Generative Adversarial Networks

Reading about making sushi is easy; actually cooking a new kind of sushi is harder than we might think. In deep learning, the creative process is harder, but not impossible. We have seen how to build models that can classify numbers, using dense, convolutional, or recurrent networks, and today we will see how to build a model that can create numbers. This chapter introduces a learning approach known as generative adversarial networks, which belong to the family of adversarial learning and generative models. The chapter explains the concepts of generators and discriminators and why having good approximations of the distribution of the training data can lead to the success of the model in other areas such as data augmentation. By the end of the chapter, you will know why adversarial training is important; you will be able to code the necessary mechanisms...