Generative Adversarial Nets

This paper introduces generative adversarial networks (GANs), a framework for estimating generative models via an adversarial process. Two networks are trained simultaneously: a generator that produces new data instances, and a discriminator that evaluates them. The generator learns to create realistic data that can fool the discriminator, while the discriminator learns to distinguish between real and generated data. This adversarial process drives both networks to improve until the generator's output is indistinguishable from real data.