GANs

GANs are one of the major inventions of the last decade in the opinion of lots of deep learning researchers. They are inherently different from other generative networks, especially with the way that they are trained. The first paper about adversarial networks for generating data appeared in 2014, authored by Ian Goodfellow. GANs are considered as an unsupervised learning algorithm where a supervised learning algorithm learns to reason the function y'=f(x) with labeled data, y.

This type of supervised learning algorithm is inherently discriminatory, which means that it learns to model the conditional probability distribution function where it says what the probability of something is given the state of something else. For instance, what's the probability of the location of a house if the price for buying that house is $100,000? GANs generate output from random distribution and hence the change in the random input lets the output be different.

GANs get the sample...