Research map / Probabilistic and scientific modeling / Dynamical systems
Embed and Emulate: Learning to estimate parameters of dynamical systems with uncertainty quantification
Ruoxi Jiang, Rebecca Willett
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 Embed & Emulate learned embedding and emulator to EnKI instead of Runge-Kutta solver with moment vector | EnKI · Lorenz-96 n=1000 | MAPE (%) for F: 15.48 -> 6.59 | Improved |
| replace Runge-Kutta solver by learned emulator in EnKI | EnKI · Lorenz-96 n=1000 | total time (min): 8,000 -> 95.0 | Improved |
| train Embed & Emulate with contrastive objective instead of supervised regression (n=1000) | regression head · Lorenz-96 n=1000 | MAPE (%) for c: 11.46 -> 9.34 | Improved |
| use Embed & Emulate instead of supervised regression with fewer training samples (n=500) | regression head · Lorenz-96 n=500 | MAPE (%) for F: 11.97 -> 11.19 | Improved |
| increase training size from 500 to 1000 (ablation) | Embed & Emulate regression head · Lorenz-96 | MAPE (%) for F: 11.19 -> 6.30 | Improved |
| use EnKI rather than regression head for uncertainty-aware estimates (ablation) | Embed & Emulate · Lorenz-96 n=1000 | MAPE (%) for F: 6.30 -> 6.59 | No clear effect |
| use EnKI posterior rather than regression head point estimates (ablation) | Embed & Emulate · Lorenz-96 n=1000 | CRPS for F: 0.412 -> 0.360 | Improved |
| use Embed & Emulate + EnKI instead of NPE-C | EnKI · Lorenz-96 n=1000 | CRPS for F: 0.593 -> 0.360 | Improved |
| use Embed & Emulate + EnKI instead of SNL+ | EnKI · Lorenz-96 n=1000 | CRPS for F: 1.304 -> 0.360 | Improved |
| remove the regression head loss (ablation) | Embed & Emulate · Lorenz-96 n=4000 | MAPE (%) for F: 8.39 -> 16.75 | Hurt |
| add empirical Bayes prior from regression head (ablation) | Embed & Emulate · Lorenz-96 n=4000 | MAPE (%) for F: 8.39 -> 3.39 | Improved |
| use learned Embed & Emulate under observation noise r=0.1, 0.2 | EnKI with Embed & Emulate · Lorenz-96 noisy observations | posterior consistency | Improved |
Closest papers by meaning
- Extracting Governing Equations from Latent Dynamics via Multi-View Contrastive Learning (NeurIPS 2026 · similarity 0.88)
- Learning Hybrid Dynamics Models with Simulator-Informed Latent States (AAAI 2024 · similarity 0.88)
- Identifying latent state transitions in non-linear dynamical systems (ICLR 2025 · similarity 0.87)
- The Dynamic-Probabilistic Consistency Gap in Chaotic Surrogate Modeling (NeurIPS 2026 · similarity 0.86)
- Learning Physics From Video: Unsupervised Physical Parameter Estimation for Continuous Dynamical Systems (CVPR 2025 · similarity 0.86)
- Contrast encodes inductive bias: separating slow noise from dynamics in predictive representation learning (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.
