Research map / Learning theory and algorithms / Prediction and decision losses
Memory-Efficient Learning of Stable Linear Dynamical Systems for Prediction and Control
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 |
|---|---|---|---|
| learn stable LDS with SOC instead of constraint generation | CG · steam sequence r=40 | reconstruction error e(A) (%): 24.90 -> 5.59 | Helped |
| learn stable LDS with SOC instead of WLS | WLS · coffee cup r=40 | reconstruction error e(A) (%): 562.17 -> 3.30 | Helped |
| learn stable LDS with SOC instead of CG/WLS | CG, WLS · UCLA, UCSD, DynTex | best error frequency / avg error | Helped |
| use SOC gradient descent instead of quadratic programs | CG, WLS · coffee cup, r=150 | memory: 3.78 GB -> 5.04 MB | Helped |
| use SOC instead of CG/WLS | CG, WLS · low subspace dimension r<20 | execution time | Hurt |
| use SOC instead of CG | CG · high dimension r>20 | execution time | Helped |
| use SOC instead of CG | CG · Franka Panda, 100 measurements | error increase vs LS (%): 11.66 -> 0.0001 | Helped |
| also update control matrix B (SOC) rather than keep B=B_ls (ablation) | CG/WLS · Franka Panda stable LDS | error increase vs LS (%): 50.14 -> 0.01 (50 measurements) | Helped |
| use SOC stable model for LQR instead of WLS | WLS · Franka Panda figure-8 tracking | average tracking error: 38.73 -> 0.0799 | Helped |
| use SOC stable model instead of CG for LQR | CG · Franka Panda figure-8 tracking | average tracking error: 0.0810 -> 0.0799 | No clear effect |
| use stable model instead of unconstrained LS | LS · Franka Panda figure-8 tracking | average tracking error: 7556 -> 0.0799 | Helped |
| use SOC instead of WLS on discontiguous measurements | WLS · Franka Panda | reconstruction error | Helped |
Closest papers by meaning
- A Memory Efficient Unified Algorithm for Online Learning of Linear Dynamical Systems (NeurIPS 2026 · similarity 0.90)
- SpectraLDS: Provable Distillation for Linear Dynamical Systems (NeurIPS 2025 · similarity 0.88)
- Neural Contractive Dynamical Systems (ICLR 2024 · similarity 0.88)
- A Novel Schur-Decomposition-Based Weight Projection Method for Stable State-Space Neural-Network Architectures (NeurIPS 2026 · similarity 0.87)
- Stable Port-Hamiltonian Neural Networks (NeurIPS 2025 · similarity 0.86)
- Learning Deep Dissipative Dynamics (AAAI 2025 · 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.