Deep Convolutional GANs

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.

Generative models are not limited though to GANs. Variational Autoencoders (VAEs) (the secret sauce of OpenAI’s DALL-E), which can learn the underlying distribution of data and can generate new samples by sampling from that distribution, and auto-regressive models (the secret sauce of LLMs), which generate data one element at a time, conditioned on the previous elements, are also among a long list of well-known generative models. However, GANs leverage their ability to generate highly realistic and diverse samples that resemble...