Research map / Probabilistic and scientific modeling / Variational and generative sampling
Learning Probabilistic Models from Generator Latent Spaces with Hat EBM
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 |
|---|---|---|---|
| Refine SN-GAN samples with Hat EBM | SN-GAN · CIFAR-10 | FID: 18.58 -> 14.04 | Helped |
| Refine SN-GAN samples with Hat EBM | SN-GAN · CelebA 64 | FID: 6.13 -> 5.98 | Helped |
| Hat EBM instead of DDLS for refinement | SN-GAN refinement · CIFAR-10 | FID: 14.59 -> 14.04 | Helped |
| Hat EBM instead of DDLS for refinement | SN-GAN refinement · CelebA 64 | FID: 6.06 -> 5.98 | Helped |
| Remove residual image Y from Hat EBM (ablation) | Hat EBM · CIFAR-10 refinement | training stability | Hurt |
| Scale Hat EBM by doubling channels (ablation) | Hat EBM · ImageNet 128 | FID: 40.24 -> 29.37 | Helped |
| Anneal generator and hat learning rate by 10x after 250K updates (ablation) | Hat EBM · ImageNet 128 | FID | Helped |
| Hat EBM versus other EBMs on CIFAR-10 | Improved CD EBM · CIFAR-10 | FID: 25.1 -> 19.30 | Helped |
| Hat EBM versus SN-GAN on ImageNet | SN-GAN · ImageNet 128 | FID: 65.7 -> 40.24 | Helped |
| Hat EBM versus SN-GAN on CelebA | SN-GAN · CelebA 64 | FID: 6.1 -> 11.57 | Hurt |
| Hat EBM versus DDPM on CIFAR-10 | DDPM · CIFAR-10 | FID: 3.2 -> 19.30 | Hurt |
| Scaled Hat EBM versus scaled SS-GAN | SS-GAN scaled · ImageNet 128 | FID: 23.4 -> 29.37 | Hurt |
Closest papers by meaning
- Learning Latent Space Energy-Based Prior Model (NeurIPS 2020 · similarity 0.90)
- Learning Latent Space Hierarchical EBM Diffusion Models (ICML 2024 · similarity 0.87)
- Improving Adversarial Energy-Based Model via Diffusion Process (ICML 2024 · similarity 0.86)
- Energy Matching: Unifying Flow Matching and Energy-Based Models for Generative Modeling (NeurIPS 2025 · similarity 0.86)
- Learning Energy-Based Models by Cooperative Diffusion Recovery Likelihood (ICLR 2024 · similarity 0.86)
- Joint Learning of Energy-based Models and their Partition Function (ICML 2025 · 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.