This document aim to present the basics of neural networks along with small examples and code samples to illustrate them. Nothing like TensorFlow or any existing neural net frameworks will be used to grasp the concept behind each network types. The code will be written from scratch using the less external dependencies as possible.
I'll not dive into an extensive introduction of what's a neural network and how it works, there is plenty of existing resources to get your hands on that. But for the need of this presentation, let's define a basic neural network:
As shown above, this network:
- take some inputs (xn nodes)
- has a single layer of neurons (noted on above)
- produce as many outputs values as on's neurons.
Still it can be even more simplified, that's what we'll see trying to solve a first problem in Part 1 - The OR gate.