This example illustrates building a neural network to predict customer
behaviors off of financial data from three independent institutions -- "SAN",
"JPM" and "PNB".  The datasets are utilized to jointly train a deep learning
Convolutional Neural Network (CNN) without exposing any private data.  The
trained network is then used to perform inference both independently of the
TripleBlind ecosystem and also for secure remote inferences through TripleBlind.

The training data comes from three different banks using the same banking core,
so the format is identical.  The data also includes a "target" column which
is used to identify customers exhibiting the behavior we are training the
network to recognize.

At the end of the training sequence, the model is owned by the organization
which initiated the training.  This trained model is available as an Asset in
the TripleBlind ecosystem which the owner can download for independent usage as
a PyTorch model.  Additionally, the owner can allow others to utilize this
algorithm Asset remotely without losing control of the model while remaining
isolated from that user's private data.
