Generative Adversarial Networks

This lecture covers the theory behind Generative Adversarial Networks (GANs), focusing on the equilibrium between the generator and discriminator. It explains the discriminator's role in maximizing the loss function and the generator's role in minimizing it. The lecture delves into concepts like Kullback-Leibler (KL) and Jensen-Shannon (JS) divergences, as well as the Wasserstein distance. It also explores various GAN variants, such as Conditional GAN (CGAN) and CycleGAN, and their applications. Additionally, the lecture introduces Diffusion Models as an alternative to GANs, highlighting their differences and applications.