Programming Machine Learning

What You Just Learned… and the Road Ahead

Let’s consider what we’ve achieved in this second part of the book. We built a neural network from scratch, we understood how and why it works, and we even worked through advanced details such as mini-batch gradient descent and testing. In this last chapter, we pushed the network as far as we managed, by standardizing its input data and tuning its hyperparameters.

Real-world ML development is more complicated than we described in these few pages, just like real-world coding is more complicated than the toy problems in programming books. However, now you have an idea of how to reach the goal that we set for ourselves at the beginning of Part II: get accurate classifications from a neural network. Remember ...