Research map / Probabilistic and scientific modeling / Variational and generative sampling
GILBO: One Metric to Measure Them All
Alexander A Alemi, Ian Fischer
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 |
|---|---|---|---|
| Measure model complexity with GILBO alongside FID | GANs on MNIST · MNIST | FID | Mixed |
| Compare GAN and VAE complexity at the same FID | GANs versus VAEs · MNIST | GILBO | Hurt |
| Use GILBO to distinguish models with equal FID | CelebA GANs · CelebA | GILBO: ∼41 nats; ∼70 nats; ∼104 nats | Mixed |
| Replace the amortized encoder with separately optimized per-image encoders | CelebA GANs · CelebA | GILBO: 41 -> 87; 70 -> 111; 104 -> 155 | Improved |
| Use a variational encoder approximation to estimate generative mutual information | GANs · CelebA | GILBO | No clear effect |
| Use the same encoder architecture and training procedure across compared models | GILBO comparisons · CelebA | GILBO reproducibility: below 2% of the mean GILBO | Improved |
Closest papers by meaning
- On the Statistical Properties of Generative Adversarial Models for Low Intrinsic Data Dimension (JMLR 2025 · similarity 0.86)
- Complexity Matters: Rethinking the Latent Space for Generative Modeling (NeurIPS 2023 · similarity 0.85)
- Stein Latent Optimization for Generative Adversarial Networks (ICLR 2022 · similarity 0.85)
- Closed-Form Factorization of Latent Semantics in GANs (CVPR 2021 · similarity 0.85)
- Statistically Optimal Generative Modeling with Maximum Deviation from the Empirical Distribution (ICML 2024 · similarity 0.85)
- CLR-GAN: Improving GANs Stability and Quality via Consistent Latent Representation and Reconstruction (ECCV 2024 · similarity 0.85)
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.
