Research map / Image generation / GANs
IB-GAN: Disentangled Representation Learning with Information Bottleneck Generative Adversarial Networks
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 an intermediate stochastic representation and an information-bottleneck constraint to the InfoGAN objective | InfoGAN · dSprites | disentanglement score: 0.61 ± 0.03 -> 0.80 ± 0.07 | Helped |
| Add an intermediate stochastic representation and an information-bottleneck constraint to the InfoGAN objective | InfoGAN · Color-dSprites | disentanglement score: 0.55 ± 0.08 -> 0.79 ± 0.05 | Helped |
| Add an intermediate stochastic representation and an information-bottleneck constraint to the InfoGAN objective | InfoGAN · CelebA | FID: 9.3 -> 7.4 | Helped |
| Add an intermediate stochastic representation and an information-bottleneck constraint to the InfoGAN objective | InfoGAN · 3D Chairs | FID: 25.6 -> 25.5 | No clear effect |
| Set β to 0, removing the upper-bound constraint on generative mutual information (ablation) | IB-GAN · dSprites | disentanglement score | Hurt |
| Set β to 1, strongly constraining the generative mutual information (ablation) | IB-GAN · dSprites | disentanglement | Hurt |
| Tune β within the reported effective range [0.071,~0.212] (ablation) | IB-GAN · dSprites | disentanglement score | Helped |
| Tune β to 0.141 (ablation) | IB-GAN · dSprites | disentanglement score: 0.80 | Helped |
| Tune β to 0.071 (ablation) | IB-GAN · Color-dSprites | disentanglement score: 0.79 | Helped |
Closest papers by meaning
- InfoGAN-CR and ModelCentrality: Self-supervised Model Training and Selection for Disentangling GANs (ICML 2020 · similarity 0.90)
- OOGAN: Disentangling GAN with One-Hot Sampling and Orthogonal Regularization (AAAI 2020 · similarity 0.89)
- InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets (NeurIPS 2016 · similarity 0.88)
- Distilling Representations from GAN Generator via Squeeze and Span (NeurIPS 2022 · similarity 0.88)
- Disentangled Image Generation Through Structured Noise Injection (CVPR 2020 · similarity 0.87)
- MCGAN: Enhancing GAN Training with Regression-Based Generator Loss (AAAI 2025 · 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.