How did Generative Adversarial Networks get their name?

Question

How did Generative Adversarial Networks get their name? I understand why this model is a generative model, but I do not understand what adversarial means in this context.

Answer 1

In GANs, there are two networks. The first network generates fake data. The second network is shown examples of both real data and fake data generated by the first network. Its goal is to determine whether its input is real or fake. The second network is trained to better distinguish real from fake data, and the first network is trained to produce fake data that better fools the second network. The overall training procedure amounts to a competition between the two networks, which is why the model is called 'adversarial'. An equilibrium point of this competition occurs if the first network learns to perfectly model the 'true' distribution, at which point the second network can do no better than chance.

Answer 2

From the paper that introduced GANs {1}:

In the proposed adversarial nets framework, the generative model is pitted against an adversary: a discriminative model that learns to determine whether a sample is from the model distribution or the data distribution. The generative model can be thought of as analogous to a team of counterfeiters, trying to produce fake currency and use it without detection, while the discriminative model is analogous to the police, trying to detect the counterfeit currency. Competition in this game drives both teams to improve their methods until the counterfeits are indistiguishable from the genuine articles.

Two side notes:

• the term "network" is misleading as neither of the generative model nor the discriminative model have to be a neural network. (same issue with the term "memory networks": Where is the network in memory networks?)
• Jurgen Schmidhuber claims to have performed similar work earlier in that direction. He called it predictability minimization. (Were generative adversarial networks introduced by Jürgen Schmidhuber?)

---

References:

• {1} Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, Yoshua Bengio. Generative Adversarial Networks. arXiv:1406.2661 https://arxiv.org/abs/1406.2661