Research map / Label-efficient and robust learning / Domain adaptation and generalization
RadialGAN: Leveraging multiple datasets to improve target-specific predictive models using Generative Adversarial Networks
Jinsung Yoon, James Jordon, Mihaela Schaar
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 RadialGAN translations from other datasets to augment the target training data | Target-specific predictive models on MAGGIC · MAGGIC, M=3, 5, and 7 | AUC gain over Target-only: M=3: .0154$\pm$.0091; M=5: .0292$\pm$.0009; M=7: .0297$\pm$.0071 | Improved |
| Use RadialGAN translations from other datasets to augment the target training data | Target-specific predictive models on MAGGIC · MAGGIC, M=3, 5, and 7 | APR gain over Target-only: M=3: .0243$\pm$.0096; M=5: .0310$\pm$.0096; M=7: .0287$\pm$.0073 | Improved |
| Increase the number of datasets used from M=3 to M=7 (ablation) | RadialGAN · MAGGIC | AUC gain over Target-only: .0154$\pm$.0091 -> .0297$\pm$.0071 | Improved |
| Increase the number of datasets used from M=3 to M=7 (ablation) | RadialGAN · MAGGIC | APR gain over Target-only: .0243$\pm$.0096 -> .0287$\pm$.0073 | Mixed |
| Augment with samples generated by a GAN trained only on the target dataset | Target-only predictive model · MAGGIC studies | Prediction gain | Hurt |
| Use RadialGAN instead of target-only GAN to generate additional training samples | Target-specific predictive model · MAGGIC studies | Prediction gain | Improved |
| Use RadialGAN when feature mismatch is absent (ablation) | Target-specific predictive model · MAGGIC, Setting A | AUC gain over Target-only: .0169 | Improved |
| Use RadialGAN when distribution mismatch is absent (ablation) | Target-specific predictive model · MAGGIC, Setting B | AUC gain over Target-only: .0313 | Improved |
| Use a pairwise translation method that does not address distribution mismatch | Target-specific predictive model · MAGGIC, Setting B | AUC gain over Target-only: .0249 | Improved |
| Use StarGAN, which does not naturally address feature mismatch, in the no-distribution-mismatch setting | Target-specific predictive model · MAGGIC, Setting B | AUC gain over Target-only: .0020 | Hurt |
Closest papers by meaning
- MCL-GAN: Generative Adversarial Networks with Multiple Specialized Discriminators (NeurIPS 2022 · similarity 0.87)
- MineGAN: Effective Knowledge Transfer From GANs to Target Domains With Few Images (CVPR 2020 · similarity 0.87)
- MineGAN++: Mining Generative Models for Efficient Knowledge Transfer to Limited Data Domains (IJCV 2023 · similarity 0.87)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.87)
- Training Generative Adversarial Networks with Limited Data (NeurIPS 2020 · similarity 0.87)
- MGAN: Training Generative Adversarial Nets with Multiple Generators (ICLR 2018 · 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.
