An antagonistic generative network, also known as a GAN (Generative Adversarial Network), is a type of artificial neural network architecture used in the field of machine learning to generate new and original data from input data.

The structure of a GAN consists of two neural networks: a generator and a discriminator. The generator receives a set of input data (e.g. images or text) and generates new samples that resemble the original samples. The discriminator, on the other hand, evaluates the quality of the generated samples and tries to distinguish whether they are real or fake.

During training, the generator attempts to fool the discriminator by generating samples that are increasingly similar to the original samples, while the discriminator attempts to detect differences between the generated samples and the original samples. This process of competition and feedback between the generator and the discriminator continues until the generator is able to generate samples that are virtually indistinguishable from the original samples.

GANs have been used in a wide variety of applications, such as image and video generation, speech and music synthesis, natural language processing, and game modeling. In addition, GANs have proven useful for improving the quality of input data and for style transfer.