Research map / Probabilistic and scientific modeling / Variational and generative sampling
Learning Latent Space Energy-Based Prior Model
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 |
|---|---|---|---|
| replace Gaussian prior with latent EBM prior (vs VAE) | generator model · SVHN | FID: 46.78 -> 29.44 | Helped |
| replace Gaussian prior with latent EBM prior (vs VAE) | generator model · CIFAR-10 | FID: 106.37 -> 70.15 | Helped |
| replace Gaussian prior with latent EBM prior (vs VAE) | generator model · CelebA | FID: 65.75 -> 37.87 | Helped |
| learn EBM prior in single latent layer vs 5-layer latent prior SRI (L=5) | generator model · SVHN | FID: 35.23 -> 29.44 | Helped |
| EBM prior vs two-stage learned prior RAE | generator model · SVHN | MSE: 0.014 -> 0.008 | Helped |
| EBM prior vs two-stage learned prior 2sVAE | generator model · CIFAR-10 | FID: 72.90 -> 70.15 | Helped |
| EBM prior vs FB-VAE | text generator · SNLI | forward perplexity: 39.19 -> 27.81 | Helped |
| EBM prior vs FB-VAE | text generator · PTB | reverse perplexity: 204.11 -> 181.54 | Helped |
| EBM prior vs ARAE/FB-VAE on NLL | text generator · SNLI | NLL: 28.90 vs 28.82 | No clear effect |
| EBM prior vs FB-VAE | text generator · Yahoo | NLL: 319.96 -> 321.18 | Hurt |
| EBM prior used for anomaly detection vs BiGAN-sigma | posterior log-density · MNIST held-out digit 4 | AUPRC: 0.443 -> 0.630 | Helped |
| compare to fixed isotropic Gaussian prior with same generator (ablation) | EBM prior · SVHN | FID | Helped |
Closest papers by meaning
- Learning Joint Latent Space EBM Prior Model for Multi-Layer Generator (CVPR 2023 · similarity 0.93)
- Learning Latent Space Hierarchical EBM Diffusion Models (ICML 2024 · similarity 0.90)
- Learning Energy-Based Prior Model with Diffusion-Amortized MCMC (NeurIPS 2023 · similarity 0.90)
- Joint Learning of Energy-based Models and their Partition Function (ICML 2025 · similarity 0.86)
- Learning Energy-Based Models by Cooperative Diffusion Recovery Likelihood (ICLR 2024 · similarity 0.85)
- Energy-Based Modelling for Discrete and Mixed Data via Heat Equations on Structured Spaces (NeurIPS 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.