While we've now discussed Go and the libraries available for it, we haven't yet discussed what constitutes a neural network. Toward the end of the previous chapter, we used Gorgonia to construct a graph that, when executed by an appropriate VM, performs several basic operations (specifically, addition and multiplication) on a series of matrices and vectors.

We will now talk about how to build a neural network and get it working. This will teach you about the components necessary to build the more advanced neural network architectures we will be discussing later in this book.

This chapter will cover the following topics:

• A basic neural network
• Activation functions
• Gradient descent and backpropagation
• Advanced gradient descent algorithms