Mastering Python for Finance - Second Edition

Phase 2 – training our model

The following steps describe the process of training our model:

1. Create a TensorFlow Session object to encapsulate the environment in which a neural network model operates:

In [ ]:    session = tf.InteractiveSession()

Here, we are specifying a session for use in an interactive context, in this case a Jupyter notebook. A regular tf.Session is non-interactive and requires an explicit Session object to be passed using the with keyword when running operations. InteractiveSession removes this need and is more convenient as it reuses the session variable.

2. TensorFlow requires that all global variables are to be initialized before training. Do this using the session.run command:

In [ ]: session.run(tf.global_variables_initializer()) ...