Hidden Layer

 The value of neural networks - and tensorflow, as a library - becomes

evident if we rerun the banknotes project, we worked up as a simple

case of supevised learning. On 500 banknotes, the computer caught the 

true ones, but might miss 5 straggler false bills. Not anymore.

Looking at the data as a neural network problem:

One has to install tensorflow with pip.

Code for a neural network:

Running for 20 iterations:

Running for 30:

There is an 8-unit invisible layer...