Research map / Probabilistic and scientific modeling / Differential equations and PDEs
Fractional is Better: Learnable Derivative Orders in Neural Operator Learning
Fares B. Mehouachi, Saif Jabari
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 learnable fractional derivative input features (d-augmentation) | FNO · Burgers (nu=0.1), noisy (sigma=0.05) | relative L2 error (%): 0.250 -> 0.190 | Improved |
| add learnable fractional derivative input features (d-augmentation) | FNO · Burgers (nu=0.001), clean | relative L2 error (%): 1.138 -> 0.837 | Improved |
| add learnable fractional derivative input features (d-augmentation) | FNO · NS2D (nu=1e-5), clean | relative L2 error (%): 3.479 -> 3.428 | Improved |
| add learnable fractional derivative input features (d-augmentation) | Transolver · Burgers (nu=0.001), clean | relative L2 error (%): 18.988 -> 4.633 | Improved |
| add learnable fractional derivative input features (d-augmentation) | Transolver · KdV, noisy (sigma=0.05) | relative L2 error (%): 16.334 -> 9.588 | Improved |
| add learnable fractional derivative input features (d-augmentation) | TFNO · Darcy, clean | relative L2 error (%): 5.223 -> 4.843 | Improved |
| add learnable fractional derivative input features (d-augmentation) | LocalNO · KdV, clean | relative L2 error (%): 2.587 -> 2.352 | Improved |
| add learnable fractional derivative input features (d-augmentation) | CNO · Poisson (CNO suite), noisy (sigma=0.05) | relative L2 error (%): 3.664 -> 1.293 | Improved |
| add learnable fractional derivative input features (d-augmentation) | CNO · Wave (CNO suite), noisy (sigma=0.05) | relative L2 error (%): 1.990 -> 0.900 | Improved |
| add learnable fractional derivative input features (d-augmentation) | CNO · Darcy (CNO suite), clean | relative L2 error (%): 0.763 -> 0.752 | Improved |
| add learnable fractional derivative input features (d-augmentation) | CNO · Airfoil (CNO suite), clean, median relative L1 error | median relative L1 error (%): 0.446 -> 0.443 | No clear effect |
| lower resolution of the input grid (N=2048 down to N=64) (ablation) | d-FNO vs FNO · Noisy Burgers | relative improvement over FNO: 39% -> 14% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Transolver — Transolver: A Fast Transformer Solver for PDEs on General Geometries (ICML 2024) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Derivative-enhanced Deep Operator Network (NeurIPS 2024 · similarity 0.90)
- Efficient Training of Neural Fractional-Order Differential Equation via Adjoint Backpropagation (AAAI 2025 · similarity 0.89)
- Optimal deep learning of holomorphic operators between Banach spaces (NeurIPS 2024 · similarity 0.88)
- Neural Variable-Order Fractional Differential Equation Networks (AAAI 2025 · similarity 0.88)
- Derivative-Informed Training of Neural Operators On-the-Fly via Sketched Tangent Consistency (NeurIPS 2026 · similarity 0.88)
- Improved Operator Learning by Orthogonal Attention (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.
