Research map / Graph learning / Graph algorithms and clustering
Scalable Graph Networks for Particle Simulations
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 |
|---|---|---|---|
| Replace the fully connected interaction graph with a hierarchical graph | DeltaGN · Coulomb, 100 particles | 20-step rollout RMSE: 0.0045 ± 2.7 · 10^{-4} -> 0.0052 ± 1.61 · 10^{-4} | Hurt |
| Replace the fully connected interaction graph with a hierarchical graph | DeltaGN · Coulomb, 1000 particles | 20-step rollout RMSE: 0.1744 ± 0.0234 -> 0.0058 ± 1.45 · 10^{-4} | Helped |
| Replace the fully connected interaction graph with a hierarchical graph | DeltaGN · Coulomb, 1000 particles | 200-step rollout RMSE: NA -> 0.1252 ± 0.0020 | Helped |
| Use a hierarchical graph with HOGN rather than a fully connected graph | HOGN · Gravitational, 100 particles | 20-step rollout RMSE: (2.09 ± 0.26) · 10^{-4} -> 0.0046 ± 0.0042 | Hurt |
| Use 2 hierarchy levels instead of 3 (ablation) | Hierarchical DeltaGN · Gravitational, 100 particles | 20-step rollout RMSE [10^{-3}]: 1.571 ± 0.017 -> 1.706 ± 0.184 | Hurt |
| Use 4 hierarchy levels instead of 3 (ablation) | Hierarchical DeltaGN · Gravitational, 100 particles | 20-step rollout RMSE [10^{-3}]: 1.571 ± 0.017 -> 1.545 ± 0.023 | Helped |
| Use 5 hierarchy levels instead of 3 (ablation) | Hierarchical DeltaGN · Gravitational, 100 particles | 20-step rollout RMSE [10^{-3}]: 1.571 ± 0.017 -> 1.592 ± 0.018 | Hurt |
| Use the hierarchical graph instead of 15-nearest-neighbour edges | DeltaGN · Gravitational, 1000 particles, 24 hours | 20-step rollout RMSE: 0.0086 ± 2.15 · 10^{-5} -> 0.0014 ± 2.73 · 10^{-5} | Helped |
| Use the hierarchical graph instead of 15-nearest-neighbour edges | DeltaGN · Gravitational, 1000 particles, 24 hours | 200-step rollout RMSE: 0.1565 ± 2.35 · 10^{-4} -> 0.05119 ± 0.0011 | Helped |
| Use roughly log4(N) hierarchy levels rather than too few or too many (ablation) | Hierarchical DeltaGN · Gravitational, 100 particles | Forward pass time: 2 levels: 28 ms; 3 levels: 23 ms; 4 levels: 27 ms; 5 levels: 30 ms | Mixed |
| Train with batch size 1 rather than the standard larger batch (ablation) | DeltaGN and hierarchical models · Particle simulations | Accuracy: a ~ 5 time decrease in accuracy for the other models | Hurt |
| Use hierarchical interactions rather than only 15 nearest neighbours | DeltaGN · Unseen particle counts, more than 100 particles | 20-step rollout RMSE | Helped |
Closest papers by meaning
- Graph Mamba Operator: A Latent Simulator for Interacting Particle Systems (NeurIPS 2026 · similarity 0.88)
- EvoMesh: Adaptive Physical Simulation with Hierarchical Graph Evolutions (ICML 2025 · similarity 0.86)
- How Particle System Theory Enhances Hypergraph Message Passing (NeurIPS 2025 · similarity 0.85)
- Long-Short-Range Message-Passing: A Physics-Informed Framework to Capture Non-Local Interaction for Scalable Molecular Dynamics Simulation (ICLR 2024 · similarity 0.85)
- MaNGO — Adaptable Graph Network Simulators via Meta-Learning (NeurIPS 2025 · similarity 0.85)
- Physics-Informed DeepONets for drift-diffusion on metric graphs: simulation and parameter identification (ICML 2025 · similarity 0.84)
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.