Research map / Image generation / GANs
Spectral Normalization for Generative Adversarial Networks
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 |
|---|---|---|---|
| Apply spectral normalization to discriminator weights | Standard CNN GAN · CIFAR-10 | Inception score: 6.68±.06 -> 7.42±.08 | Helped |
| Apply spectral normalization to discriminator weights | Standard CNN GAN · STL-10 | FID: 55.1 -> 53.1 | Helped |
| Apply spectral normalization instead of orthonormal regularization | ResNet GAN · CIFAR-10 | Inception score: 7.92±.04 -> 8.22±.05 | Helped |
| Apply spectral normalization instead of orthonormal regularization | ResNet GAN · STL-10 | FID: 42.4±.99 -> 40.1±.50 | Helped |
| Use hinge loss (ablation) | SN-GANs · CIFAR-10 | Inception score: 7.42±.08 -> 7.58±.12 | Helped |
| Use hinge loss (ablation) | SN-GANs · CIFAR-10 | FID: 29.3 -> 25.5 | Helped |
| Train for 2x updates (ablation) | SN-GANs · STL-10 | Inception score: 8.28±.09 -> 8.69±.09 | Helped |
| Train for 2x updates (ablation) | SN-GANs · STL-10 | FID: 53.1 -> 47.5 | Helped |
| Increase final-layer feature-map dimension (ablation) | SN-GANs and orthonormal-regularized GANs · STL-10 | Inception score | Mixed |
| Increase discriminator updates per generator update (ablation) | Weight-normalized GAN · CIFAR-10 and STL-10 | Inception score | Hurt |
| Increase discriminator updates per generator update (ablation) | SN-GANs · CIFAR-10 and STL-10 | Inception score | No clear effect |
| Use aggressive learning rates and momentum (ablation) | WGAN-GP · CIFAR-10 and STL-10 | Inception score | Hurt |
Closest papers by meaning
- Why Spectral Normalization Stabilizes GANs: Analysis and Improvements (NeurIPS 2021 · similarity 0.94)
- On Computation and Generalization of Generative Adversarial Networks under Spectrum Control (ICLR 2019 · similarity 0.93)
- Sparsity Aware Normalization for GANs (AAAI 2021 · similarity 0.92)
- Consistency Regularization for Generative Adversarial Networks (ICLR 2020 · similarity 0.90)
- A Large-Scale Study on Regularization and Normalization in GANs (ICML 2019 · similarity 0.90)
- CHAIN: Enhancing Generalization in Data-Efficient GANs via lipsCHitz continuity constrAIned Normalization (CVPR 2024 · similarity 0.88)
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.