Understanding Generative Adversarial Networks (GANs)
In a Generative Adversarial Network (GAN) a generator creates fake data and a discriminator examines its inputs for authenticity.
Imagine two artists - one is a forger trying to create an exact replica of a famous painting, and the other is an art detective trying to spot the forgery. The forger starts by making a copy and gives it to the detective. The detective identifies the differences between the original and the copy. The forger then uses this information to improve their next attempt. This process continues until the forger becomes so good that the detective can no longer tell the difference between the forgery and the original.
In a Generative Adversarial Network (GAN), these two artists are represented by two parts: a generator (the forger) and a discriminator (the detective). The generator creates fake data to pass off as real. The discriminator examines its inputs for authenticity; it has to figure out whether each instance of data it sees is like the original or a forgery.
The generator and discriminator are both neural networks, a type of software inspired by the brain, and they're trained together. The generator gets better at producing fakes, and the discriminator gets better at spotting them. Over time, the generator can produce very convincing "forgeries" that look a lot like the original data.