Research map / Probabilistic and scientific modeling / Variational and generative sampling
Training Energy-Based Normalizing Flow with Score-Matching Objectives
Chen-Hao Chao, Wei-Fang Sun, Yen-Chang Hsu, Zsolt Kira, Chun-Yi Lee
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 EBFlow with SSM instead of ML training | FC-based flow · MNIST | Batch/Sec.: 8.00 -> 33.11 | Improved |
| Use EBFlow with FDSSM instead of ML training | FC-based flow · CIFAR-10 | Batch/Sec.: 5.05 -> 62.50 | Improved |
| Use EBFlow with FDSSM instead of ML training | CNN-based flow · CIFAR-10 | Batch/Sec.: 0.02 -> 39.84 | Improved |
| Use SSM training instead of ML training | FC-based flow · MNIST | NLL: 1092.4 ± 0.1 -> 1092.8 ± 0.3 | No clear effect |
| Use SSM training instead of ML training | FC-based flow · CIFAR-10 | NLL: 11912.9 ± 10.5 -> 11917.7 ± 15.5 | No clear effect |
| Use DSM training instead of ML training | FC-based flow · MNIST | NLL: 1092.4 ± 0.1 -> 1099.2 ± 0.2 | Hurt |
| Use FDSSM training instead of ML training | FC-based flow · CIFAR-10 | NLL: 11912.9 ± 10.5 -> 12347.8 ± 6.8 | Hurt |
| Use FDSSM training instead of ML training | CNN-based flow · CIFAR-10 | NLL: 11408.7 ± 26.7 -> 11766.0 ± 36.8 | Hurt |
| Add match-after-preprocessing (MaP) (ablation) | EBFlow trained with score matching · Ablation study; dataset not specified in provided text | overall performance | Improved |
| Add exponential moving average (EMA) (ablation) | EBFlow trained with score matching · Ablation study; dataset not specified in provided text | variance | Improved |
Closest papers by meaning
- Energy Matching: Unifying Flow Matching and Energy-Based Models for Generative Modeling (NeurIPS 2025 · similarity 0.88)
- An Error Analysis of Flow Matching for Deep Generative Modeling (ICML 2025 · similarity 0.87)
- Learning normalized image densities via dual score matching (NeurIPS 2025 · similarity 0.86)
- Faster Inference of Flow-Based Generative Models via Improved Data-Noise Coupling (ICLR 2025 · similarity 0.86)
- Normalizing Flows are Capable Generative Models (ICML 2025 · similarity 0.86)
- Markovian Flow Matching: Accelerating MCMC with Continuous Normalizing Flows (NeurIPS 2024 · similarity 0.86)
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.
