Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Learning Space-Time Continuous Latent Neural PDEs from Partially Observed States
Valerii Iakovlev, Markus Heinonen, Harri Lähdesmäki
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 linear instead of k-NN interpolation (ablation) | latent neural PDE · Shallow Water coarser grid | test MAE: 0.046 ± 0.002 -> 0.034 ± 0.001 | Improved |
| use linear instead of k-NN interpolation (ablation) | latent neural PDE · Shallow Water finer grid | test MAE: 0.031 ± 0.003 -> 0.017 ± 0.003 | Improved |
| use linear instead of k-NN interpolation (ablation) | latent neural PDE · Navier-Stokes coarser grid | test MAE: 0.087 ± 0.006 -> 0.069 ± 0.009 | Improved |
| use IDW instead of k-NN interpolation (ablation) | latent neural PDE · Navier-Stokes coarser grid | test MAE: 0.087 ± 0.006 -> 0.066 ± 0.006 | Improved |
| use linear instead of k-NN interpolation (ablation) | latent neural PDE · Scalar Flow finer grid | test MAE: 0.040 ± 0.016 -> 0.026 ± 0.006 | Improved |
| change interpolation method (k-NN, linear, IDW) (ablation) | latent neural PDE · Shallow Water original grid | test MAE: 0.017 ± 0.002 -> 0.016 ± 0.002 | No clear effect |
| change interpolation method (k-NN, linear, IDW) (ablation) | latent neural PDE · Scalar Flow original grid | test MAE: 0.019 ± 0.001 -> 0.018 ± 0.000 | No clear effect |
| test on coarser grid than training grid (linear) (ablation) | latent neural PDE · Shallow Water | test MAE: 0.016 ± 0.002 -> 0.034 ± 0.001 | Hurt |
| test on finer grid than training grid (linear) (ablation) | latent neural PDE · Shallow Water | test MAE: 0.016 ± 0.002 -> 0.017 ± 0.003 | Mixed |
| increase latent state dimension d from 1 to 5 (ablation) | latent neural PDE · Shallow Water, Navier-Stokes, Scalar Flow | test MAE | Improved |
| use latent-space model (d=3,5) instead of data-space model (d=1) (ablation) | latent neural PDE · partially observed data | test MAE | Improved |
| train on partially observed instead of fully observed Shallow Water (vs. best baseline) | Ours · Shallow Water | test MAE: 0.014 ± 0.002 -> 0.016 ± 0.002 | Improved |
Closest papers by meaning
- CALM-PDE: Continuous and Adaptive Convolutions for Latent Space Modeling of Time-dependent PDEs (NeurIPS 2025 · similarity 0.90)
- Space-Time Continuous PDE Forecasting using Equivariant Neural Fields (NeurIPS 2024 · similarity 0.90)
- AROMA: Preserving Spatial Structure for Latent PDE Modeling with Local Neural Fields (NeurIPS 2024 · similarity 0.89)
- Stable Long-Horizon PDE Forecasting via Latent Structured Spectral Propagators (NeurIPS 2026 · similarity 0.89)
- P$^2$C$^2$Net: PDE-Preserved Coarse Correction Network for efficient prediction of spatiotemporal dynamics (NeurIPS 2024 · similarity 0.88)
- Vectorized Conditional Neural Fields: A Framework for Solving Time-dependent Parametric Partial Differential Equations (ICML 2024 · similarity 0.88)
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.
