def fit(self, train_X, train_Y, val_X=None, val_Y=None, graph=None):
"""Fit the model to the data.
Parameters
----------
train_X : array_like, shape (n_samples, n_features)
Training data.
train_Y : array_like, shape (n_samples, n_classes)
Training labels.
val_X : array_like, shape (N, n_features) optional, (default = None).
Validation data.
val_Y : array_like, shape (N, n_classes) optional, (default = None).
Validation labels.
graph : tf.Graph, optional (default = None)
Tensorflow Graph object.
Returns
-------
"""
if len(train_Y.shape) != 1:
num_classes = train_Y.shape[1]
else:
raise Exception("Please convert the labels with one-hot encoding.")
g = graph if graph is not None else self.tf_graph
with g.as_default():
# Build model
self.build_model(train_X.shape[1], num_classes)
with tf.Session() as self.tf_session:
# Initialize tf stuff
summary_objs = tf_utils.init_tf_ops(self.tf_session)
self.tf_merged_summaries = summary_objs[0]
self.tf_summary_writer = summary_objs[1]
self.tf_saver = summary_objs[2]
# Train model
self._train_model(train_X, train_Y, val_X, val_Y)
# Save model
_weight = self.tf_session.run(self.encoding_w_)
for matrix in _weight:
np.savetxt(utils.get_root_path(False) + '/save/' + str(matrix.shape[0]) + 'to' + str(matrix.shape[1]) + '.txt', matrix)
self.tf_saver.save(self.tf_session, self.model_path)
def fit(self, train_X, train_Y=None, val_X=None, val_Y=None, graph=None):
"""Fit the model to the data.
Parameters
----------
train_X : array_like, shape (n_samples, n_features)
Training data.
train_Y : array_like, shape (n_samples, n_features)
Training reference data.
val_X : array_like, shape (N, n_features) optional, (default = None).
Validation data.
val_Y : array_like, shape (N, n_features) optional, (default = None).
Validation reference data.
graph : tf.Graph, optional (default = None)
Tensorflow Graph object.
Returns
-------
"""
g = graph if graph is not None else self.tf_graph
with g.as_default():
# Build model
self.build_model(train_X.shape[1])
with tf.Session() as self.tf_session:
# Initialize tf stuff
summary_objs = tf_utils.init_tf_ops(self.tf_session)
self.tf_merged_summaries = summary_objs[0]
self.tf_summary_writer = summary_objs[1]
self.tf_saver = summary_objs[2]
# Train model
self._train_model(train_X, train_Y, val_X, val_Y)
# Save model
weight = self.tf_session.run(self.W)
np.savetxt(
root + '/save/' + str(weight.shape[0]) + 'to' +
str(weight.shape[1]) + '.txt', weight)
self.tf_saver.save(self.tf_session, self.model_path)
def fit(self, train_X, train_Y, val_X=None, val_Y=None, graph=None):
"""Fit the model to the data.
Parameters
----------
train_X : array_like, shape (n_samples, n_features)
Training data.
train_Y : array_like, shape (n_samples, n_classes)
Training labels.
val_X : array_like, shape (N, n_features) optional, (default = None).
Validation data.
val_Y : array_like, shape (N, n_classes) optional, (default = None).
Validation labels.
graph : tf.Graph, optional (default = None)
Tensorflow Graph object.
Returns
-------
"""
if len(train_Y.shape) != 1:
num_classes = train_Y.shape[1]
else:
raise Exception("Please convert the labels with one-hot encoding.")
g = graph if graph is not None else self.tf_graph
with g.as_default():
# Build model
self.build_model(train_X.shape[1], num_classes)
with tf.Session() as self.tf_session:
# Initialize tf stuff
summary_objs = tf_utils.init_tf_ops(self.tf_session)
self.tf_merged_summaries = summary_objs[0]
self.tf_summary_writer = summary_objs[1]
self.tf_saver = summary_objs[2]
# Train model
self._train_model(train_X, train_Y, val_X, val_Y)
# Save model
self.tf_saver.save(self.tf_session, self.model_path)