Research map / Image generation / GANs
Data-Efficient GAN Training Beyond (Just) Augmentations: A Lottery Ticket Perspective
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 |
|---|---|---|---|
| Use a 67.24% sparse BigGAN ticket instead of the dense model | BigGAN · Tiny-ImageNet 64×64, 10% data | FID: 11.30 FID | Helped |
| Add AdvAug to sparse tickets | BigGAN tickets · Tiny-ImageNet 64×64, 10% data | FID: 73.23 ± 0.05 -> 70.91 ± 0.05 | Helped |
| Use a 36.00% sparse ticket with AdvAug instead of dense BigGAN | BigGAN · ImageNet 128×128, 25% data | FID: 25.37 ± 0.07 -> 23.14 ± 0.07 | Helped |
| Combine DiffAug and AdvAug with sparse tickets | BigGAN tickets · CIFAR-10, 10% data | FID: 25.41 -> 14.38 | Helped |
| Combine DiffAug and AdvAug with sparse tickets | BigGAN tickets · CIFAR-100, 10% data | FID: 30.96 -> 22.37 | Helped |
| Add AdvAug to GAN ticket training | StyleGAN-V2 tickets · CIFAR-100, 10% data | FID: 12.20 ± 0.03 -> 12.11 ± 0.05 | Helped |
| Add AdvAug to dense StyleGAN-V2 with ADA and RLC | Dense StyleGAN-V2 · CIFAR-100, 10% data | FID: 12.87 ± 0.03 -> 13.01 ± 0.02 | Hurt |
| Prune to 48.80% sparsity and add DiffAug plus AdvAug | StyleGAN-V2 · Obama 100-shot | FID: 89.18 -> 52.86 | Helped |
| Replace dense model with a 67.24% sparse ticket (ablation) | SNGAN · CIFAR-10, 10% data | FID: 41.25 -> 32.47 | Helped |
| Use a lower IMP pruning ratio (ablation) | SNGAN ticket finding · CIFAR-10, 10% data | FID | Helped |
| Apply AdvAug to either the discriminator or generator (ablation) | SNGAN tickets · CIFAR-10, 10% data | FID | Helped |
| Apply AdvAug to both discriminator and generator rather than only one (ablation) | SNGAN tickets · CIFAR-10, 10% data | FID | Helped |
Closest papers by meaning
- GANs Can Play Lottery Tickets Too (ICLR 2021 · similarity 0.93)
- Winning Lottery Tickets in Deep Generative Models (AAAI 2021 · similarity 0.92)
- Re-GAN: Data-Efficient GANs Training via Architectural Reconfiguration (CVPR 2023 · similarity 0.91)
- Data-Efficient Double-Win Lottery Tickets from Robust Pre-training (ICML 2022 · similarity 0.88)
- SMaRt: Improving GANs with Score Matching Regularity (ICML 2024 · similarity 0.87)
- RG-GAN: Dynamic Regenerative Pruning for Data-Efficient 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.