Research map / Learning theory and algorithms / Sample complexity
AdaGAN: Boosting Generative Models
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 sequential discriminator-based reweighting and mixture components (T=3) | GAN · Toy Gaussian mixture, 5 modes | coverage C: 0.72 -> 0.91 | Helped |
| Add sequential discriminator-based reweighting and mixture components (T=10) | GAN · Toy Gaussian mixture, 10 modes | coverage C: 0.59 -> 1.00 | Helped |
| Use a uniform mixture of independently trained generators instead of sequential reweighting (ablation) | GAN, 5-mode toy Gaussian mixture · Toy Gaussian mixture, 5 modes | coverage C: 0.91 -> 1.00 | Hurt |
| Use a uniform mixture of independently trained generators instead of sequential reweighting (ablation) | GAN, 10-mode toy Gaussian mixture · Toy Gaussian mixture, 10 modes | coverage C: 0.89 -> 1.00 | No clear effect |
| Increase the number of sequential boosted components from T=3 to T=10 (ablation) | AdaGAN · Toy Gaussian mixture, 5 modes | coverage C: 0.91 -> 1.00 | Helped |
| Increase the number of sequential boosted components from T=3 to T=10 (ablation) | AdaGAN · Toy Gaussian mixture, 10 modes | coverage C: 0.86 -> 1.00 | Helped |
| Use the top-fraction heuristic based on the previous generator rather than the proposed reweighting (ablation) | GAN, 5-mode toy Gaussian mixture · Toy Gaussian mixture, 5 modes | coverage C: 0.99 vs 1.00 | No clear effect |
| Use the top-fraction heuristic based on the previous generator rather than the proposed reweighting (ablation) | GAN, 10-mode toy Gaussian mixture · Toy Gaussian mixture, 10 modes | coverage C: 1.00 vs 1.00 | No clear effect |
| Use the proposed reweighting and mixture (T=10) rather than a single vanilla GAN | GAN, 5-mode toy Gaussian mixture · Toy Gaussian mixture, 5 modes | log likelihood L: -23.65 -> -5.25 | Helped |
| Use the proposed reweighting and mixture (T=10) rather than a single vanilla GAN | GAN, 10-mode toy Gaussian mixture · Toy Gaussian mixture, 10 modes | log likelihood L: -55.51 -> -5.92 | Helped |
| Add unrolling steps to the underlying GAN (ablation) | AdaGAN with unrolled GAN · Toy Gaussian mixture | coverage C | No clear effect |
| Use the Jensen-Shannon objective rather than the modified GAN objective (ablation) | AdaGAN base GAN · Toy Gaussian mixture | coverage C | Helped |
Closest papers by meaning
- MGAN: Training Generative Adversarial Nets with Multiple Generators (ICLR 2018 · similarity 0.92)
- DeLiGAN : Generative Adversarial Networks for Diverse and Limited Data (CVPR 2017 · similarity 0.91)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.91)
- Training Generative Adversarial Networks with Limited Data (NeurIPS 2020 · similarity 0.91)
- A Large-Scale Study on Regularization and Normalization in GANs (ICML 2019 · similarity 0.91)
- RG-GAN: Dynamic Regenerative Pruning for Data-Efficient Generative Adversarial Networks (AAAI 2024 · similarity 0.90)
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.