Research map / Image generation / GANs
Discriminator Rejection Sampling
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 DRS to GAN samples | GAN · 25 Gaussians | high-quality samples (%): $70 \pm 9$ -> $90 \pm 2$ | Helped |
| Apply DRS to GAN samples | GAN · 25 Gaussians | recovered modes: $24.8 \pm 0.4$ -> $24.8 \pm 0.4$ | No clear effect |
| Apply DRS | SAGAN · ImageNet | Inception Score: $52.34 \pm 0.45$ -> $61.44\pm 0.09$ | Helped |
| Apply DRS | SAGAN · ImageNet | FID: $18.21 \pm 0.14$ -> $17.14\pm0.09$ | Helped |
| Apply DRS | Improved-SAGAN · ImageNet | Inception Score: $62.36\pm 0.35$ -> $\mathbf{76.08\pm 0.30}$ | Helped |
| Apply DRS | Improved-SAGAN · ImageNet | FID: $14.79 \pm 0.06$ -> $\mathbf{13.57 \pm 0.13}$ | Helped |
| Add discriminator fine-tuning to convergence (ablation) | Probabilistic DRS · 25 Gaussians | within 4 std (%): 81.2 -> 89.8 | Helped |
| Replace probabilistic rejection with hard-threshold rejection (ablation) | DRS with discriminator fine-tuning · 25 Gaussians | recovered modes: 25 -> 20 | Hurt |
| Replace probabilistic rejection with hard-threshold rejection (ablation) | DRS without discriminator fine-tuning · 25 Gaussians | recovered modes: 25 -> 4 | Hurt |
| Apply hard-threshold rejection without discriminator fine-tuning (ablation) | GAN · 25 Gaussians | within 4 std (%): 75.6 -> 99.8 | Helped |
| Apply hard-threshold rejection without discriminator fine-tuning (ablation) | GAN · 25 Gaussians | recovered modes: 25 -> 4 | Hurt |
| Decrease the ImageNet DRS acceptance rate (ablation) | DRS · ImageNet | Inception Score | Mixed |
Closest papers by meaning
- Improving the Training of Data-Efficient GANs via Quality Aware Dynamic Discriminator Rejection Sampling (CVPR 2025 · similarity 0.92)
- Mixed batches and symmetric discriminators for GAN training (ICML 2018 · similarity 0.89)
- Your GAN is Secretly an Energy-based Model and You Should Use Discriminator Driven Latent Sampling (NeurIPS 2020 · similarity 0.88)
- Collaborative Sampling in Generative Adversarial Networks (AAAI 2020 · similarity 0.88)
- Direct Discriminative Optimization: Your Likelihood-Based Visual Generative Model is Secretly a GAN Discriminator (ICML 2025 · similarity 0.87)
- Discriminative Forests Improve Generative Diversity for Generative Adversarial Networks (AAAI 2024 · similarity 0.86)
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.