This is a demo implementation of a multi layer neural network. Automatic differentiation is used to compute the gradients during learning.
The goal is to have a very simple and straight forward implementation for demoing and learning the inner workings of DNNs, i.e. learning by doing for myself.