Research map / Image generation / GANs
Generative Adversarial Ranking Nets
Yinghua Yao, Yuangang Pan, Jing Li, Ivor W. Tsang, Xin Yao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Use GARNet ranker-based adversarial training instead of GAN-2 trained on the top subset | GAN on preference data · MNIST (preferred digits) | mean digit (MD): 0.15 -> 0.00 | Improved |
| Use GARNet instead of vanilla GAN-0 trained on all data | GAN · MNIST (preferred digits) | mean digit (MD): 4.49 -> 0.00 | Improved |
| Add a ranker to the GAN (GAN-RK) instead of GARNet's ranker-as-discriminator | GAN · MNIST (preferred digits) | mean digit (MD): 4.63 vs 0.00 | Hurt |
| Use GARNet instead of FBGAN | GAN · MNIST (preferred digits) | diversity / mode collapse | Improved |
| Use GARNet instead of GAN-2 | GAN · LFW (smile) | mean score (MS): 1.14 -> 2.29 | Improved |
| Use GARNet instead of GAN-2 | GAN · LFW (smile) | FID: 37.40 -> 22.22 | Improved |
| Use GARNet instead of GAN-2 | GAN · UT-Zap50K (comfort) | FID: 77.99 -> 39.16 | Improved |
| Use GARNet instead of GAN-2 | GAN · UT-Zap50K (comfort) | mean score (MS): -0.91 -> -0.90 | No clear effect |
| Use GARNet instead of GAN-2 | GAN · UT-Zap50K (open) | mean score (MS): 0.99 -> 3.75 | Improved |
| Use GARNet instead of GAN-2 | GAN · UT-Zap50K (sporty) | mean score (MS): 4.48 -> 7.07 | Improved |
| Use GARNet instead of FBGAN | GAN · UT-Zap50K (sporty) | FID: 431.44 -> 32.87 | Improved |
Closest papers by meaning
- Adversarial Ranking for Language Generation (NeurIPS 2017 · similarity 0.87)
- Strategic Candidacy in Generative AI Arenas (ICML 2026 · similarity 0.85)
- Perturbation-Invariant Adversarial Training for Neural Ranking Models: Improving the Effectiveness-Robustness Trade-Off (AAAI 2024 · similarity 0.85)
- Sampling-Decomposable Generative Adversarial Recommender (NeurIPS 2020 · similarity 0.85)
- Preference learning made easy: Everything should be understood through win rate (ICML 2025 · similarity 0.85)
- Activation Maximization Generative Adversarial Nets (ICLR 2018 · similarity 0.85)
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.
