Research map / Image generation / GANs
DeGAN: Data-Enriching GAN for Retrieving Representative Samples from a Trained Classifier
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 |
|---|---|---|---|
| Add the frozen classifier's class guidance and balance constraints to proxy-data GAN training (ablation) | DeGAN for CIFAR-10 knowledge distillation using CIFAR-100 proxy data · CIFAR-10, 90 non-overlapping proxy classes | accuracy: 66.24 -> 80.55 | Helped |
| Add the frozen classifier's class guidance and balance constraints to proxy-data GAN training (ablation) | DeGAN for Fashion-MNIST knowledge distillation · Fashion-MNIST, CIFAR-10 proxy | accuracy: 79.67 -> 83.79 | Helped |
| Add the frozen classifier's class guidance and balance constraints to proxy-data GAN training (ablation) | DeGAN for CIFAR-100 knowledge distillation · CIFAR-100, CIFAR-10 proxy | accuracy: 39.77 -> 65.25 | Helped |
| Enrich proxy samples instead of using them directly | DeGAN knowledge distillation · CIFAR-10, 90 non-overlapping CIFAR-100 proxy classes | accuracy: 74.58 -> 80.55 | Helped |
| Enrich proxy samples instead of using them directly | DeGAN knowledge distillation · CIFAR-10, SVHN proxy | accuracy: 45.18 -> 55.05 | Helped |
| Use DeGAN-generated old-class examples rather than new-class proxy data directly for distillation | Single-step class-incremental learning · CIFAR-100 | accuracy: 65.03 -> 68.65 | Helped |
| Use random noise as proxy data and enrich it with DeGAN (ablation) | DeGAN knowledge distillation · CIFAR-10 | accuracy: 11.63 -> 23.26 | Helped |
| Reduce proxy coverage from 90 non-overlapping classes to 6 background classes (ablation) | DeGAN knowledge distillation · CIFAR-10 | accuracy: 80.55 -> 59.53 | Hurt |
Closest papers by meaning
- MineGAN: Effective Knowledge Transfer From GANs to Target Domains With Few Images (CVPR 2020 · similarity 0.87)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.87)
- DigGAN: Discriminator gradIent Gap Regularization for GAN Training with Limited Data (NeurIPS 2022 · similarity 0.87)
- DeLiGAN : Generative Adversarial Networks for Diverse and Limited Data (CVPR 2017 · similarity 0.87)
- RG-GAN: Dynamic Regenerative Pruning for Data-Efficient Generative Adversarial Networks (AAAI 2024 · similarity 0.86)
- Taming the Tail in Class-Conditional GANs: Knowledge Sharing via Unconditional Training at Lower Resolutions (CVPR 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.