Imagine, you were a postal worker. Your would be job to deliver letters. Most of the time, the recipient's name and address would be printed and quite legible, and your job becomes quite easy. But come Thanksgiving and Christmas, the number of envelopes with handwritten addresses increases as people give their personal touches and flourishes. And, to be frank, some people (me included) just have terrible handwriting.

Blame it on schools for no longer emphasizing cursive handwriting if you must, but the problem remains: handwriting is hard to read and interpret. God forbid you have to deliver a letter penned by a doctor (good luck doing that!).

Imagine, instead, if you had built a machine learning system that allows you to read handwriting. That's what we will be doing this chapter and the next; we will be building a type of machine-learning algorithm known as an artificial neural network, and in the next chapter, we will be expanding on the concept with deep learning.

In this chapter, we will learn the basics of neural networks, see how it's inspired by biological neurons, find a better way of representing them, and finally apply neural networks on handwriting to recognize digits.