generative adversarial networks

A machine learning framework in which two neural net­works contest with each other in the form of a zero-sum game, where one agent’s gain is another agent’s loss. A GAN learns to generate new data with the same statistics as the training set.