Research map / Probabilistic and scientific modeling / Differential equations and PDEs
CFO: Learning Continuous-Time PDE Dynamics via Flow-Matched Neural Operators
Xianglong Hou, Xinquan Huang, Paris Perdikaris
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 |
|---|---|---|---|
| train on only 25% irregular time points instead of AR on full data | QuinticCFO · Burgers' | relative L2 error: 3.34(14)e-2 -> 7.09(110)e-3 | Improved |
| train on only 25% irregular time points instead of AR on full data | QuinticCFO · Diffusion-Reaction | relative L2 error: 4.23(59)e-1 -> 5.32(65)e-2 | Improved |
| train on only 25% irregular time points instead of AR on full data | QuinticCFO · Shallow Water | relative L2 error: 9.04(36)e-2 -> 1.55(22)e-2 | Improved |
| train on only 25% irregular time points instead of AR on full data | QuinticCFO · Lorenz | relative L2 error: 9.04(280)e-2 -> 6.82(48)e-2 | Improved |
| use quintic spline instead of linear spline (ablation) | CFO · Lorenz, 25% keep rate | relative L2 error: 9.39(77)e-2 -> 6.82(48)e-2 | Improved |
| use quintic spline instead of linear spline (ablation) | CFO · Burgers', 100% keep rate | relative L2 error: 5.75(72)e-3 -> 5.89(79)e-3 | No clear effect |
| use quintic spline instead of linear spline (ablation) | CFO · Diffusion-Reaction, 25% keep rate | relative L2 error: 7.25(48)e-2 -> 5.32(65)e-2 | Improved |
| increase NFE from 100% to 200% of AR budget (RK4) (ablation) | QuinticCFO · Shallow Water | relative L2 error: 0.0154 -> 0.0060 | Improved |
| increase NFE from 200% to 400% of AR budget (RK4) (ablation) | QuinticCFO · Shallow Water | relative L2 error: 0.0060 -> 0.0061 | No clear effect |
| reduce NFE to 50% of AR budget (RK4) (ablation) | QuinticCFO · Lorenz | relative L2 error: 0.0655 -> 0.0919 | Hurt |
| increase noise magnitude gamma_0 above 1e-4 (ablation) | CFO · SWE and DR | relative L2 error | Hurt |
| use flow matching instead of Neural ODE training | CFO · Burgers' | relative L2 error: 0.0275 -> 0.00589 | Improved |
Closest papers by meaning
- Mesh-Free Convolution: Learned Spectral Attenuation and Transport (NeurIPS 2026 · similarity 0.89)
- Stable Long-Horizon PDE Forecasting via Latent Structured Spectral Propagators (NeurIPS 2026 · similarity 0.89)
- On the Benefits of Memory for Modeling Time-Dependent PDEs (ICLR 2025 · similarity 0.89)
- Learning semilinear neural operators: A unified recursive framework for prediction and data assimilation. (ICLR 2024 · similarity 0.88)
- Active Learning with Selective Time-Step Acquisition for PDEs (ICML 2025 · similarity 0.88)
- CALM-PDE: Continuous and Adaptive Convolutions for Latent Space Modeling of Time-dependent PDEs (NeurIPS 2025 · 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.
