(X)[1])
# Generate a different random seed per tree
random_states = self._random_state.randint(0,
4294967295,
size=self.n_estimators,
dtype='u8')
self.ensemble = [
ARFRegBaseLearner(
index_original=i,
estimator=ARFHoeffdingTreeRegressor(
max_byte_size=self.max_byte_size,
memory_estimate_period=self.memory_estimate_period,
grace_period=self.grace_period,
split_confidence=self.split_confidence,
tie_threshold=self.tie_threshold,
binary_split=self.binary_split,
stop_mem_management=self.stop_mem_management,
remove_poor_atts=self.remove_poor_atts,
no_preprune=self.no_preprune,
leaf_prediction=self.leaf_prediction,
nominal_attributes=self.nominal_attributes,
learning_ratio_perceptron=self.learning_ratio_perceptron,
learning_ratio_decay=self.learning_ratio_decay,
learning_ratio_const=self.learning_ratio_const,
max_features=self.max_features,
random_state=random_states[i]),
instances_seen=self.instances_seen,
drift_detection_method=self.drift_detection_method,
warning_detection_method=self.warning_detection_method,
performance_metric=self.weighted_vote_strategy,
drift_detection_criteria=self.drift_detection_criteria,
is_background_learner=False) for i in range(self.n_estimators)
]
def predict(self, X):
"""Predicts the target value using mean class or the perceptron.
Parameters
----------
X: numpy.ndarray of shape (n_samples, n_features)
Samples for which we want to predict the labels.
Returns
-------
list
Predicted target values.
"""
r, _ = get_dimensions(X)
try:
predictions = np.zeros((r, self._n_targets), dtype=np.float64)
except AttributeError:
warnings.warn("Calling predict without previously fitting the model at least once.\n"
"Predictions will default to a column array filled with zeros.")
return np.zeros((r, 1))
for i in range(r):
node = self._tree_root.filter_instance_to_leaf(X[i], None, -1).node
if isinstance(node, SplitNode):
# If not leaf, use mean as response
predictions[i, :] = node.stats[1] / node.stats[0] if len(node.stats) > 0 else 0.0
continue
predictions[i, :] = node.predict_one(X[i], tree=self)
return predictions
def partial_fit(self, X, y, classes=None, sample_weight=None):
"""
Fit an array of observations. Splits input into individual observations and
passes to a helper function _partial_fit. Randomly weights observations depending on
Config.
"""
if self.classes is None and classes is not None:
self.classes = classes
if y is not None:
row_cnt, _ = get_dimensions(X)
if sample_weight is None:
sample_weight = np.ones(row_cnt)
if row_cnt != len(sample_weight):
raise ValueError(
'Inconsistent number of instances ({}) and weights ({}).'.
format(row_cnt, len(sample_weight)))
for i in range(row_cnt):
if sample_weight[i] != 0.0:
self._train_weight_seen_by_model += sample_weight[i]
self.ex += 1
if self.rand_weights and self.poisson >= 1:
# Use weights similar to ARF.
# This just uses similar avg grace period etc
# without having to calculate those parameters.
k = self.poisson
sample_weight[i] = k
self._partial_fit(X[i], y[i], sample_weight[i])
def predict(self, X):
predictions = deque()
r, _ = get_dimensions(X)
y_pred = self.stream.current_sample_y
for i in range(r):
predictions.append(y_pred)
return np.array(predictions)
def _partial_fit(self, X, y, classes=None, sample_weight=None):
self._n_samples_seen += 1
_, n_features = get_dimensions(X)
if not self.ensemble:
self._init_ensemble(n_features)
for i in range(len(self.ensemble)):
# Get prediction for instance
y_pred = np.asarray([np.argmax(self.ensemble[i].predict_proba(X))])
# Update performance evaluator
self.ensemble[i].performance_evaluator.add_result(y[0], y_pred[0], sample_weight[0])
# Train using random subspaces without resampling,
# i.e. all instances are used for training.
if self.training_method == self._TRAIN_RANDOM_SUBSPACES:
self.ensemble[i].partial_fit(X=X, y=y, classes=classes,
sample_weight=np.asarray([1.]),
n_samples_seen=self._n_samples_seen,
random_state=self._random_state)
# Train using random patches or resampling,
# thus we simulate online bagging with Poisson(lambda=...)
else:
k = self._random_state.poisson(lam=self.lam)
if k > 0:
self.ensemble[i].partial_fit(X=X, y=y, classes=classes,
sample_weight=np.asarray([k]),
n_samples_seen=self._n_samples_seen,
random_state=self._random_state)
def predict(self, X):
predictions = deque()
r, _ = get_dimensions(X)
y_proba = np.zeros((r, len(Data.classes)))
for i in range(r):
session_vector = Data.session_vector[-self.max_session_size:]
for pos, y_o_idx in enumerate(session_vector):
weight = self.w_mc if y_o_idx == session_vector[-1] else 1
y_proba_current = self.ht.predict_proba(np.array([[y_o_idx]]))
y_proba_current *= weight / (len(session_vector) - pos)
y_proba += y_proba_current
y_proba[i][Data.session_vector[-1]] = 0.0
nonzero = np.flatnonzero(y_proba[i])
if len(nonzero > 0):
sorted_desc = np.argsort(y_proba[i][nonzero])[::-1]
sorted_ids = nonzero[sorted_desc]
if not Data.allow_reminders:
sorted_ids = sorted_ids[~np.isin(sorted_ids, Data.
session_vector)]
if not Data.allow_repeated:
session = X[i, Data.sid]
sorted_ids = sorted_ids[
~np.isin(sorted_ids, self._rec_tracker[session])]
self._rec_tracker[session].extend(
sorted_ids[:Data.rec_size])
y_pred = Data.classes[sorted_ids[:Data.rec_size]]
else:
y_pred = np.array([])
predictions.append(y_pred)
return np.array(predictions)