We just accomplished an important part of our learning journey regarding deep learning architectures—GANs! Throughout this chapter, we got more familiar with GANs and their variants. We started with what GANs are; the evolution paths of GANs; and how they became so popular in data synthesis, such as image generation, audio, and video generation. We also explored four GAN architectures, that is, vanilla GANs, deep convolutional GANs, conditional GANs, and the information-maximizing GANs. We implemented each individual GAN model from scratch and used them to generate digital images that appear to be real.
GANs are a great invention of deep learning that's been made in recent years. In the next chapter, we will talk about other recent advancements in deep learning, including Bayesian neural networks, capsule networks, and meta-learning.
...
