Pre-Publication V2. A class of machine learning architecture consisting of two neural networks, a Generator and a Discriminator, trained in opposition to each other. The Generator generates produces synthetic content (images, video, audio, or other data) with intents to produce outputs that are attempting to be statistically indistinguishable from real-world examples. The Discriminator evaluates generated outputs to classify whether outputs are, attempting to distinguish synthetic or content from genuine samples. Through iterative feedback, each network drives the other to improve until the Generator produces outputs that the Discriminator can no longer reliably identify as synthetic. GANs are foundational to many synthetic media applications including image synthesis, video generation, voice cloning, and deepfake production.
Deliberation Summary:
Notation that the definition referred to a term not otherwise defined in the set of definitions, and that any new term should be added; the section was deleted for added clarity and simplification.
Pre-Publication V1. A class of machine learning architecture consisting of two neural networks, a Generator and a Discriminator trained in opposition to each other. The Generator produces synthetic content (images, video, audio, or other data) attempting to be statistically indistinguishable from real-world examples. The Discriminator evaluates outputs, attempting to distinguish synthetic content from genuine samples. Through iterative feedback, each network drives the other to improve until the Generator produces outputs the Discriminator can no longer reliably identify as synthetic. GANs are foundational to many synthetic media applications including image synthesis, video generation, voice cloning, and deepfake production.