Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Functional Space Analysis of Local GAN Convergence
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Tune the loss and gradient-penalty parameters using the dataset's estimated Poincaré constant | LSGAN-like GAN on two-Gaussian mixtures · Gaussian mixtures with D=3 and D=4 | Earth Mover's Distance | Helped |
| Use sub-optimal gradient penalty strength (ablation) | LSGAN-like GAN on two-Gaussian mixtures · Gaussian mixtures with D=3 and D=4 | Earth Mover's Distance | Hurt |
| Add zoom-in augmentation at a suitable intermediate strength | SNDCGAN · CIFAR-10 | FID | Helped |
| Add cutout augmentation at a suitable intermediate strength | SNDCGAN · CIFAR-10 | FID | Helped |
| Increase translation augmentation strength (ablation) | SNDCGAN · CIFAR-10 | FID | Hurt |
| Add color-based augmentation | SNDCGAN · CIFAR-10 | FID | Hurt |
| Increase the instance-selection truncation quantile (ablation) | GAN data on CIFAR-10 · CIFAR-10 | Δξ_min: 0 -> 0.95 -> -0.04 -> 1.24 -> \mathbf{2.90} -> 2.74 | Mixed |
| Use stronger instance selection by removing low-likelihood samples | GAN data on CIFAR-10 · CIFAR-10 | ξ_min | Helped |
| Increase zoom-out augmentation strength (ablation) | SNDCGAN · CIFAR-10 | FID | No clear effect |
Closest papers by meaning
- Training Generative Adversarial Networks by Solving Ordinary Differential Equations (NeurIPS 2020 · similarity 0.91)
- On the Limitations of First-Order Approximation in GAN Dynamics (ICML 2018 · similarity 0.90)
- A Solvable High-Dimensional Model of GAN (NeurIPS 2019 · similarity 0.90)
- Smoothness and Stability in GANs (ICLR 2020 · similarity 0.90)
- The GAN is dead; long live the GAN! A Modern GAN Baseline (NeurIPS 2024 · similarity 0.87)
- Exploring the Precise Dynamics of Single-Layer GAN Models: Leveraging Multi-Feature Discriminators for High-Dimensional Subspace Learning (NeurIPS 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.