Research map / Image generation / GANs
Detecting Overfitting of Deep Generative Networks via Latent Recovery
Check 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 10 random restarts for latent recovery instead of a single initialization (ablation) | MESCH · real images (train) | recovery success rate: 68% to 98% | Helped |
| Use 10 random restarts for latent recovery instead of a single initialization (ablation) | MESCH · real images (test) | recovery success rate: 67% to 99% | Helped |
| Use 10 random restarts for latent recovery instead of a single initialization (ablation) | MESCH · generated images | recovery success rate: 67% to 96% | Helped |
| Use LBFGS rather than SGD for latent recovery (ablation) | latent recovery · generated images | recovery error | Helped |
| Use LBFGS rather than Adam for latent recovery (ablation) | latent recovery · generated images | convergence speed | Helped |
| Use VGG perceptual loss instead of Euclidean loss (ablation) | latent recovery · CelebA-HQ | recovery optimization | Hurt |
| Apply pooling by a factor of 32 during latent recovery (ablation) | latent recovery · image recovery | convergence | No clear effect |
| Increase GLO training data size from 1024 to 16384 | GLO · MNIST | MRE-gap (train vs val): 6.78e-01 -> 6.45e-03 | Helped |
| Increase GLO training data size from 1024 to 16384 | GLO · CIFAR-10 | MRE-gap (train vs val): 5.84e-01 -> 6.45e-03 | Helped |
Closest papers by meaning
- Memorization Precedes Generation: Learning Unsupervised GANs with Memory Networks (ICLR 2018 · similarity 0.90)
- Optimizing the Latent Space of Generative Networks (ICML 2018 · similarity 0.89)
- Complexity Matters: Rethinking the Latent Space for Generative Modeling (NeurIPS 2023 · similarity 0.89)
- Generalized Adversarially Learned Inference (AAAI 2021 · similarity 0.88)
- A Large-Scale Study on Regularization and Normalization in GANs (ICML 2019 · similarity 0.88)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · 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.