An Overview of Generative Adversarial Networks
Yiqi Zeng, Dongchi Xue
Abstract
Artificial intelligence is still the pervasive topic in the field of science and technology. Scientists have been thinking about how to rely on artificial intelligence to transform traditional industries into auto-industries which could significantly improve efficiency and reduce costs. Some researchers used to think that only humans could create images, text and voice and so on, through the GAN[l] model for sample study, however, AI could replace human beings to do all the art creation. Generating adversarial network (GAN) is a kind of neural network belonging to the category of unsupervised learning, which is suitable to solve the tasks of generating images from text, improving image resolution, drug matching and so on. This article would give an introduction of GAN, its structure, applications and some current drawbacks and issues.