Generative Adversarial Networks are all about generating new content. GANs are capable of learning some distribution and creating a new sample from that distribution. That sample might just be a new point on a line that isn't present in our training data, but it could also be a new point in a very complex dataset. GANs have been used to generate new music, sounds, and images. According to Yann LeCun, adversarial training is the coolest thing since sliced bread (https://www.quora.com/session/Yann-LeCun/1). I'm not sure that sliced bread is especially cool, but Yann LeCun is a very cool guy so I'll take his word for it. Regardless, GANs are incredibly popular and while perhaps not as practical as some of the other topics we've covered in a business setting yet, they deserve some consideration in our survey of deep learning techniques.

In 2014, Ian Goodfellow et al. wrote a paper called Generative Adversarial Nets (https://arxiv.org/pdf/1406.2661.pdf) that proposed a framework that used the adversarial training of two deep networks, each trying to defeat the other. This framework is composed of two separate networks: a discriminator and a generator.

The discriminator is looking at real data from a training set and fake data from the generator. It's job is to classify each as incoming instance of data as either real or fake.

The generator attempts to fool the discriminator into thinking the data it is generating is real.

The generator and the discriminator are locked into a game where they each try to outsmart the other. This competition drives each network to improve until eventually the output of the generator is indistinguishable from the data in the training set, by the discriminator. When both the generator and discriminator are configured correctly they arrive at a Nash equilibrium where both are unable to find an advantage over the other.