Research map / Probabilistic and scientific modeling / Dynamical systems
Spatial Attention Kinetic Networks with E(n)-Equivariance
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 |
|---|---|---|---|
| add spatial attention (neural linear combinations of edge vectors) (ablation) | SAKE · MD17 Aspirin | Energy RMSE (meV): 8.08 -> 5.91 | Helped |
| add spatial attention (neural linear combinations of edge vectors) (ablation) | SAKE · MD17 Aspirin | Force RMSE (meV/Å): 17.48 -> 8.09 | Helped |
| add velocity and position update (ablation) | SAKE · MD17 Aspirin | Force RMSE (meV/Å): 10.31 -> 8.09 | Helped |
| add semantic attention (ablation) | SAKE · MD17 Aspirin | Energy RMSE (meV): 7.88 -> 5.91 | Helped |
| remove spatial attention from EGNN-style backbone with auxiliary modules | EGNN · MD17 Aspirin | Energy RMSE (meV): 16.02 -> 589.34 when distance smearing removed; spatial attention is not in EGNN | Hurt |
| SAKE vs EGNN on N-body charged particles | SAKE · N-body charged particles | MSE: 0.0071 -> 0.0049 | Helped |
| SAKE vs GMN on walking motion capture | SAKE · Walking motion capture | MAE: 43.9 -> 14.59 | Helped |
| SAKE vs NequIP on MD17 energy/force | SAKE · MD17 (8 molecules) | Energy and force MAE | Mixed |
| SAKE on QM9 extensive vs intensive properties | SAKE · QM9 | MAE | Mixed |
Closest papers by meaning
- Equivariant Spatio-Temporal Attentive Graph Networks to Simulate Physical Dynamics (NeurIPS 2023 · similarity 0.89)
- Equivariant Graph Neural Operator for Modeling 3D Dynamics (ICML 2024 · similarity 0.88)
- Equivariant Graph Hierarchy-Based Neural Networks (NeurIPS 2022 · similarity 0.88)
- SEGNO: Generalizing Equivariant Graph Neural Networks with Physical Inductive Biases (ICLR 2024 · similarity 0.88)
- Generalizing Convolutional Neural Networks for Equivariance to Lie Groups on Arbitrary Continuous Data (ICML 2020 · similarity 0.87)
- Equivariant Neural Tangent Kernels (ICML 2025 · 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.