Research map / Probabilistic and scientific modeling / Dynamical systems
Learning Efficient Surrogate Dynamic Models with Graph Spline Networks
Chuanbo Hua, Federico Berto, Michael Poli, Stefano Massaroli, Jinkyoo Park
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 |
|---|---|---|---|
| add end-to-end OSC training instead of applying OSC only after training (ablation) | MGN · Heat Equation, Damped Wave, Navier Stokes, Ocean Currents, and Black Sea | MSE: 1.68±0.18 -> 1.14±0.11; 4.98±0.29 -> 4.60±0.27; 2.58±0.28 -> 2.21±0.27; 3.82±0.49 -> 3.61±0.38; 4.37±0.33 -> 4.23±0.17 | Improved |
| add adaptive collocation to end-to-end OSC (ablation) | MGN+OSC · Heat Equation, Damped Wave, Navier Stokes, Ocean Currents, and Black Sea | MSE: 1.14±0.11 -> 1.07±0.28; 4.60±0.27 -> 4.51±0.31; 2.21±0.27 -> 2.02±0.30; 3.61±0.38 -> 3.34±0.44; 4.23±0.17 -> 3.91±0.27 | Improved |
| use coarse spatial and temporal predictions with OSC interpolation | GraphSplineNets · Heat Equation | MSE (×10^-3): 2.92±0.23 -> 1.07±0.28 | Improved |
| use coarse spatial and temporal predictions with OSC interpolation | GraphSplineNets · Damped Wave | MSE (×10^-1): 6.49±0.62 -> 4.51±0.31 | Improved |
| use coarse spatial and temporal predictions with OSC interpolation | GraphSplineNets · Navier Stokes | MSE (×10^-1): 3.45±0.24 -> 2.02±0.30 | Improved |
| use OSC rather than MGN alone | MGN · 3D Navier-Stokes | MSE: (4.52 ± 0.23) × 10^{-1} -> (3.98 ± 0.27) × 10^{-1} | Improved |
| increase the number of collocation points (ablation) | MGN+OSC and MGN+OSC+Adaptive · four datasets in collocation-count sensitivity analysis | MSE | Improved |
| add adaptive collocation when using fewer collocation points (ablation) | MGN+OSC · four datasets in collocation-count sensitivity analysis | MSE | Improved |
| use fewer neural-network rollout steps and interpolate skipped times with OSC (ablation) | MGN · long-range rollouts | MSE | Improved |
| increase the number of message-passing layers (ablation) | MGN and GraphSplineNets · message-passing layer sensitivity analysis | MSE | Improved |
| replace OSC with linear interpolation (ablation) | MGN+OSC · interpolation-method sensitivity analysis | MSE | Hurt |
| use OSC rather than linear, cubic, or B-spline collocation (ablation) | MGN · interpolation-method sensitivity analysis | MSE | Improved |
Closest papers by meaning
- Physics-Informed DeepONets for drift-diffusion on metric graphs: simulation and parameter identification (ICML 2025 · similarity 0.84)
- EvoMesh: Adaptive Physical Simulation with Hierarchical Graph Evolutions (ICML 2025 · similarity 0.84)
- Learning Distributions of Complex Fluid Simulations with Diffusion Graph Networks (ICLR 2025 · similarity 0.84)
- SEGNO: Generalizing Equivariant Graph Neural Networks with Physical Inductive Biases (ICLR 2024 · similarity 0.84)
- PURE: Prompt Evolution with Graph ODE for Out-of-distribution Fluid Dynamics Modeling (NeurIPS 2024 · similarity 0.84)
- Model-Agnostic Knowledge Guided Correction for Improved Neural Surrogate Rollout (ICLR 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.
