mlsanityData to 14 Oct 2026

Combating Mode Collapse in GAN training: An Empirical Analysis using Hessian Eigenvalues

arXiv 2012.09673 ↗First posted 17 Dec 2020cs.LG cs.CVcited by 27

Not on the map. Why?

Abstract (arXiv)

Generative adversarial networks (GANs) provide state-of-the-art results in image generation. However, despite being so powerful, they still remain very challenging to train. This is in particular caused by their highly non-convex optimization space leading to a number of instabilities. Among them, mode collapse stands out as one of the most daunting ones. This undesirable event occurs when the model can only fit a few modes of the data distribution, while ignoring the majority of them. In this work, we combat mode collapse using second-order gradient information. To do so, we analyse the loss surface through its Hessian eigenvalues, and show that mode collapse is related to the convergence towards sharp minima. In particular, we observe how the eigenvalues of the G are directly correlated with the occurrence of mode collapse. Finally, motivated by these findings, we design a new optimization algorithm called nudged-Adam (NuGAN) that uses spectral information to overcome mode collapse, leading to empirically more stable convergence properties.

Key facts. Cited by 27 papers in the map (data through 14 Oct 2026). First posted on arXiv 17 Dec 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 2; 2022: 3; 2023: 5; 2024: 5; 2025: 10; 2026: 2.

27 papers.

Sort:

By area:

  1. 2024DiffuseLoco: Real-Time Legged Locomotion Control with Diffusion from Offline Datasetscited by 49
  2. 2024HSIGene: A Foundation Model For Hyperspectral Image Generationcited by 22
  3. 2023Idempotent Generative Networkcited by 18
  4. 2022A survey on GANs for computer vision: Recent research, analysis and taxonomycited by 15
  5. 2025Metrics that matter: Evaluating image quality metrics for medical image generationcited by 12
  6. 2024StainDiffuser: MultiTask Dual Diffusion Model for Virtual Stainingcited by 11
  7. 2022Score-Based Generative Models for Molecule Generationcited by 10
  8. 2024xTED: Cross-Domain Adaptation via Diffusion-Based Trajectory Editingcited by 9
  9. 2023Fine-tuning Language Models with Generative Adversarial Reward Modellingcited by 8
  10. 2025AutoMedPrompt: A New Framework for Optimizing LLM Medical Prompts Using Textual Gradientscited by 7
  11. 2023Collaborative Synthesis of Patient Records through Multi-Visit Health State Inferencecited by 6
  12. 2022Generative Adversarial Networks for Image Augmentation in Agriculture: A Systematic Reviewcited by 3
15 more
  1. 2021Joint Semi-supervised 3D Super-Resolution and Segmentation with Mixed Adversarial Gaussian Domain Adaptationcited by 2
  2. 2025Score-based Idempotent Distillation of Diffusion Modelscited by 2
  3. 2025SynQuE: Estimating Synthetic Dataset Quality Without Annotationscited by 2
  4. 2023Generative Adversarial Networks for Malware Detection: a Surveycited by 1
  5. 2021Regularized Generative Adversarial Networkcited by 0
  6. 2023Learning Representative Trajectories of Dynamical Systems via Domain-Adaptive Imitationcited by 0
  7. 2024RDM: Recurrent Diffusion Model for Human Motion Generationcited by 0
  8. 2025LapLoss: Laplacian Pyramid-based Multiscale loss for Image Translationcited by 0
  9. 2025Beyond Anonymization: Object Scrubbing for Privacy-Preserving 2D and 3D Vision Taskscited by 0
  10. 2025Generative diffusion model surrogates for mechanistic agent-based biological modelscited by 0
  11. 2025IMPLICITSTAINER: Resolution Agnostic Data-Efficient Virtual Staining Using Neural Implicit Functionscited by 0
  12. 2025StolenLoRA: Exploring LoRA Extraction Attacks via Synthetic Datacited by 0
  13. 2025Diffusion-Based Synthetic Brightfield Microscopy Images for Enhanced Single Cell Detectioncited by 0
  14. 2026A Wasserstein GAN-based climate scenario generator for risk management and insurance: the case of soil subsidencecited by 0
  15. 2026ROMS-IMLE: A Minimalist Approach to Competitive Single-Step Generative Modellingcited by 0