Research map / Probabilistic and scientific modeling / Variational and generative sampling
E(n) Equivariant Normalizing Flows
Victor Garcia Satorras, Emiel Hoogeboom, Fabian Bernd Fuchs, Ingmar Posner, Max Welling
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 |
|---|---|---|---|
| E(n) equivariant dynamics instead of non-equivariant GNF-att-aug (ablation) | E-NF · DW4, 10^2 samples | NLL: 8.81 -> 8.31 | Improved |
| E(n) equivariant dynamics instead of GNF-att-aug (ablation) | E-NF · DW4, 10^5 samples | NLL: 7.61 -> 7.48 | Improved |
| E-NF instead of GNF-att-aug (ablation) | E-NF · LJ13, 10 samples | NLL: 41.09 -> 33.12 | Improved |
| E-NF instead of GNF-att-aug (ablation) | E-NF · LJ13, 10^4 samples | NLL: 30.93 -> 30.41 | Improved |
| data augmentation with rotations for non-equivariant GNF-att (ablation) | GNF-att · DW4, 10^2 samples | NLL: 11.65 -> 8.81 | Improved |
| data augmentation with rotations for GNF-att (ablation) | GNF-att · LJ13, 10^2 samples | NLL: 36.22 -> 31.50 | Improved |
| attention (edge inference) added to GNF (ablation) | GNF · DW4, 10^4 samples | NLL: 10.38 -> 9.34 | Improved |
| E-NF instead of Kernel dynamics | E-NF · DW4, 10^2 samples | NLL: 8.74 -> 8.31 | Improved |
| E-NF instead of Simple dynamics | E-NF · LJ13, 10 samples | NLL: 33.67 -> 33.12 | Improved |
| E-NF instead of GNF-att-aug (ablation) | E-NF · QM9 | NLL: -29.3 -> -59.7 | Improved |
| E-NF instead of GNF-att-aug (ablation) | E-NF · QM9 | atom stability: 75% -> 85% | Improved |
| E-NF instead of GNF-att-aug (ablation) | E-NF · QM9 | molecule stability: 0.5% -> 4.9% | Improved |
Closest papers by meaning
- Flow Equivariant Recurrent Neural Networks (NeurIPS 2025 · similarity 0.88)
- On the Generalization of Equivariant Graph Neural Networks (ICML 2024 · similarity 0.87)
- Equivariant Graph Neural Operator for Modeling 3D Dynamics (ICML 2024 · similarity 0.86)
- Scalable Non-Equivariant 3D Molecule Generation via Rotational Alignment (ICML 2025 · similarity 0.86)
- ET-Flow: Equivariant Flow-Matching for Molecular Conformer Generation (NeurIPS 2024 · similarity 0.85)
- SEGNO: Generalizing Equivariant Graph Neural Networks with Physical Inductive Biases (ICLR 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.
