Research map / Probabilistic and scientific modeling / Dynamical systems
Interpretable Functional Koopman Learning with Non-Markovian Closure for Spatiotemporal Systems
Wanfeng Lu, He Ma, Wei Lin, Qunxi Zhu
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 Mori-Zwanzig memory correction (finite memory) to linear Koopman backbone (ablation) | MERLIN · Navier-Stokes test horizon | MSE: 9.527e-3 -> 2.035e-3 | Improved |
| add memory correction to linear backbone (ablation) | MERLIN · Wave test horizon | MSE: 6.150e-4 -> 1.659e-4 | Improved |
| add memory correction to linear backbone (ablation) | MERLIN · SST test horizon | MSE: 1.025e-1 -> 7.510e-2 | Improved |
| use finite memory (LSTM) instead of leaky memory (ablation) | MERLIN · Kuramoto-Sivashinsky test horizon | MSE: 4.529e-1 -> 4.393e-2 | Improved |
| use finite memory instead of leaky memory (ablation) | MERLIN · Wave and Navier-Stokes | MSE: 2.074e-4 -> 1.659e-4 (Wave test) | No clear effect |
| add learnable Fourier embeddings to encoder projections (ablation) | MERLIN · Navier-Stokes training horizon | MSE: 4.590e-4 -> 4.867e-5 | Improved |
| add Stiefel constraint to projection head in ROM (ablation) | MERLIN ROM · Wave ROM d=32 | total training loss: 1.125e-2 -> 5.419e-4 | Improved |
| increase number of CLS tokens K (latent dimension D=16K) (ablation) | MERLIN with FMM · Navier-Stokes | test MSE: 1.218e-3 (K=1) to 2.323e-3 (K=12) | No clear effect |
| increase ROM latent dimension from 4 to 8 (NS) (ablation) | MERLIN ROM · Navier-Stokes | MSE | Improved |
| MERLIN vs FNO on in-horizon accuracy | MERLIN · Wave training horizon | MSE: 1.012e-5 (FNO) vs 6.194e-5 (MERLIN FMM) | Hurt |
Closest papers by meaning
- Deep-Koopman-KANDy: Dictionary Discovery for Deep-Koopman Operators with Kolmogorov-Arnold Networks for Dynamics (NeurIPS 2026 · similarity 0.87)
- StreamPhy: Streaming Inference of High-Dimensional Physical Dynamics via State Space Models (NeurIPS 2026 · similarity 0.86)
- ResKoopNet: Learning Koopman Representations for Complex Dynamics with Spectral Residuals (ICML 2025 · similarity 0.86)
- KoNODE: Koopman-Driven Neural Ordinary Differential Equations with Evolving Parameters for Time Series Analysis (ICML 2025 · similarity 0.86)
- Course Correcting Koopman Representations (ICLR 2024 · similarity 0.85)
- Conformal Online Learning of Deep Koopman Linear Embeddings (NeurIPS 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.
