Research map / Learning theory and algorithms / Combinatorial and search optimization
An Algorithm for Learning Switched Linear Dynamics from Data
Guillaume O Berger, Monal Narasimhamurthy, Kandai Watanabe, Morteza Lahijanian, Sriram Sankaranarayanan
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 proposed gap-based algorithm instead of kLR | Switched linear ID · Benchmark 1, N=20 | one-step error: 0.34 -> 0.17 | Improved |
| use proposed algorithm instead of kLR | Switched linear ID · Benchmark 3, N=100 | one-step error: 0.48 -> 0.23 | Improved |
| use proposed algorithm instead of kLR | Switched linear ID · Benchmark 3, N=100 | training time (s): 0.41 -> 506.29 | Hurt |
| use proposed algorithm instead of kLR | Switched linear ID · Benchmark 2, N=500 | one-step error: 0.27 -> 0.06 | Improved |
| use proposed algorithm instead of kLR | Switched linear ID · Acrobot | one-step error: 1.80 -> 0.005 | Improved |
| use proposed algorithm instead of kLR | Switched linear ID · Cartpole | one-step error: 0.04 -> 0.016 | Improved |
| use proposed algorithm instead of kLR | Switched linear ID · Acrobot | training time (s): 0.18 -> 7.56 | Hurt |
| increase data points N from 20 to 500 | Proposed approach · Benchmark 1 | one-step error: 0.17 -> 0.05 | Improved |
| increase data points N from 100 to 500 | Proposed approach · Benchmark 3 | training time (s): 506.29 -> 1184.16 | Hurt |
| increase N from 20 to 500 | k-Linear Regression · Benchmark 2 | one-step error: 0.29 -> 0.27 | No clear effect |
| replace MILP solver with proposed algorithm as N grows | Proposed approach vs MILP (Gurobi) · synthetic microbenchmarks | computation time | Improved |
| vary state dimension n or number of modes m | Proposed vs MILP · synthetic microbenchmarks | computation time | No clear effect |
Closest papers by meaning
- Stability Analysis of Switched Linear Systems with Neural Lyapunov Functions (AAAI 2024 · similarity 0.88)
- A Memory Efficient Unified Algorithm for Online Learning of Linear Dynamical Systems (NeurIPS 2026 · similarity 0.86)
- Identifying latent state transitions in non-linear dynamical systems (ICLR 2025 · similarity 0.85)
- Learning system dynamics without forgetting (ICLR 2025 · similarity 0.85)
- In-context Learning of Linear Dynamical Systems with Transformers: Approximation Bounds and Depth-separation (NeurIPS 2025 · similarity 0.85)
- Learning to Relax: Setting Solver Parameters Across a Sequence of Linear System Instances (ICLR 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.
