Research map / Probabilistic and scientific modeling / Differential equations and PDEs
From Basis to Basis: Gaussian Particle Representation for Interpretable PDE Operators
Zhihao Li, Yu Feng, Zhilu Lai, Wei Wang
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 Gaussian Field instead of plain MLP encoder/decoder (ablation) | GPO · NS2D | relative L2: 7.44×10−2 -> 3.90×10−2 | Improved |
| add PG operator on top of Gaussian field (ablation) | GPO · NS2D | relative L2: 8.57×10−2 -> 3.90×10−2 | Improved |
| use a single Gaussian per site rather than plain MLP (ablation) | GPO · NS2D | relative L2: 7.44×10−2 -> 6.28×10−2 | Improved |
| increase modal budget G from 16 to 64 (ablation) | GPO · NS2D | relative L2: 4.21×10−2 -> 3.84×10−2 | Improved |
| use Gaussian kernel instead of inverse-quadric RBF (best alternative) (ablation) | GPO · NS2D | relative L2: 4.86×10−2 -> 3.90×10−2 | Improved |
| remove center regularizer L_mu (ablation) | GPO · particle geometry | center error: 4.35×10−3 -> 1.87×10−2 | Hurt |
| relax center regularizer weight to 0.1x (ablation) | GPO · NS2D | relative L2 | Mixed |
| strengthen center regularizer weight to 5x (ablation) | GPO · NS2D | relative L2 | Mixed |
| widen FNO to match GPO parameters (ablation) | FNO · Turbulent | relative L2: 4.82×10−1 -> 4.78×10−1 | No clear effect |
| increase FNO retained modes (ablation) | FNO · Turbulent | relative L2: 4.82×10−1 -> 4.63×10−1 | Improved |
| use GPO instead of M2NO | GPO · NS2D | relative L2: 2.76E-02 vs 3.02E-02 | Hurt |
| use GPO instead of best baseline | GPO · ERA5-temp | relative L2: 3.35E-03 -> 2.26E-03 | Improved |
Closest papers by meaning
- RIGNO: A Graph-based Framework For Robust And Accurate Operator Learning For PDEs On Arbitrary Domains (NeurIPS 2025 · similarity 0.89)
- Poseidon: Efficient Foundation Models for PDEs (NeurIPS 2024 · similarity 0.86)
- Data-Efficient Operator Learning via Unsupervised Pretraining and In-Context Learning (NeurIPS 2024 · similarity 0.86)
- Latent Neural Operator for Solving Forward and Inverse PDE Problems (NeurIPS 2024 · similarity 0.86)
- PIG: Physics-Informed Gaussians as Adaptive Parametric Mesh Representations (ICLR 2025 · similarity 0.86)
- Universal Physics Transformers: A Framework For Efficiently Scaling Neural Operators (NeurIPS 2024 · similarity 0.85)
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.
