notes about gan-based researches
Generative Adversarial Networks (GAN) have two models, generator and discriminator. The generator fools the discriminator by making fake data, and the discriminator determines which data is fake or real. The continuous generator makes an image somewhere between two data for example between a bird and a cat. However, in real data distribution, there are no such data since it consists of a disconnected manifold. Thus, to cover this problem, the following papers provide solutions.