Work

The development of generative adversarial networks in modern deep learning

Year: 2025
Type: article
Abstract: This paper offers a broad historical context for Generative Adversarial Networks (GANs), from their technological and theoretical origins in early generative model paradigms through to the sophisticated architectures that underpin contemporary research.This paper contains a wide range of areas including GANs, including core probabilistic and neural generative models that preceded adversarial training.It explores the conceptual breakthrough of GANs in 2014 and identifies key architectural advancements which have shaped their path from creation to the present, that are, Conditional GANs, DCGAN, StyleGAN, and BigGAN.It also reviews current real-world applications in various fields and identifies emerging trends like hybrid generative models and ethical issues surrounding artificial media.The discussion ends with an examination of the state of current research challenges and frontiers, such as interpretability, training stability, data efficiency, and AI responsible deployment.The study provides a historically informed yet futurist account of the evolution and potential of GANs in machine learning.
Language: English
Open Access status: green
OpenAlex ID: W7223511296
FWCI:
Cites: 18
Cited by:
Related to: 10