Research map / Probabilistic and scientific modeling / Dynamical systems
ModLaNets: Learning Generalisable Dynamics via Modularity and Physical Inductive Bias
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 |
|---|---|---|---|
| Use modular element-energy networks with Lagrangian dynamics rather than a whole-system baseline | ModLaNet · Double-pendulum systems | Average prediction MSE_q: 3.42×10^{0} ± 1.92×10^{0} -> 1.28×10^{0} ± 1.25×10^{0} | Helped |
| Use modular element-energy networks with Lagrangian dynamics rather than a whole-system baseline | ModLaNet · Three-body systems | Average prediction MSE_E: 2.08×10^{-2} ± 2.25×10^{-2} -> 1.10×10^{-3} ± 1.61×10^{-3} | Helped |
| Use ModLaNet instead of HNN | Double-pendulum systems · Double-pendulum systems | Average prediction MSE_E: 1.24×10^{+1} ± 2.37×10^{+1} -> 4.60×10^{-2} ± 5.02×10^{-2} | Helped |
| Use ModLaNet instead of HNN | Three-body systems · Three-body systems | Average prediction MSE_q: 9.55×10^{-2} ± 5.53×10^{-2} -> 3.96×10^{-2} ± 2.94×10^{-2} | Helped |
| Fine-tune on a small dataset of the more complex system | ModLaNet extensions to multi-pendulum systems · Multi-pendulum systems | MSE_q: 16%-26% | Helped |
| Fine-tune on a small dataset of the more complex system | ModLaNet extensions to multi-pendulum systems · Multi-pendulum systems | MSE_E: 96% | Helped |
| Fine-tune on a small dataset of the more complex system | ModLaNet extensions to multi-body systems · Multi-body systems | MSE_q: 28%-52% | Helped |
| Fine-tune on a small dataset of the more complex system | ModLaNet extensions to multi-body systems · Multi-body systems | MSE_E | No clear effect |
| Increase ModLaNet double-pendulum training data from 10k to 20k (ablation) | ModLaNet · Double-pendulum systems | Test loss: 3.47×10^{-2} -> 3.02×10^{-2} | Helped |
| Increase ModLaNet double-pendulum training data from 10k to 40k (ablation) | ModLaNet · Double-pendulum systems | Test loss: 3.47×10^{-2} -> 7.82×10^{-2} | Hurt |
| Increase ModLaNet double-pendulum training data to 100k (ablation) | ModLaNet · Double-pendulum systems | Test loss: fail | Hurt |
| Increase ModLaNet double-pendulum training data to 1m (ablation) | ModLaNet · Double-pendulum systems | Test loss: fail | Hurt |
Closest papers by meaning
- Learning Articulated Rigid Body Dynamics with Lagrangian Graph Neural Network (NeurIPS 2022 · similarity 0.89)
- Learning Modular Simulations for Homogeneous Systems (NeurIPS 2022 · similarity 0.88)
- A Riemannian Framework for Learning Reduced-order Lagrangian Dynamics (ICLR 2025 · similarity 0.88)
- Poisson-Dirac Neural Networks for Modeling Coupled Dynamical Systems across Domains (ICLR 2025 · similarity 0.87)
- Towards Cross Domain Generalization of Hamiltonian Representation via Meta Learning (ICLR 2024 · similarity 0.86)
- SEGNO: Generalizing Equivariant Graph Neural Networks with Physical Inductive Biases (ICLR 2024 · 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.