mlsanityData to 7 Oct 2026

A Closer Look at the Optimization Landscapes of Generative Adversarial Networks

arXiv 1906.04848 ↗First posted 11 Jun 2019cs.LG stat.MLcited by 25

Not on the map. Why?

Abstract (arXiv)

Generative adversarial networks have been very successful in generative modeling, however they remain relatively challenging to train compared to standard deep neural networks. In this paper, we propose new visualization techniques for the optimization landscapes of GANs that enable us to study the game vector field resulting from the concatenation of the gradient of both players. Using these visualization techniques we try to bridge the gap between theory and practice by showing empirically that the training of GANs exhibits significant rotations around Local Stable Stationary Points (LSSP), similar to the one predicted by theory on toy examples. Moreover, we provide empirical evidence that GAN training converge to a stable stationary point which is a saddle point for the generator loss, not a minimum, while still achieving excellent performance.

Key facts. Cited by 25 papers in the map (data through 7 Oct 2026). First posted on arXiv 11 Jun 2019.

Builds on

No reference list yet.

Built on by

Cited by year

Citing papers per year: 2021: 14; 2022: 4; 2023: 3; 2024: 3; 2025: 1.

25 papers.

Sort:

By area:

  1. 2023CoLa-Diff: Conditional Latent Diffusion Model for Multi-Modal MRI Synthesiscited by 64
  2. 2021Fair Normalizing Flowscited by 38
  3. 2024ExtremeCast: Boosting Extreme Value Prediction for Global Weather Forecastcited by 19
  4. 2021On the Convergence of Stochastic Extragradient for Bilinear Games using Restarted Iteration Averagingcited by 18
  5. 2021Understanding Overparameterization in Generative Adversarial Networkscited by 14
  6. 2023SAN: Inducing Metrizability of GAN with Discriminative Normalized Linear Layercited by 14
  7. 2021Complex Momentum for Optimization in Gamescited by 11
  8. 2021Last-Iterate Convergence of Saddle-Point Optimizers via High-Resolution Differential Equationscited by 8
  9. 2024Phy-Diff: Physics-guided Hourglass Diffusion Model for Diffusion MRI Synthesiscited by 8
  10. 2021Fair Representations by Compressioncited by 7
  11. 2022Domain Adversarial Training: A Game Perspectivecited by 7
  12. 2022Minimax Optimality (Probably) Doesn't Imply Distribution Learning for GANscited by 6
13 more
  1. 2021Multi-Resolution Continuous Normalizing Flowscited by 4
  2. 2023Fixed-kinetic Neural Hamiltonian Flows for enhanced interpretability and reduced complexitycited by 4
  3. 2021A Differential Geometry Perspective on Orthogonal Recurrent Modelscited by 3
  4. 2021Generative Minimization Networks: Training GANs Without Competitioncited by 3
  5. 2021On Characterizing GAN Convergence Through Proximal Duality Gapcited by 3
  6. 2021Relaxed Conditional Image Transfer for Semi-supervised Domain Adaptationcited by 2
  7. 2021Generative Adversarial Network: Some Analytical Perspectivescited by 2
  8. 2021Convergence of GANs Training: A Game and Stochastic Control Methodologycited by 2
  9. 2021Generative Adversarial Networks via a Composite Annealing of Noise and Diffusioncited by 1
  10. 2022When is Momentum Extragradient Optimal? A Polynomial-Based Analysiscited by 1
  11. 2024Local convergence of simultaneous min-max algorithms to differential equilibrium on Riemannian manifoldcited by 1
  12. 2025Towards Scalable Topological Regularizerscited by 1
  13. 2022On the Nash equilibrium of moment-matching GANs for stationary Gaussian processescited by 0