notes.txt

- View GCN in Tensorboard

- Train TAGCN for small k, c, n

- Build small TAGCN using the data




optimizer = tf.train.AdamOptimizer(learning_rate=FLAGS.learning_rate)


loss += FLAGS.weight_decay * tf.nn.l2_loss(var)


loss += masked_softmax_cross_entropy(self.outputs, self.placeholders['labels'],
		                                          self.placeholders['labels_mask'])


accuracy = masked_accuracy(self.outputs, self.placeholders['labels'],
                                        self.placeholders['labels_mask'])

opt_op = optimizer.minimize(self.loss)


# tf.sess

for epoch in range(FLAGS.epochs):


	outs = sess.run([opt_op, loss, accuracy], feed_dict=feed_dict)


	print("Epoch:", '%04d' % (epoch + 1), "train_loss=", "{:.5f}".format(outs[1]),
          "train_acc=", "{:.5f}".format(outs[2]))




#tf.nn.softmax(self.outputs)