Research map / Probabilistic and scientific modeling / Differential equations and PDEs
MDBench: Benchmarking Data-Driven Methods for Model Discovery
Amirmohammad Ziaei Bideh, Aleksandra Georgievska, Jonathan Gryak
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 |
|---|---|---|---|
| increase noise from clean to SNR 40 dB | MD methods · PDE vs ODE datasets | log NMSE ratio: 3.13 ± 2.09 vs 0.88 ± 2.12 | Hurt |
| use linear model WSINDy instead of PySR | WSINDy · PDEs SNR >= 30 dB | NMSE: W=455.0, p < 10^-4 | Improved |
| use linear model WSINDy instead of Operon | WSINDy · PDEs SNR >= 30 dB | NMSE: W=466.0, p < 10^-4 | Improved |
| use GP method PySR instead of ESINDy | PySR · ODEBench SNR >= 30 dB | NMSE: W=9774, p <= 0.0001 | Improved |
| use sparse linear SINDy instead of EQL | SINDy · ODEBench noisy data | complexity: W=6951.5, p<10^-8 | Improved |
| increase noise from SNR 30 dB to 20 dB or lower | ODEFormer · ODEBench | NMSE | Hurt |
| increase spatial dimension from 1D to 3D | PDEFIND · Heat (Solar) | runtime: 40x and 465x slower | Hurt |
| increase spatial dimension from 1D to 3D | PySR · Heat (Solar) | runtime: 33x in 2D and 12x in 3D | Hurt |
| add custom forcing functions and six state variables | linear MD methods · Heat (Laser) and RD (Cylinder) | equation recovery | Hurt |
| use pretrained transformer End2end without dynamical-system pretraining | End2end · ODEBench | NMSE: W=12795, p<10^-10 | Hurt |
Closest papers by meaning
- SPDEBench: An Extensive Benchmark for Learning Stochastic PDEs (NeurIPS 2026 · similarity 0.89)
- CausalDynamics: A large‐scale benchmark for structural discovery of dynamical causal models (NeurIPS 2025 · similarity 0.88)
- DISCO: learning to DISCover an evolution Operator for multi-physics-agnostic prediction (ICML 2025 · similarity 0.86)
- APEBench: A Benchmark for Autoregressive Neural Emulators of PDEs (NeurIPS 2024 · similarity 0.86)
- No Equations Needed: Learning System Dynamics Without Relying on Closed-Form ODEs (ICLR 2025 · similarity 0.86)
- Skip the Equations: Learning Behavior of Personalized Dynamical Systems Directly From Data (ICML 2025 · 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.
