What should GAN in AI stand for?

GANs were introduced as AI framework where two learning models (a generator and a discriminator) compete in a two-player game. Goodfellow et al. (2014) were the first to explicitly define this two-network, mini-max optimization game and coined it as “Generative Adversarial Network (GAN)” focusing on adversarial applications. Since then, some GANs research focused on adversarial applications such as deepfakes for images, audio, and videos, while a diverse range of non-adversarial GAN applications also emerged. It is to be noted that the GAN formulation is inherently neutral which uses the well-established concepts of mini-max, bi-level optimization and game theory.  

Accordingly, we argue that an appropriate interpretation of the term GAN should be “Generative Associated Networks” since it tries to associate a generated instance to the real one by minimizing their divergence. Such an interpretation reflects its broader role in generating and associating meaningful representations beyond adversarial learning.  There are different types of GAN (based on their defining characteristics) which are used for generating synthetic data, hyper-realistic images, videos, simulation and other transformative digital content. Their influence extends across domains such as healthcare, art, and data augmentation, making these techniques useful in many AI-based applications. 

I will discuss the evolution of GANs—from Basic-GAN to EVO-GAN and DRD-GAN and illustrate with different applications to justify why GAN should stand for “Generative Associated Networks”. 

References:

• What should GAN in AI stand for? D Dasgupta, A Roy - Authorea Preprints, 2025

• DRD-GAN: A Novel Distributed Conditional Wasserstein Deep Convolutional Relativistic Discriminator GAN with Improved Convergence. Arunava Roy, Dipankar Dasgupta. ACM Transactions on Probabilistic Machine Learning, Volume 1, Issue 1, Pages 1 – 34, 09 December 2024 . https://doi.org/10.1145/3655030.

• A Critical Analysis of Distributed-GANs: Approaches, Challenges and Future Directions. A Roy, D Dasgupta - Authorea Preprints, 2024 - TechRxiv; DOI: 10.36227/techrxiv.172289155.54627068/v1.

• Generative Adversarial Networks. Goodfellow et al. Proceedings of the 31st International Conference on Neural Information Processing Systems (NIPS), 2014.