Introduction

In this chapter, you will explore generative models, which are types of unsupervised learning algorithms that generate completely new artificial data. Generative models differ from predictive models in that they aim to generate new samples from the same distribution of training data. While the purpose of these models may be very different from those covered in other chapters, you can and will use many of the concepts learned in prior chapters, including loading and preprocessing various data files, hyperparameter tuning, and building convolutional and recurrent neural networks (RNNs). In this chapter, you will learn about one way to generate new samples from a training dataset, which is to use LSTM models to complete sequences of data based on initial seed data.

Another way that you will learn about is the concept of two neural networks competing against one another in an adversarial way, that is, a generator generating samples and a discriminator trying to distinguish...