Research map / Probabilistic and scientific modeling / Variational and generative sampling
Monte Carlo Neural PDE Solver for Learning PDEs via Probabilistic Representation
Rui Zhang, Qi Meng, Rongchan Zhu, Yue Wang, Wenlei Shi, Shihua Zhang, Zhi-Ming Ma, Tie‐Yan Liu
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 Heun's method instead of Euler for the convection process (ablation) | MCNP-10 · 2D Navier-Stokes E3 (nu=1e-4) | relative l2 error (%): 11.572 -> 3.101 | Improved |
| use Heun's method instead of Euler for the convection process (ablation) | MCNP-20 · 2D Navier-Stokes E4 (nu=1e-5) | relative l2 error (%): 9.249 -> 5.776 | Improved |
| add interpolation trick (ablation) | MCNP-10 · 2D Navier-Stokes E4 (nu=1e-5) | relative l2 error (%): 61.719 -> 5.825 | Improved |
| add interpolation trick (ablation) | MCNP-20 · 2D Navier-Stokes E3 (nu=1e-4) | relative l2 error (%): 8.970 -> 2.999 | Improved |
| add interpolation trick (ablation) | MCNP-10 · 2D Navier-Stokes E3 (nu=1e-4) | relative l2 error (%): 3.168 -> 3.101 | No clear effect |
| refine time step from 10 to 20 frames (ablation) | MCNP · 1D convection-diffusion E1 | relative l2 error (%): 0.090 -> 0.119 | Hurt |
| refine time step from 10 to 100 frames (ablation) | PINO · 1D convection-diffusion E4 | relative l2 error (%): 6.287 -> 0.257 | Improved |
| use MCNP-10 instead of PINO-100 | MCNP · 2D Navier-Stokes E2 (nu=1e-5) | relative l2 error (%): 7.554 -> 7.055 | Improved |
| use MCNP-100 instead of PINO-100 with Neumann boundary | MCNP · 1D Allen-Cahn E3 | relative l2 error (%): 1.705 -> 0.394 | Improved |
| use MCNP-20 instead of MCNP-100 (ablation) | MCNP · 1D Allen-Cahn E5 (eps=0.0001) | relative l2 error (%): 0.303 -> 0.285 | No clear effect |
| unsupervised MCNP-20 vs supervised FNO with more training trajectories (500 to 2000) (ablation) | FNO · 2D Navier-Stokes E3 (nu=1e-4) | relative l2 error (%): 4.754 -> 2.619; MCNP-20 2.999 | Mixed |
| use MCNP instead of Eulerian PSM++ solver (more steps) | MCNP · Navier-Stokes E1 | relative l2 error (%): 4.102 -> 2.397 | Improved |
Closest papers by meaning
- MC^2: Monte Carlo Correction for Fast Elliptic PDE Solving (NeurIPS 2026 · similarity 0.90)
- Text2PDE: Latent Diffusion Models for Accessible Physics Simulation (ICLR 2025 · similarity 0.89)
- P$^2$C$^2$Net: PDE-Preserved Coarse Correction Network for efficient prediction of spatiotemporal dynamics (NeurIPS 2024 · similarity 0.88)
- MultiPDENet: PDE-embedded Learning with Multi-time-stepping for Accelerated Flow Simulation (ICML 2025 · similarity 0.87)
- Mechanistic PDE Networks for Discovery of Governing Equations (ICML 2025 · similarity 0.87)
- CALM-PDE: Continuous and Adaptive Convolutions for Latent Space Modeling of Time-dependent PDEs (NeurIPS 2025 · similarity 0.87)
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.
