Research map / Probabilistic and scientific modeling / Dynamical systems
LEADS: Learning Dynamical Systems that Generalize Across Environments
Yuan Yin, Ibrahim Ayed, Emmanuel de Bezenac, Nicolas Baskiotis, patrick gallinari
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 |
|---|---|---|---|
| use LEADS instead of FT-NODE | dynamical-system modeling · Lotka-Volterra (LV), m=10, n=1·K | MSE test: 3.40$\pm$2.64 e-3 -> 1.16$\pm$0.99 e-3 | Improved |
| use LEADS instead of FT-NODE | dynamical-system modeling · Gray-Scott (GS), m=3, n=1·K | MSE test: 3.86$\pm$3.36 e-3 -> 2.08$\pm$2.88 e-3 | Improved |
| use LEADS instead of GBML-like | dynamical-system modeling · Navier-Stokes (NS), m=4, n=8·K | MSE test: 7.37$\pm$4.80 e-3 -> 5.95$\pm$3.65 e-3 | Improved |
| add the Ω(g_e) penalty (ablation) | LEADS · Lotka-Volterra (LV), m=10, n=1·K | MSE test: 3.07$\pm$2.58 e-3 -> 1.16$\pm$0.99 e-3 | Improved |
| add the Ω(g_e) penalty (ablation) | LEADS · Gray-Scott (GS), m=3, n=1·K | MSE test: 5.53$\pm$3.43 e-3 -> 2.08$\pm$2.88 e-3 | Improved |
| add the Ω(g_e) penalty (ablation) | LEADS · Navier-Stokes (NS), m=4, n=8·K | MSE test: 7.10$\pm$4.24 e-3 -> 5.95$\pm$3.65 e-3 | Improved |
| increase the number of environments | LEADS · Lotka-Volterra (LV), 1 to 8 environments | test error | Improved |
| increase the number of environments | One-For-All · Lotka-Volterra (LV), 1 to 8 environments | test error | Hurt |
| train the novel-environment residual using a pretrained shared function | LEADS on novel Lotka-Volterra environments · novel LV environments | test error | Improved |
| use only the pretrained shared function without adaptation | Pre-trained-f-Only · novel LV environments | test error | Hurt |
Closest papers by meaning
- Learning Interpretable Hierarchical Dynamical Systems Models from Time Series Data (ICLR 2025 · similarity 0.89)
- Environment Inference for Learning Generalizable Dynamical System (NeurIPS 2025 · similarity 0.88)
- Out-of-Domain Generalization in Dynamical Systems Reconstruction (ICML 2024 · similarity 0.87)
- Learning Hybrid Dynamics Models with Simulator-Informed Latent States (AAAI 2024 · similarity 0.87)
- When are dynamical systems learned from time series data statistically accurate? (NeurIPS 2024 · similarity 0.86)
- Generalization Dynamics of Linear Diffusion Models (NeurIPS 2026 · 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.
