EUVIMEDEUVIMED entdecken
Crossref

Nested Annealed Training Scheme for Generative Adversarial Networks

Chang Wan, Ming-Hsuan Yang, Minglu Li, Yunliang Jiang, Zhonglong Zheng

IEEE Transactions on Circuits and Systems for Video Technology · 2025-01 · 35 (1) · 670-683

DOI: 10.1109/tcsvt.2024.3456476 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Circuits and Systems for Video Technology, 2025-01. Verfasst von Chang Wan und 4 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 66 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Natural Science Foundation of Zhejiang Province.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Recently, researchers have proposed many deep generative models, including generative adversarial networks (GANs) and denoising diffusion models. Although significant breakthroughs have been made and empirical success has been achieved with the GAN, its mathematical underpinnings remain relatively unknown. This paper focuses on a rigorous mathematical theoretical framework: the composite-functional-gradient GAN (CFG). Specifically, we reveal the theoretical connection between the CFG model and score-based models. We find that the CFG discriminator’s training objective is equivalent to finding an optimal $D(\mathrm {x})$ . The optimal $D(\mathrm {x})$ ’s gradient differentiates the integral of the differences between the score functions of real and synthesized samples. Conversely, training the CFG generator involves finding an optimal $G(\mathrm {x})$ that minimizes this difference. In this paper, we aim to derive an annealed weight preceding the CFG discriminator’s weight. This new explicit theoretical explanation model is called the annealed CFG method. To overcome the annealed CFG method’s limitation, as the method is not readily applicable to the state-of-the-art (SOTA) GAN model, we propose a nested annealed training scheme (NATS). This scheme keeps the annealed weight from the CFG method and can be seamlessly adapted to various GAN models, no matter their structural, loss, or regularization differences. We conduct thorough experimental evaluations on various benchmark datasets for image generation. The results show that our annealed CFG and NATS methods significantly improve the synthesized samples’ quality and diversity. This improvement is clear when comparing the CFG method and the SOTA GAN models.

Förderung

  • National Natural Science Foundation of China
  • Natural Science Foundation of Zhejiang Province

Zitieren

APA

Wan, C., Yang, M., Li, M., Jiang, Y., & Zheng, Z. (2025). Nested Annealed Training Scheme for Generative Adversarial Networks. IEEE Transactions on Circuits and Systems for Video Technology, 35(1), 670-683. https://doi.org/10.1109/tcsvt.2024.3456476

Vancouver

Wan C, Yang M, Li M, Jiang Y, Zheng Z. Nested Annealed Training Scheme for Generative Adversarial Networks. IEEE Transactions on Circuits and Systems for Video Technology. 2025;35(1):670-683. doi:10.1109/tcsvt.2024.3456476

BibTeX
@article{Wan2025,
  author = {Wan, Chang and Yang, Ming-Hsuan and Li, Minglu and Jiang, Yunliang and Zheng, Zhonglong},
  title = {Nested Annealed Training Scheme for Generative Adversarial Networks},
  journal = {IEEE Transactions on Circuits and Systems for Video Technology},
  year = {2025},
  volume = {35},
  number = {1},
  pages = {670--683},
  doi = {10.1109/tcsvt.2024.3456476}
}
Kostenlos starten →