Research map / Probabilistic and scientific modeling / Dynamical systems
Physics-Inspired All-Pair Interaction Learning for 3D Dynamics Modeling
Kai Yang, Yuqi Huang, Junheng Tao, Wanyu Wang, Qitian Wu
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 |
|---|---|---|---|
| replace the learnable pairwise mappings with fixed ones, such as an all-one matrix (adding learnable mappings) (ablation) | PAINET attention · Motion Capture Run | A-MSE | Improved |
| add the attention module (ablation) | PAINET · Motion Capture Run | A-MSE | Improved |
| use parallel equivariant decoder instead of EGNN-recurrent (ablation) | PAINET decoder · Motion Capture Run | A-MSE and time | Improved |
| use the EGNN decoder instead of MLP-add or MLP-concat (ablation) | PAINET decoder · Motion Capture Run | A-MSE | Improved |
| increase decoder layers from 1 to 10 (ablation) | PAINET decoder · Motion Capture | F-MSE | No clear effect |
| use more than one attention layer per time step (1 to 6) (ablation) | PAINET attention · Motion Capture | A-MSE | Hurt |
| use PAINET instead of GF-NODE | motion capture S2S · Motion Capture Walk (S2S) | F-MSE: 15.7 -> 8.45 | Improved |
| use PAINET instead of EGNO | motion capture S2T · Motion Capture Run (S2T) | A-MSE: 5.70 -> 3.33 | Improved |
| use PAINET instead of GF-NODE | MD17 benzene S2S · MD17 benzene (S2S) | F-MSE: 6.23 -> 4.65 | Improved |
| use PAINET instead of EGNO | Adk protein dynamics · Adk (S2T) | A-MSE: 1.805 -> 1.654 | Improved |
Closest papers by meaning
- Equivariant Graph Neural Operator for Modeling 3D Dynamics (ICML 2024 · similarity 0.87)
- Integrating GNN and Neural ODEs for Estimating Non-Reciprocal Two-Body Interactions in Mixed-Species Collective Motion (NeurIPS 2024 · similarity 0.86)
- SEGNO: Generalizing Equivariant Graph Neural Networks with Physical Inductive Biases (ICLR 2024 · similarity 0.86)
- Long-Short-Range Message-Passing: A Physics-Informed Framework to Capture Non-Local Interaction for Scalable Molecular Dynamics Simulation (ICLR 2024 · similarity 0.86)
- Generalist Equivariant Transformer Towards 3D Molecular Interaction Learning (ICML 2024 · similarity 0.85)
- Large Language-Geometry Model: When LLM meets Equivariance (ICML 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.
