Variations of GANs
In this section, we will explore the main variation of GANs. For an interesting practical application of GANs, please refer to Chapter 12 and Case Study 3 – Predictive Analytics to see how Amazon utilized GANs for fraud transaction prediction. For more applications, please refer to Generative Adversarial Networks in the built environment: A comprehensive review of the application of GANs across data types and scales (https://www.sciencedirect.com/science/article/abs/pii/S0360132322007089).
Conditional GAN (cGAN)
A typical GAN generates images given a random noise vector. However, in many scenarios, we really want to control the attributes and properties of the generated synthetic samples. For example, suppose you are deploying a GAN to generate human faces. The standard GAN architecture has no way to let you specify some attributes of the generated faces such as gender, age, eye color, and hair length. Using cGAN, we can condition the GAN on these attributes...
