Generative Adversarial Networks (GANs) are a class of artificial intelligence algorithms used in machine learning. They are a system of two neural networks, a generative model and a discriminative model. The generative model is used to generate data, while the discriminative model is used to determine whether the generated data is real or fake. GANs use a game dynamic, wherein the generative model attempts to create data that appears real, while the discriminative model attempts to recognize the data as fake.
GANs are used in areas that require artificial synthesis of data, including speech synthesis, image generation, and music generation. The generative model is trained to produce data which can be indistinguishable from real data. The discriminative model is trained to be able to distinguish between real and artificially generated data. The models work together in a loop, wherein each model gains knowledge from the other, which allows it to improve its own performance. GANs can generate data with much higher resolution than traditional methods.
GANs are relatively new concepts in the field of computer science, having been first proposed in 2014. They have since seen much research, and are being used in a variety of areas, including computer vision, natural language processing, and generative design. GANs have shown great potential in a variety of areas, and are likely to continue to be one of the more popular algorithms in the coming years.