Research map / Probabilistic and scientific modeling / Dynamical systems
Generalizing to New Dynamical Systems via Frequency Domain Adaptation
Tiexin Qin, Hong Yan, Haoliang Li
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 |
|---|---|---|---|
| replace ReLU with Swish activation (ablation) | FNSDA · Lotka-Volterra (inter-trajectory) | RMSE (x1e-2): 12.234 -> 3.736 | Improved |
| replace ReLU with Swish activation (ablation) | FNSDA · Glycolytic Oscillator | RMSE (x1e-2): inter 12.583 -> 8.541; extra 10.502 -> 14.918 | Mixed |
| add cosine annealing scheduler and warmup (ablation) | FNSDA · Lotka-Volterra | RMSE (x1e-2): further decreased the error by 5.0 | Improved |
| add regularization on c_e (ablation) | FNSDA · Lotka-Volterra | RMSE (x1e-2): additional error reduction of 0.3 | Improved |
| make W_env tunable during adaptation (ablation) | FNSDA · Lotka-Volterra | RMSE | No clear effect |
| use automatic Fourier mode partition instead of low-modes-only update (ablation) | FNSDA · Lotka-Volterra (inter-trajectory) | RMSE | Improved |
| update c_e during adaptation instead of fixing it (ablation) | FNSDA · Lotka-Volterra (inter-trajectory) | RMSE (x1e-2): 40.539 -> 3.736 | Improved |
| update beta_e during adaptation instead of fixing it (ablation) | FNSDA · Lotka-Volterra (inter-trajectory) | RMSE (x1e-2): 3.836 -> 3.736 | Improved |
| use L2 instead of L1 regularization on c_e (ablation) | FNSDA · Lotka-Volterra (inter-trajectory) | RMSE (x1e-2): 86.774 -> 3.736 | Improved |
| vary c_e dimension from 10 to 5 (ablation) | FNSDA · Lotka-Volterra (inter-trajectory) | RMSE (x1e-2): 3.736 -> 12.966 | Hurt |
| vary regularization lambda from 1e-4 to 1e-3 (ablation) | FNSDA · Lotka-Volterra (inter-trajectory) | RMSE (x1e-2): 3.736 -> 4.631 | No clear effect |
Closest papers by meaning
- From Fourier to Neural ODEs: Flow Matching for Modeling Complex Systems (ICML 2024 · similarity 0.87)
- PDEfuncta: Spectrally-Aware Neural Representation for PDE Solution Modeling (NeurIPS 2025 · similarity 0.86)
- Learning Robust Spectral Dynamics for Temporal Domain Generalization (NeurIPS 2025 · similarity 0.86)
- F-Adapter: Frequency-Adaptive Parameter-Efficient Fine-Tuning in Scientific Machine Learning (NeurIPS 2025 · similarity 0.86)
- Learning Interpretable Hierarchical Dynamical Systems Models from Time Series Data (ICLR 2025 · similarity 0.86)
- Out-of-Domain Generalization in Dynamical Systems Reconstruction (ICML 2024 · similarity 0.86)
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.
