In this chapter, we learned about generative models and what makes them different from discriminative models. We also discussed the different kinds of autoencoders, including deep, variational, and convolutional. In addition, we learned about a new type of generative model, called a generative adversarial network (GAN). After learning about all these generative models, we saw how we could train them ourselves in TensorFlow for generating handwritten digits and saw the different quality images that they can each produce.

In Chapter 7, Transfer Learning, we will learn about transfer learning and how it can help us speed up training.