Research map / Probabilistic and scientific modeling / Variational and generative sampling
Learning Hierarchical Features with Joint Latent Space Energy-Based Prior
Jiali Cui, Ying Nian Wu, Tian Han
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 |
|---|---|---|---|
| add joint latent EBM prior to hierarchical generator (vs Gaussian-prior VLAE) | multi-layer generator · SVHN latent classifier | top-layer accuracy: 52.14% -> 86.63% | Improved |
| increase prior Langevin steps from 15 to 60 (ablation) | joint EBM prior · CelebA-64 | FID: 56.42 -> 32.15 | Improved |
| increase Langevin steps beyond 60 (ablation) | joint EBM prior · CelebA-64 | FID: 32.15 -> 30.78 | No clear effect |
| increase hidden units of the EBM (nef) from 0 to 100 (ablation) | joint EBM prior · SVHN | FID: 43.95 -> 24.16 | Improved |
| use joint EBM prior instead of baselines for generation | generator · SVHN | FID: 26.19 -> 24.16 vs Multi-NCP | Improved |
| joint EBM prior versus Multi-NCP on SVHN reconstruction | generator · SVHN | MSE: 0.004 (Multi-NCP) vs 0.008 (ours) | Hurt |
| use joint EBM prior for anomaly detection instead of HVAE | inference model · MNIST anomaly detection | AUPRC: 0.494 -> 0.722 (digit 1) | Improved |
| use joint EBM prior instead of HVAE with specialized LLR decision | OOD detection · Fashion-MNIST vs MNIST | AUPRC: 0.984 (HVAE LLR) vs 0.893 (ours) | Hurt |
| use joint EBM prior versus pro-VLAE with progressive learning | disentanglement · 3DShapes | MIG: 0.357 -> 0.554 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- BIVA — BIVA: A Very Deep Hierarchy of Latent Variables for Generative Modeling (NeurIPS 2019) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning Joint Latent Space EBM Prior Model for Multi-Layer Generator (CVPR 2023 · similarity 0.94)
- Learning Latent Space Hierarchical EBM Diffusion Models (ICML 2024 · similarity 0.92)
- Towards Identifiability of Hierarchical Temporal Causal Representation Learning (NeurIPS 2025 · similarity 0.86)
- Improving Unsupervised Hierarchical Representation with Reinforcement Learning (CVPR 2024 · similarity 0.85)
- Joint Learning of Energy-based Models and their Partition Function (ICML 2025 · similarity 0.85)
- Learning Discrete Concepts in Latent Hierarchical Models (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.
