def model(dim_embedding, batch_s):
# Building model
input_layer = tflearn.input_data(shape=[None, DIM_FEATURES], dtype=tf.float64, name='input')
input_text = input_layer[:, 0:DIM_TEXT]
input_citation = input_layer[:, DIM_TEXT:DIM_TEXT + DIM_CITATION]
input_section = input_layer[:, DIM_TEXT + DIM_CITATION:DIM_TEXT + DIM_CITATION + NUM_SEC]
input_subsection = input_layer[:, DIM_TEXT + DIM_CITATION + NUM_SEC:]
textual_inf = tflearn.embedding(input_text, input_dim=142698, output_dim=dim_embedding, name='word_embedding')
section_embedding = tf.Variable(tf.random_normal([NUM_SEC, 128]), name='group_embedding')
subsection_embedding = tf.Variable(tf.random_normal([NUM_SUBS, 128]), name='group_embedding')
# section_embedding = tflearn.embedding(input_section, input_dim=NUM_SEC, output_dim=dim_embedding, name='section_embedding')
# subsection_embedding = tflearn.embedding(input_subsection, input_dim=NUM_SUBS, output_dim=dim_embedding,
# name='subsection_embedding')
class_embedding = tf.Variable(tf.random_normal([NUM_CLASS, dim_embedding]), name='class_embedding')
textual_embedding = tflearn.lstm(textual_inf, dim_embedding, dropout=0.8, name='lstm')
network = tf.concat([textual_embedding, input_citation], 1)
network = tflearn.fully_connected(network, dim_embedding, activation='elu')
network = _cat_weighted(network, section_embedding)
network = _cat_weighted(network, subsection_embedding)
network = tf.matmul(network, tf.transpose(class_embedding), name='class_weight')
# network = tflearn.softmax(network)
network = tflearn.sigmoid(network)
# network = tflearn.regression(network, optimizer='adam', learning_rate=0.001, loss='categorical_crossentropy')
# tflearn.objectives.binary_crossentropy(y_pred, y_true) Computes sigmoid cross entropy between y_pred (logits) and y_true (labels).
# network = tflearn.regression(network, optimizer='adam', learning_rate=0.001, dtype=tf.float64, batch_size=batch_s,
# loss='binary_crossentropy')
network = tflearn.regression(network, optimizer='adam', learning_rate=0.001, dtype=tf.float64, batch_size=batch_s,
loss=loss_multilabel())
return network
def model(self):
# Building model
# input_layer = tflearn.input_data(shape=[None, self.DIM_FEATURES], dtype=tf.float64, name='input')
# input_text = input_layer[:, 0:self.DIM_TEXT]
# input_citation = input_layer[:, self.DIM_TEXT:self.DIM_TEXT + self.DIM_CITATION]
# self.input_section = input_layer[:,
# self.DIM_TEXT + self.DIM_CITATION:self.DIM_TEXT + self.DIM_CITATION + self.NUM_SEC]
# self.input_subsection = input_layer[:, self.DIM_TEXT + self.DIM_CITATION + self.NUM_SEC:]
textual_inf = tflearn.embedding(self.input_text, input_dim=self.vocab_size, output_dim=self.embedding_size,
name='word_embedding')
section_embedding = tf.Variable(tf.random_normal([self.NUM_SEC, self.embedding_size]), name='group_embedding')
subsection_embedding = tf.Variable(tf.random_normal([self.NUM_SUBS, self.embedding_size]),
name='group_embedding')
# section_embedding = tflearn.embedding(input_section, input_dim=NUM_SEC, output_dim=dim_embedding, name='section_embedding')
# subsection_embedding = tflearn.embedding(input_subsection, input_dim=NUM_SUBS, output_dim=dim_embedding,
# name='subsection_embedding')
class_embedding = tf.Variable(tf.random_normal([self.NUM_CLASS, self.embedding_size]), name='class_embedding')
textual_embedding = tflearn.lstm(textual_inf, self.lstm_hidden_size, dropout=0.8, name='lstm')
network = tf.concat([textual_embedding, self.input_citation], 1)
network = tflearn.fully_connected(network, self.embedding_size, activation='elu')
self.weight_section, network = self._cat_weighted(network, section_embedding)
self.weight_subsection, network = self._cat_weighted(network, subsection_embedding)
self.weight_class = tf.matmul(network, tf.transpose(class_embedding), name='class_weight')
network = tflearn.sigmoid(self.weight_class)
return network
def encoder(inputs, hidden_layer):
nb_feature = 64
net = tflearn.conv_2d(inputs, 16, 3, strides=2)
net = tflearn.batch_normalization(net)
net = tflearn.elu(net)
print "========================"
print "enc-L1", net.get_shape()
print "========================"
net = tflearn.conv_2d(net, 16, 3, strides=1)
net = tflearn.batch_normalization(net)
net = tflearn.elu(net)
print "========================"
print "enc-L2", net.get_shape()
print "========================"
net = tflearn.conv_2d(net, 32, 3, strides=2)
net = tflearn.batch_normalization(net)
net = tflearn.elu(net)
print "========================"
print "enc-L3", net.get_shape()
print "========================"
net = tflearn.conv_2d(net, 32, 3, strides=1)
net = tflearn.batch_normalization(net)
net = tflearn.elu(net)
print "========================"
print "enc-L4", net.get_shape()
print "========================"
net = tflearn.flatten(net)
#net = tflearn.fully_connected(net, nb_feature,activation="sigmoid")
net = tflearn.fully_connected(net, nb_feature)
h = net.W
print "Encoder Weights shape", h.get_shape()
net = tflearn.batch_normalization(net)
net = tflearn.sigmoid(net)
print "========================"
print "hidden", net.get_shape()
print "========================"
return [net, h]
def encoder(inputs,hidden_layer):
net = tflearn.conv_2d(inputs, 16, 3, strides=2)
net = tflearn.batch_normalization(net)
net = tflearn.elu(net)
print "========================"
print "enc-L1",net.get_shape()
print "========================"
net = tflearn.conv_2d(net, 16, 3, strides=1)
net = tflearn.batch_normalization(net)
net = tflearn.elu(net)
print "========================"
print "enc-L2",net.get_shape()
print "========================"
net = tflearn.conv_2d(net, 32, 3, strides=2)
net = tflearn.batch_normalization(net)
net = tflearn.elu(net)
print "========================"
print "enc-L3",net.get_shape()
print "========================"
net = tflearn.conv_2d(net, 32, 3, strides=1)
net = tflearn.batch_normalization(net)
net = tflearn.elu(net)
print "========================"
print "enc-L4",net.get_shape()
print "========================"
net = tflearn.flatten(net)
#net = tflearn.fully_connected(net, nb_feature,activation="sigmoid")
net = tflearn.fully_connected(net, nb_feature)
hidden_layer = net
net = tflearn.batch_normalization(net)
net = tflearn.sigmoid(net)
print "========================"
print "hidden",net.get_shape()
print "========================"
return [net,hidden_layer]
def encoder(self,inputs,hidden_layer): net = tflearn.conv_2d(inputs, 16, 3, strides=2) net = tflearn.batch_normalization(net) net = tflearn.elu(net) print "========================" print "enc-L1",net.get_shape() print "========================" net = tflearn.conv_2d(net, 16, 3, strides=1) net = tflearn.batch_normalization(net) net = tflearn.elu(net) print "========================" print "enc-L2",net.get_shape() print "========================" net = tflearn.conv_2d(net, 32, 3, strides=2) net = tflearn.batch_normalization(net) net = tflearn.elu(net) print "========================" print "enc-L3",net.get_shape() print "========================" net = tflearn.conv_2d(net, 32, 3, strides=1) net = tflearn.batch_normalization(net) net = tflearn.elu(net) print "========================" print "enc-L4",net.get_shape() print "========================" net = tflearn.flatten(net) #net = tflearn.fully_connected(net, nb_feature,activation="sigmoid") net = tflearn.fully_connected(net, self.instance_dim) hidden_layer = net net = tflearn.batch_normalization(net) net = tflearn.sigmoid(net) print "========================" print "hidden",net.get_shape() print "========================" return [net,hidden_layer]
def encoder(inputs):
net = tflearn.conv_2d(inputs, 16, 3, strides=2)
net = tflearn.batch_normalization(net)
net = tflearn.elu(net)
print "========================"
print "enc-L1",net.get_shape()
print "========================"
net = tflearn.conv_2d(net, 16, 3, strides=1)
net = tflearn.batch_normalization(net)
net = tflearn.elu(net)
print "========================"
print "enc-L2",net.get_shape()
print "========================"
net = tflearn.conv_2d(net, 32, 3, strides=2)
net = tflearn.batch_normalization(net)
net = tflearn.elu(net)
print "enc-L3",net.get_shape()
net = tflearn.conv_2d(net, 32, 3, strides=1)
net = tflearn.batch_normalization(net)
net = tflearn.elu(net)
print "========================"
print "enc-L4",net.get_shape()
print "========================"
net = tflearn.flatten(net)
net = tflearn.fully_connected(net, nb_feature)
hidden_layer = net
net = tflearn.batch_normalization(net)
net = tflearn.sigmoid(net)
print "========================"
print "enc-hidden_L",net.get_shape()
print "========================"
return [net,hidden_layer]
