Generative Adversarial Networks, Their Various Types, A Comparative Analysis, and Applications in Different Areas

Abstract

GANs(Generative Adversarial Networks) are two neural networks that compete with each other to develop new data which hasn't been there originally. If we train the GAN on images, the resultant will be imaged with realistic characteristics. A GAN consists of two sub-models in which is called a generator model and the other model called the discriminator model. The Generator generates the data from the input data. The Discriminator checks or distinguishes whether the data is a sample from the real world or a generated one. This model was given for the unsupervised learning method, but it can also be applied to semi-supervised models. In this paper, we present a survey on GANs. First, we explain what simply generative adversarial networks(GANs) are, what are the different types of GANs, then what are the applications of generative adversarial networks(GANs) and then we explain the different areas in which GANs are being used and lastly what future GANs hold in different areas or fields or how can we use GANs in technology and research as a future evolving aspect.