mlsanityData to 7 Oct 2026

Nested Annealed Training Scheme for Generative Adversarial Networks

arXiv 2501.11318 ↗First posted 20 Jan 2025cs.CV cs.LGcited by 0

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

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)[1]. Specifically, we reveal the theoretical connection between the CFG model and score-based models. We find that the training objective of the CFG discriminator is equivalent to finding an optimal D(x). The optimal gradient of D(x) 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(x) that minimizes this difference. In this paper, we aim to derive an annealed weight preceding the weight of the CFG discriminator. This new explicit theoretical explanation model is called the annealed CFG method. To overcome the limitation of the annealed CFG method, as the method is not readily applicable to the 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 quality and diversity of the synthesized samples. This improvement is clear when comparing the CFG method and the SOTA GAN models.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 48 of its 68 references that are in the map. First posted on arXiv 20 Jan 2025.

Builds on

48 of 68 references.

Closest ancestors · 6

Same direction, newest first

  1. 2022Analytic-DPM: an Analytic Estimate of the Optimal Reverse Variance in Diffusion Probabilistic Models17 Jan
  2. 2021High-Resolution Image Synthesis with Latent Diffusion Models20 Dec
  3. 2021Tackling the Generative Learning Trilemma with Denoising Diffusion GANs15 Dec
  4. 2020Score-Based Generative Modeling through Stochastic Differential Equations26 Nov
  5. 2020Denoising Diffusion Probabilistic Models19 Jun
  6. 2020Improved Techniques for Training Score-Based Generative Models16 Jun

Other references · 42

  1. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
  2. 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,539
  3. 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,322
  4. 2021Zero-Shot Text-to-Image Generationcited by 4,276
  5. 2019Generative Modeling by Estimating Gradients of the Data Distributioncited by 4,067
  6. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  7. 2016Improved Techniques for Training GANscited by 2,938
  8. 2018Large Scale GAN Training for High Fidelity Natural Image Synthesiscited by 2,609
  9. 2017Improved Training of Wasserstein GANscited by 2,223
  10. 2022Imagen Video: High Definition Video Generation with Diffusion Modelscited by 1,765
  11. 2015LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loopcited by 1,350
  12. 2016Least Squares Generative Adversarial Networkscited by 1,000
30 more
  1. 2018Which Training Methods for GANs do actually Converge?cited by 672
  2. 2016f-GAN: Training Generative Neural Samplers using Variational Divergence Minimizationcited by 504
  3. 2022MM-Diffusion: Learning Multi-Modal Diffusion Models for Joint Audio and Video Generationcited by 225
  4. 2023Drag Your GAN: Interactive Point-based Manipulation on the Generative Image Manifoldcited by 218
  5. 2023Locally Attentional SDF Diffusion for Controllable 3D Shape Generationcited by 129
  6. 2016Mode Regularized Generative Adversarial Networkscited by 123
  7. 2017BEGAN: Boundary Equilibrium Generative Adversarial Networkscited by 121
  8. 2017Stabilizing Training of Generative Adversarial Networks through Regularizationcited by 118
  9. 2020FedGAN: Federated Generative Adversarial Networks for Distributed Datacited by 89
  10. 2022Variational inference via Wasserstein gradient flowscited by 84
  11. 2018How Well Generative Adversarial Networks Learn Distributionscited by 78
  12. 2017Gradient descent GAN optimization is locally stablecited by 65
  13. 2017Triple Generative Adversarial Netscited by 59
  14. 2017Dual Discriminator Generative Adversarial Netscited by 55
  15. 2021Variational Wasserstein gradient flowcited by 53
  16. 2023LDM3D: Latent Diffusion Model for 3Dcited by 53
  17. 2018Finding Mixed Nash Equilibria of Generative Adversarial Networkscited by 48
  18. 2017Many Paths to Equilibrium: GANs Do Not Need to Decrease a Divergence At Every Stepcited by 45
  19. 2021An error analysis of generative adversarial networks for learning distributionscited by 35
  20. 2019Deep Generative Learning via Variational Gradient Flowcited by 32
  21. 2020Smoothness and Stability in GANscited by 27
  22. 2017An Online Learning Approach to Generative Adversarial Networkscited by 13
  23. 2018Fictitious GAN: Training GANs with Historical Modelscited by 7
  24. 2018Stackelberg GAN: Towards Provable Minimax Equilibrium via Multi-Generator Architecturescited by 7
  25. 2018Composite Functional Gradient Learning of Generative Adversarial Modelscited by 6
  26. 2021Training Generative Adversarial Networks in One Stagecited by 6
  27. 2021Game of GANs: Game-Theoretical Models for Generative Adversarial Networkscited by 6
  28. 2018Gradient Layer: Enhancing the Convergence of Adversarial Training for Generative Modelscited by 5
  29. 2020Training Generative Adversarial Networks via stochastic Nash gamescited by 3
  30. 2023Error analysis of generative adversarial networkcited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2023Insights into Closed-form IPM-GAN Discriminator Guidance for Diffusion Modeling
  2. 2023DiffFlow: A Unified SDE Framework for Score-Based Diffusion Models and Generative Adversarial Networks
  3. 2025Generative Adversarial Networks Bridging Art and Machine Intelligence
  4. 2024Pruning then Reweighting: Towards Data-Efficient Training of Diffusion Models
  5. 2022Refining Generative Process with Discriminator Guidance in Score-based Diffusion Models
  6. 2024You Only Sample Once: Taming One-Step Text-to-Image Synthesis by Self-Cooperative Diffusion GANs
  7. 2023Complexity Matters: Rethinking the Latent Space for Generative Modeling
  8. 2022SPI-GAN: Denoising Diffusion GANs with Straight-Path Interpolations