Neural Graph Simulator for Complex Systems
Organizations: CTP and Department of Physics and Astronomy, Seoul National University, Seoul 08826, Korea · Department of Physics and Astronomy, Seoul National University, Seoul 08826, Korea · CCSS, KI for Grid Modernization, Korea Institute of Energy Technology, Naju, Jeonnam 58330, Korea · Department of Physics Education, Seoul National University, Seoul 08826, Korea ·
Abstract
Simulating complex systems governed by coupled ordinary differential equations (ODEs) typically requires explicit governing equations and incurs high computational cost from iterative numerical solvers. We propose the Neural Graph Simulator (NGS), a general-purpose data-driven framework that learns multivariate continuous-time dynamics directly from noisy and incomplete observations, without prior knowledge of the equations. Unlike neural ODE-based approaches that depend on repeated solver evaluations, NGS approximates the state transition integral using elapsed time intervals as direct inputs. This eliminates numerical integration while naturally supporting irregular sampling and heterogeneous time scales. Through its graph network architecture, NGS generalizes to unseen graph structures and larger system sizes, as demonstrated on independently sampled Erdős--Rényi graphs. We evaluate NGS on comprehensive benchmarks covering linear diffusion, nonlinear and chaotic dynamics, systems with stiffness-like behavior, and real-world spatiotemporal datasets. NGS combines competitive accuracy across benchmarks with substantial speedups over numerical solvers in computationally demanding settings. Notably, NGS exhibits strong extrapolation to unseen graph sizes and longer time horizons, and supports non-uniform time steps. These results establish NGS as an efficient and scalable alternative to solver-dependent neural simulators.
Figures & tables
| Model | NRMSE ( ) | NFE / Runtime (sec) | |||||
| ( , ) | ( , ) | ( , ) | ( , ) | ||||
| DOP853 | - | - | - | - | 1051 / 0.01 | 1201 / 3.92 | |
| 400 / 0.07 | 400 / 0.15 | ||||||
| 815 / 0.15 | 819 / 0.30 | ||||||
| 400 / 0.20 | 400 / 0.35 | ||||||
| 1265 / 0.65 | 1411 / 1.22 | ||||||
| Model | NRMSE ( ) | NFE / Runtime (sec) | |||||||
| ( , ) | ( , ) | ( , ) | ( , ) | ||||||
| DOP853 | - | - | - | - | 0.025 | 11970 / 0.14 | 13115 / 31.9 | ||
| 0.007 | 50 / 0.01 | 50 / 0.02 | |||||||
| 0.004 | 138 / 0.03 | 138 / 0.05 | |||||||
| 0.004 | 50 / 0.03 | 50 / 0.04 | |||||||
| 0.004 | 129 / 0.08 | 129 / 0.15 | |||||||
| Model | NRMSE ( ) | NFE / Runtime (sec) | |||||
| LSODA | / 6228 | / 12190 | / 13851 | ||||
| TG-ODE | 80 / 0.417 | 80 / 0.406 | 80 / 0.371 | ||||
| TDE-GNN | 80 / 0.495 | 80 / 0.430 | 80 / 0.419 | ||||
| NGS (ours) | 80 / 0.056 | 80 / 0.054 | 80 / 0.054 | ||||
| System | NRMSE | ||
| Heat ( ) | (1/40, 1/80) | ||
| (1/20, 1/40) | |||
| (1/20, 1/80) | |||
| Rössler ( ) | (1/2, 2/3) | ||
| (1.0, 2/3) | |||
| (1.0, 1/2) |
| Model | 15 minutes (3 steps) | 30 minutes (6 steps) | 60 minutes (12 steps) | ||||||||
| MAE | RMSE | MAPE | MAE | RMSE | MAPE | MAE | RMSE | MAPE | |||
| HA | 2.88 | 5.59 | 6.80 | 2.88 | 5.59 | 6.80 | 2.88 | 5.59 | 6.80 | ||
| SVR | 1.85 | 3.59 | 3.80 | 2.48 | 5.18 | 5.50 | 3.28 | 7.08 | 8.00 | ||
| ARIMA | 1.62 | 3.30 | 3.50 | 2.33 | 4.76 | 5.40 | 3.38 | 6.50 | 8.30 | ||
| FC-LSTM | 2.05 | 4.19 | 4.80 | 2.20 | 4.55 | 5.20 | 2.37 | 4.96 | 5.70 | ||
| DCRNN | 1.38 | 2.95 | 2.90 | 1.74 | 3.97 | 3.90 | 2.07 | 4.74 | 4.90 | ||
| Models | 0.5 hours | 7.5 hours | 11 hours |
| (1 Step) | (15 Steps) | (22 Steps) | |
| DCRNN | 1.59 0.17 | 4.10 0.27 | 4.93 0.60 |
| GWNet | 1.65 0.23 | 6.93 0.58 | 7.93 0.17 |
| GGN | 1.61 0.06 | 3.25 0.04 | 3.04 0.05 |
| SGP | 1.09 0.00 | 3.06 0.11 | 3.13 0.13 |
| NGS (ours) | 1.14 0.00 | 3.26 0.04 | 3.27 0.01 |
Appendix figures & tables6 assets
Supplementary material from the paper’s appendix.
Appendix
| System | State | Node coeff. | Edge coeff. | Global coeff. | |
| Mathematical systems | Heat diffusion | – | – | ||
| Coupled Rössler | – | ||||
| Kuramoto | – | ||||
| Real-world datasets | PEMS-BAY | 12-step speeds | Node ID | latest TID/DIW | |
| PV-US | 12-step generation history | Node ID | latest TID |
| System | Configuration | ( , ) | ( , ) | ( , ) | ( , ) | Rel. Runtime |
| Heat ( ) | NGS (Default) | |||||
| Depth | ||||||
| -conditioning | ||||||
| Without |
| Configuration | 15 minutes | 30 minutes | 60 minutes | Rel. Runtime | |||||||||
| MAE | RMSE | MAPE | MAE | RMSE | MAPE | MAE | RMSE | MAPE | |||||
| NGS (Default) | 1.29 | 2.79 | 2.71 | 1.60 | 3.70 | 3.61 | 1.86 | 4.35 | 4.36 | 1.00 | |||
| Depth | |||||||||||||
| 1.28 | 2.76 | 2.71 | 1.61 | 3.72 | 3.65 | 1.88 | 4.39 | 4.46 | 0.86 | ||||
| 1.29 | 2.76 | 2.71 | 1.60 | 3.67 | 3.59 | 1.86 | 4.33 | 4.37 | 1.15 | ||||
Dynamical Systems
Dynamical systems studies how states evolve over time, including stability, feedback, chaos, and transitions in physical, biological, and engineered systems.
See all 510 papers