Join our book community on Discord

https://packt.link/EarlyAccessCommunity

Generative neural networks have become a popular and active area of research and development. A huge amount of credit for this trend goes to a class of models that we are going to discuss in this chapter. These models are called generative adversarial networks (GANs) and were introduced in 2014. Ever since the introduction of the basic GAN model, various types of GANs have been, and are being, invented for different applications.

Essentially, a GAN is composed of two neural networks – a generator and a discriminator. Let's look at an example of the GAN that is used to generate images. For such a GAN, the task of the generator would be to generate realistic-looking fake images, and the task of the discriminator would be to tell the real images apart from the fake images.

In a joint optimization procedure, the generator would ultimately learn to generate such good fake images that the discriminator...