Hands-On Natural Language Processing with Python

Class constructor

We will define the constructor for initializing the memory network's initializer object, optimizer object, mini-batch size, and so on. We will also write high-level TensorFlow operations for the loss, prediction, and training. All of them hinge on the _inference method, which we will implement in the following sections:

class MemoryNetwork(object):    def __init__(self, sentence_size, vocab_size, candidates_size,                  candidates_vec, embedding_size, hops,                  initializer=tf.random_normal_initializer(stddev=0.1),                  optimizer=tf.train.AdamOptimizer(learning_rate=0.01),                 session=tf.Session()):        self._hops = hops        self._candidates_vec = candidates_vec                # Define placeholders for inputs to the model        self._facts = tf.placeholder( tf.int32, [None, ...