def call(self, inputs, mask=None):
'''Logic borrowed from: https://github.com/fchollet/keras/blob/master/keras/layers/core.py
'''
dotproduct_op = get_dotproduct_op(sparse_features=self.sparse_features)
if (self.sparse_features):
inputs = sparse_dropout(inputs, keep_prob=1 - self.dropout_rate, noise_shape=(self.num_elements,))
else:
inputs = tf.nn.dropout(inputs, keep_prob=1 - self.dropout_rate)
output = dotproduct_op(inputs, self.kernel)
output = tf.add(output, self.bias)
if self.activation is not None:
output = self.activation(output)
return output
def call(self, inputs, mask=None):
'''Logic borrowed from: https://github.com/fchollet/keras/blob/master/keras/layers/core.py
'''
dotproduct = get_dotproduct(sparse_features=self.sparse_features)
sparse_dotproduct = get_dotproduct(sparse_features=True)
if (self.sparse_features):
inputs = sparse_dropout(inputs,
keep_prob=1 - self.dropout_rate,
noise_shape=(self.num_elements, ))
else:
inputs = tf.nn.dropout(inputs, keep_prob=1 - self.dropout_rate)
supports = []
for i in range(len(self.supports)):
supports.append(
sparse_dotproduct(self.supports[i],
dotproduct(inputs, self.support_kernels[i])))
output = tf.add_n(supports)
output = tf.add(output, self.bias)
if self.activation is not None:
output = self.activation(output)
return output