Research map / Image generation / GANs
Small-GAN: Speeding up GAN Training using Core-Sets
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 core-set sampling to prior and target batches | SN-GAN · CIFAR-10 | FID: 18.75 -> 16.73 | Helped |
| Add core-set sampling to MEG | MEG · MNIST held-out digits 1, 4, 5, 7, and 9 | area under precision-recall curve: 0.281 -> 0.351; 0.401 -> 0.501; 0.402 -> 0.518; 0.29 -> 0.387; 0.342 -> 0.39 | Helped |
| Select target examples using image-space distances (ablation) | Small-GAN · CIFAR-10 | FID: 16.73 -> 18.09 | Hurt |
| Use Inception embeddings without random projection (ablation) | Small-GAN · CIFAR-10 | FID: 16.73 -> 17.03 | Hurt |
| Apply core-set sampling only to the prior (ablation) | Small-GAN · CIFAR-10 | FID: 16.73 -> 17.88 | Hurt |
| Apply core-set sampling only to the target distribution (ablation) | Small-GAN · CIFAR-10 | FID: 16.73 -> 17.45 | Hurt |
| Use prior and target sampling factors 2×2 instead of 4×8 (ablation) | Small-GAN · CIFAR-10 | FID: 16.73 -> 18.01 | Hurt |
| Use prior and target sampling factors 2×4 instead of 4×8 (ablation) | Small-GAN · CIFAR-10 | FID: 16.73 -> 17.8 | Hurt |
| Use prior and target sampling factors 4×2 instead of 4×8 (ablation) | Small-GAN · CIFAR-10 | FID: 16.73 -> 17.59 | Hurt |
| Use prior and target sampling factors 4×4 instead of 4×8 (ablation) | Small-GAN · CIFAR-10 | FID: 16.73 -> 17.12 | Hurt |
| Use prior and target sampling factors 8×4 instead of 4×8 (ablation) | Small-GAN · CIFAR-10 | FID: 16.73 -> 16.83 | Hurt |
| Use prior and target sampling factors 8×8 instead of 4×8 (ablation) | Small-GAN · CIFAR-10 | FID: 16.73 -> 16.9 | Hurt |
Closest papers by meaning
- Projected GANs Converge Faster (NeurIPS 2021 · similarity 0.90)
- Re-GAN: Data-Efficient GANs Training via Architectural Reconfiguration (CVPR 2023 · similarity 0.90)
- Improved Techniques for Training GANs (NeurIPS 2016 · similarity 0.89)
- Ensembling Off-the-Shelf Models for GAN Training (CVPR 2022 · similarity 0.89)
- Training Generative Adversarial Networks with Limited Data (NeurIPS 2020 · similarity 0.89)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.89)
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.