Research map / Probabilistic and scientific modeling / Variational and generative sampling
Variational (Gradient) Estimate of the Score Function in Energy-based Latent Variable Models
Fan Bao, Kun Xu, Chongxuan Li, Lanqing Hong, Jun Zhu, Bo Zhang
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 |
|---|---|---|---|
| Use VaGES-MDSM instead of BiMDSM | Deep EBLVM · CIFAR10 | FID: 29.43 ± 2.76† -> 28.93 ± 1.91† | Improved |
| Use VaGES-MDSM instead of BiMDSM | Deep EBLVM · CelebA (64×64) | FID: 32.43‡ -> 31.53 | Improved |
| Use VaGES-DSM rather than BiDSM with 5 unrolling steps | GRBM · Frey face | time consumption: about 40% time | Improved |
| Use VaGES-DSM rather than BiDSM with 5 unrolling steps | GRBM · Frey face | memory consumption: 90% memory | Improved |
| Increase the number of convolutional layers (ablation) | VaGES-SM · Deep EBLVMs | FID | Improved |
| Vary the latent-variable dimension (ablation) | VaGES-SM · Deep EBLVMs | performance | No clear effect |
| Vary the number of samples from the variational posterior (ablation) | VaGES-SM · Deep EBLVMs | performance | No clear effect |
| Vary the Langevin dynamics noise level (ablation) | VaGES-SM · Deep EBLVMs | performance | No clear effect |
| Use KL divergence instead of Fisher divergence to train the variational posterior (ablation) | VaGES-SM · Deep EBLVMs | performance | No clear effect |
| Set variational-posterior updates and Langevin corrector steps to 5 (ablation) | VaGES-SM · Deep EBLVMs | training stability | Improved |
Closest papers by meaning
- Bi-level Score Matching for Learning Energy-based Latent Variable Models (NeurIPS 2020 · similarity 0.90)
- Bi-Level Doubly Variational Learning for Energy-Based Latent Variable Models (CVPR 2022 · similarity 0.88)
- Variational Inference with Gaussian Score Matching (NeurIPS 2023 · similarity 0.88)
- EigenVI: score-based variational inference with orthogonal function expansions (NeurIPS 2024 · similarity 0.86)
- Batch and match: black-box variational inference with a score-based divergence (ICML 2024 · similarity 0.86)
- Repulsive Latent Score Distillation for Solving Inverse Problems (ICLR 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.
