Research map / Time series / Causal and generative time series
On-Line Learning of Linear Dynamical Systems: Exponential Forgetting in Kalman Filters
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use online AR(2) instead of spectral filtering or persistence | Dow Jones d0, d1, and d2 · Dow Jones, first T = 100 elements | error | Helped |
| Use spectral filtering instead of persistence | Dow Jones d0 · Dow Jones d0, first T = 100 elements | error: 1.47 vs. 940 | Hurt |
| Use spectral filtering instead of persistence | Dow Jones d1 · Dow Jones d1, first T = 100 elements | error: 3.58 vs. 299 | Hurt |
| Use spectral filtering instead of persistence | Dow Jones d2 · Dow Jones d2, first T = 100 elements | error: 11.02 vs. 4689 | Hurt |
| Increase process-noise variance | AR(2) approximation of the Kalman filter · Synthetic LDS with varying process and observation noise | approximation to Kalman filter | Helped |
| Increase regression depth (ablation) | AR(s + 1) · Synthetic LDS | prediction error | Helped |
| Reduce process-noise magnitude (ablation) | AR(s + 1) approximation of the Kalman filter · Synthetic LDS with varying process and observation noise | convergence with regression depth | Hurt |
| Set process noise to zero (ablation) | Kalman filter with G = F = 1 · One-dimensional LDS, W = 0 | remainder-term decay | Hurt |
Closest papers by meaning
- Predictively-Oriented Kalman Filtering (NeurIPS 2026 · similarity 0.88)
- Outlier-robust Kalman Filtering through Generalised Bayes (ICML 2024 · similarity 0.86)
- Learning invariant representations of time-homogeneous stochastic dynamical systems (ICLR 2024 · similarity 0.86)
- A Memory Efficient Unified Algorithm for Online Learning of Linear Dynamical Systems (NeurIPS 2026 · similarity 0.86)
- Finite Sample Analyses for Continuous-time Linear Systems: System Identification and Online Control (NeurIPS 2025 · similarity 0.85)
- Consistent Long-Term Forecasting of Ergodic Dynamical Systems (ICML 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.