How do GANs work and what makes them so useful in the context of machine learning? Are there any other ways for machines to learn?
In: Technology
A generative-adversarial network is, conceptually, two different AIs tied together.
The first system takes in examples of patterns, and learns to produce similar kinds of patterns as output.
The second system is shown both the real examples and the first system’s attempts, and learns to distinguish genuine from fake.
The data from the second system is *fed back in* to the first one, so it can refine its efforts to produce better fakes… and *its* data is fed into the second system, so it can get better at detecting fakes.
It’s like a forger and a forensics expert training each other to get progressively better at their jobs, until eventually you have a grandmaster forger that even a grandmaster forensic expert can’t catch out.
Latest Answers