Research map / Label-efficient and robust learning / Recognition and long-tail
A Classification-Based Study of Covariate Shift in GAN Distributions
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 |
|---|---|---|---|
| Train the classifier on GAN-generated rather than true data | The studied GANs · CelebA, Smiling (Y/N) | test accuracy (%): 85.6 (true data) vs 55.8–65.3 (GAN data) | Hurt |
| Train the classifier on GAN-generated rather than true data | The studied GANs · LSUN, Bedroom/Kitchen | test accuracy (%): 64.1 (true data) vs 51.2–56.2 (GAN data) | Hurt |
| Oversample GAN data 10-fold | GAN-generated training datasets · CelebA and LSUN | test accuracy | No clear effect |
| Increase annotator granularity by considering more classes | The mode-collapse evaluation · CelebA and LSUN | mode balance | Hurt |
| Use ALI rather than the other studied GAN setups | GAN mode distribution · CelebA and LSUN | mode collapse | Helped |
| Use WGAN or ALI rather than the other studied GANs | GAN sample diversity · CelebA and LSUN | classification-based diversity | Helped |
| Evaluate diversity using the classifier-based tests in addition to perceptual quality | GAN evaluation · CelebA and LSUN | diversity assessment | Helped |
Closest papers by meaning
- Quantitatively Evaluating GANs With Divergences Proposed for Training (ICLR 2018 · similarity 0.91)
- Are GANs Created Equal? A Large-Scale Study (NeurIPS 2018 · similarity 0.91)
- A Large-Scale Study on Regularization and Normalization in GANs (ICML 2019 · similarity 0.90)
- A Neural Tangent Kernel Perspective of GANs (ICML 2022 · similarity 0.90)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · similarity 0.89)
- The GAN is dead; long live the GAN! A Modern GAN Baseline (NeurIPS 2024 · similarity 0.88)
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.