def setup(self):
tf.reset_default_graph()
self.x = tf.placeholder(dtype=tf.float32,
shape=[None,self.no_of_features],
name="input")
self.y = tf.placeholder(dtype=tf.float32, shape=[None, 1], name="labels")
self.lr = tf.placeholder("float", shape=[])
self.is_train = tf.placeholder(tf.bool, shape=[])
if self.logits==None:
self.logits=self.get_model(self.x,self.is_train)
else:
self.logits=self.logits(self.x,self.is_train)
with tf.name_scope('Output'):
self.cross_entropy = ops.get_loss(self.logits, self.y, self.loss_type)
if self.regularization_type != None:
self.cross_entropy = ops.get_regularization(self.cross_entropy, self.regularization_type,
self.regularization_coefficient)
self.prediction = self.logits
tf.summary.scalar("Cross_Entropy", self.cross_entropy)
with tf.name_scope('Optimizer'):
if self.optimizer==None:
# learningRate = tf.train.exponential_decay(learning_rate=learning_rate, global_step=1,
# decay_steps=shape[0], decay_rate=0.97, staircase=True,
# name='Learning_Rate')
# optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cross_entropy)
# optimizer = tf.train.MomentumOptimizer(lr, .9, use_nesterov=True).minimize(cross_entropy)
self.optimizer = tf.train.AdamOptimizer(self.lr).minimize(self.cross_entropy)
# optimizer = tf.train.AdadeltaOptimizer(lr).minimize(cross_entropy)
self.session = tf.InteractiveSession()
return
def setup(self):
tf.reset_default_graph()
self.test_result = []
self.train_result = []
self.x = tf.placeholder(dtype=tf.float32,
shape=[None, self.no_of_features],
name="input")
self.y = tf.placeholder(dtype=tf.float32,
shape=[None, self.no_of_classes],
name="labels")
self.lr = tf.placeholder("float", shape=[])
self.is_train = tf.placeholder(tf.bool, shape=[])
self.logits = self.model(self.x, self.is_train)
with tf.name_scope('Output'):
self.cross_entropy = ops.get_loss(self.logits, self.y,
self.loss_type)
hypothesis = tf.nn.softmax(self.logits, name="softmax")
self.prediction = tf.argmax(hypothesis, 1, name='Prediction')
correct_prediction = tf.equal(self.prediction,
tf.argmax(self.y, 1),
name='Correct_prediction')
self.accuracy = tf.reduce_mean(tf.cast(correct_prediction,
tf.float32),
name='Accuracy')
tf.summary.scalar("Cross_Entropy", self.cross_entropy)
tf.summary.scalar("Accuracy", self.accuracy)
with tf.name_scope('Optimizer'):
# learningRate = tf.train.exponential_decay(learning_rate=learning_rate, global_step=1,
# decay_steps=shape[0], decay_rate=0.97, staircase=True,
# name='Learning_Rate')
if (self.optimizer_type == 'adadelta'):
self.optimizer = tf.train.AdadeltaOptimizer(self.lr).minimize(
self.cross_entropy)
elif (self.optimizer_type == 'gdecent'):
self.optimizer = tf.train.GradientDescentOptimizer(
self.lr).minimize(self.cross_entropy)
elif (self.optimizer_type == 'momentum'):
self.optimizer = tf.train.MomentumOptimizer(
self.lr, .9,
use_nesterov=True).minimize(self.cross_entropy)
else:
self.optimizer = tf.train.AdamOptimizer(self.lr).minimize(
self.cross_entropy)
return
def setup(self):
tf.reset_default_graph()
self.x = tf.placeholder(
dtype=tf.float32,
shape=[None, self.sequence_length, self.sequence_dimensions],
name="input")
self.y = tf.placeholder(dtype=tf.float32,
shape=[None, self.no_of_classes],
name="labels")
self.lr = tf.placeholder("float", shape=[])
self.is_train = tf.placeholder(tf.bool, shape=[])
if self.logits == None:
self.logits = self.get_model(self.x, self.is_train)
else:
self.logits = self.logits(self.x, self.is_train)
with tf.name_scope('Output'):
self.cross_entropy = ops.get_loss(self.logits, self.y,
self.loss_type)
if self.regularization_type != None:
self.cross_entropy = ops.get_regularization(
self.cross_entropy, self.regularization_type,
self.regularization_coefficient)
self.probability = tf.nn.softmax(self.logits, name="softmax")
self.prediction = tf.argmax(self.probability, 1, name='Prediction')
correct_prediction = tf.equal(self.prediction,
tf.argmax(self.y, 1),
name='Correct_prediction')
self.accuracy = tf.reduce_mean(tf.cast(correct_prediction,
tf.float32),
name='Accuracy')
tf.summary.scalar("Cross_Entropy", self.cross_entropy)
tf.summary.scalar("Accuracy", self.accuracy)
with tf.name_scope('Optimizer'):
if self.optimizer == None:
# learningRate = tf.train.exponential_decay(learning_rate=learning_rate, global_step=1,
# decay_steps=shape[0], decay_rate=0.97, staircase=True,
# name='Learning_Rate')
# optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cross_entropy)
# optimizer = tf.train.MomentumOptimizer(lr, .9, use_nesterov=True).minimize(cross_entropy)
self.optimizer = tf.train.AdamOptimizer(self.lr).minimize(
self.cross_entropy)
# optimizer = tf.train.AdadeltaOptimizer(lr).minimize(cross_entropy)
self.session = tf.InteractiveSession()
return
def setup(self):
tf.reset_default_graph()
self.x = tf.placeholder(dtype=tf.float32,
shape=[None, self.no_of_features],
name="input")
self.lr = tf.placeholder("float", shape=[])
self.is_train = tf.placeholder(tf.bool, shape=[])
if self.encoder_op==None:
self.encoder_op=self.get_encoder(self.x,self.is_train)
else:
self.encoder_op=self.encoder_op(self.x,self.is_train)
if self.decoder_op==None:
self.decoder_op=self.get_decoder(self.x,self.is_train)
else:
self.decoder_op=self.decoder_op(self.x,self.is_train)
with tf.name_scope('Output'):
self.cross_entropy = ops.get_loss(self.decoder_op, self.x, self.loss_type)
if self.regularization_type != None:
self.cross_entropy = ops.get_regularization(self.cross_entropy, self.regularization_type,
self.regularization_coefficient)
self.cosine_similarity=distance_metric.cosine_similarity(self.decoder_op,self.x)
print(self.cosine_similarity.get_shape().as_list())
tf.summary.scalar("Cross_Entropy", self.cross_entropy)
tf.summary.scalar("Accuracy", tf.reduce_mean(self.cosine_similarity))
with tf.name_scope('Optimizer'):
if self.optimizer==None:
# learningRate = tf.train.exponential_decay(learning_rate=learning_rate, global_step=1,
# decay_steps=shape[0], decay_rate=0.97, staircase=True,
# name='Learning_Rate')
# optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cross_entropy)
# optimizer = tf.train.MomentumOptimizer(lr, .9, use_nesterov=True).minimize(cross_entropy)
self.optimizer = tf.train.AdamOptimizer(self.lr).minimize(self.cross_entropy)
# optimizer = tf.train.AdadeltaOptimizer(lr).minimize(cross_entropy)
self.session = tf.InteractiveSession()
return