General Adversarial Network
Is a form of semi-supervised learning, using two depp-learning networks in competition with each other.
One network tries to create new data points that mimic the training data and the other evaluates whether they are part of the training data or fakes. These two networks operate in a positive feedback loop, ~as the discriminator gets better at separating fakes from the originals, the generator improves its ability to create convincing fakes~.