Research map / Probabilistic and scientific modeling / Dynamical systems
Switching Linear Dynamics for Variational Bayes Filtering
Philip Becker-Ehmck, Jan Peters, Patrick Van Der Smagt
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 |
|---|---|---|---|
| Use normally distributed rather than Concrete switching variables (ablation) | Switching Linear Dynamics · Reacher | MSE: 1.06E-02 -> 3.39E-03; 5.73E-02 -> 1.85E-02; 1.56E-01 -> 4.97E-02 | Improved |
| Use normally distributed rather than Concrete switching variables (ablation) | Switching Linear Dynamics · 3-Ball Maze | MSE: 2.28E-03 -> 1.30E-03; 1.22E-02 -> 5.52E-03; 3.40E-02 -> 1.38E-02 | Improved |
| Use the proposed stochastic switching-dynamics model instead of DVBF Fusion | Switching Linear Dynamics · Reacher | MSE: 4.90E-03 -> 3.39E-03; 2.97E-02 -> 1.85E-02; 8.25E-02 -> 4.97E-02 | Mixed |
| Use the proposed stochastic switching-dynamics model instead of DVBF Fusion | Switching Linear Dynamics · 3-Ball Maze | MSE: 4.33E-03 -> 1.30E-03; 2.03E-02 -> 5.52E-03; 4.88E-02 -> 1.38E-02 | Improved |
| Use a hard forward pass rather than the soft weighted mixture of transition matrices (ablation) | Switching Linear Dynamics · Simulated tasks | prediction accuracy | Hurt |
| Use Concrete switching variables as the discretization interval increases (ablation) | Switching Linear Dynamics · Single-agent maze at varying temporal discretization scales | absolute prediction error growth | Hurt |
| Use Concrete switching variables instead of normally distributed switching variables (ablation) | Switching Linear Dynamics · Single-agent maze at small discretization intervals | absolute prediction error | Improved |
| Use current observations only rather than the full future observation sequence for the inverse measurement model (ablation) | Online filtering inference · Simulated tasks | prediction quality | No clear effect |
| Represent switching variables stochastically rather than treating transition dynamics deterministically (ablation) | Switching Linear Dynamics · 3-Ball Maze | wall-interaction representation | Improved |
| Use a stochastic switching-dynamics model | Switching Linear Dynamics · FitzHugh-Nagumo | normalized multi-step predictive performance | No clear effect |
Closest papers by meaning
- Identifying latent state transitions in non-linear dynamical systems (ICLR 2025 · similarity 0.89)
- Extracting Latent State Representations with Linear Dynamics from Rich Observations (ICML 2022 · similarity 0.89)
- Active Learning for Nonlinear System Identification with Guarantees (JMLR 2022 · similarity 0.88)
- Dynamics Are Learned, Not Told: Semi-Supervised Discovery of Latent Dynamics Geometries For Zero-Shot Policy Adaptation (ICML 2026 · similarity 0.88)
- End-to-End Identifiable and Consistent Recurrent Switching Dynamical Systems (NeurIPS 2026 · similarity 0.87)
- The Laplacian Keyboard: Beyond the Linear Span (NeurIPS 2026 · 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.
