Backprogragation isn't implemented yet

A naive implementation of a Convolutional Network in Python 3 that learn how to classify the CIFAR-10 dataset.

While librairies like Theano, TensorFlow and [insert the others 1000 librairies here] are awesome and allow for an amazing abstraction of the actual Neural Network construction, training and testing, they often hide the complexity of those model and prevent me from understanding the inner working of these algorithms. To circumvent that I decided to make a naive implementation!