Research map / Image generation / GANs
SGAN: An Alternative Training of 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 |
|---|---|---|---|
| Train a global pair against five independently trained local pairs | WGAN · CIFAR10 | Inception score: 1.21(.002) -> 4.44(.058) | Helped |
| Train a global pair against five independently trained local pairs | DRAGAN · CIFAR10 | Inception score: 3.40(.036) -> 4.80(.069) | Helped |
| Train a global pair against five independently trained local pairs | GAN · CIFAR10 | Inception score: 3.58(.020) -> 4.73(.051) | Helped |
| Remove messenger discriminators and train the global generator without copies of the local discriminators (ablation) | SGAN · 8-GMM | log-likelihood | Hurt |
| Use weight sharing among the networks instead of separate networks (ablation) | SGAN with DCGAN · Real-world image datasets | sample quality | Hurt |
| Use weight sharing among the networks instead of separate networks (ablation) | SGAN with some GAN variants · Real-world image datasets | sample quality | Mixed |
| Use RMSProp instead of Adam (ablation) | GANs on toy datasets · Toy datasets | performance | No clear effect |
| Update the global generator after each local-pair update rather than at the standard SGAN frequency (ablation) | SGAN · Toy datasets | performance | No clear effect |
Closest papers by meaning
- TRAINING GENERATIVE ADVERSARIAL NETWORKS VIA PRIMAL-DUAL SUBGRADIENT METHODS: A LAGRANGIAN PERSPECTIVE ON GAN (ICLR 2018 · similarity 0.91)
- MGAN: Training Generative Adversarial Nets with Multiple Generators (ICLR 2018 · similarity 0.91)
- Improved Techniques for Training GANs (NeurIPS 2016 · similarity 0.91)
- MSG-GAN: Multi-Scale Gradients for Generative Adversarial Networks (CVPR 2020 · similarity 0.91)
- Training Generative Adversarial Networks with Limited Data (NeurIPS 2020 · similarity 0.91)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.90)
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.