We've all heard of deep learning, but how many of us know what a Deep Belief Network is? Let's start this chapter by answering that very question. A Deep Belief Network is a very advanced form of machine learning, one whose meaning is rapidly evolving. As a machine learning developer, it's important that you have a bit of exposure to this concept so that you are familiar with it when you encounter it or it encounters you!

In machine learning, a Deep Belief Network is technically a deep neural network. We should state that the meaning of deep, when it comes to deep learning or deep belief, means that the network is composed of multiple layers (hidden units). In a Deep Belief Network, these connections span internally between each neuron within a layer, but not between different layers. A Deep Belief Network can be trained to learn unsupervised in order to probabilistically reconstruct the network's inputs. The layers then function as 'feature detectors' to recognize or classify images, letters, and so on. You can also watch a Deep Belief Network dream, which is a very interesting topic in and of itself.

In this chapter we will cover:

• Restricted Boltzmann Machines
• Creating and training a Deep Belief Network in C#