This example uses a distributed version of the the NASA C-MAPSS (Commercial
Modular Aero-Propulsion System Simulation) Aircraft Engine dataset.  Each party
owns engine sensor data over time until the engine reaches failure. The goal is
to train a model that will predict the Remaining Useful Life (RUL) of an engine
at a given time step.

There are two examples using this dataset. One example uses a feed forward
neural network (example/CMAPSS_NN) and the other reshapes the data in
preprocessing so that a convolutional neural network can be used
(example/CMAPSS_CNN).  These examples demonstrate how to train models that
predict a continuous dependent variable.

In the preprocess script the RUL values are calculated for every sample of data
in the train and test sets. The "id" and "cycle" columns are dropped, the data
is scaled and for the CNN model the data is segmented into 24 X 30 windows of
data to be processed by the convolutional layers.

The training of these networks defaults to only 5 epochs and training against
datasets one and three. To get a more accurate model the training will need to
be increased to 250+ epochs.
