import tensorflow as tf
def sparse_dropout(x, keep_prob, noise_shape=None, seed=None, name=None):
'''borrowed logic and implementation from https://github.com/tensorflow/tensorflow/blob/r1.4/tensorflow/python/ops/nn_ops.py'''
# Skipping all the assertions
if (keep_prob == 1):
return x
# uniform [keep_prob, 1.0 + keep_prob)
random_tensor = keep_prob
random_tensor += tf.random_uniform(noise_shape,
seed=seed,
dtype=x.dtype)
# 0. if [keep_prob, 1.0) and 1. if [1.0, 1.0 + keep_prob)
binary_tensor = tf.floor(random_tensor)
# Typecase necessary as mentioned on https://www.tensorflow.org/api_docs/python/tf/sparse_retain
binary_tensor = tf.cast(binary_tensor, dtype=tf.bool)
ret = tf.sparse_retain(x, binary_tensor)
ret = ret * (1 / keep_prob)
return ret
def get_dotproduct_op(sparse_features=True):
if (sparse_features):
return tf.sparse_tensor_dense_matmul
else:
return tf.matmul
def get_transpose_op(sparse_features=True):
if (sparse_features):
return tf.sparse_transpose
else:
return tf.transpose
