Research map / Architectures and efficiency / RNNs and normalization
QPKO: Differentiable QP-Embedded Deep Koopman Framework for Modeling Nonlinear Systems
Runze Tian, Peng Kou
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 QP-based mapping to compute the global linear model instead of the second-best Koopman framework | deep Koopman model · TSR (T=150, T=200) | NMAE: 46.9% and 41.1% reduction | Improved |
| use QP-based mapping instead of the second-best framework | deep Koopman model · CartPole (T=150, T=200) | NMAE: 26.5%, 27.9% reduction | Improved |
| use QP-based mapping instead of the second-best framework | deep Koopman model · VDP (T=150, T=200) | NMAE: 22.7%, 13.3% reduction | Improved |
| use QP-based mapping instead of the second-best framework | deep Koopman model · WT (T=150, T=200) | NMAE: 7.9%, 9.3% reduction | Improved |
| use end-to-end QP-based mapping instead of the second-best framework | deep Koopman model · CartPole training time | wall-clock time: 3.1x speedup | Improved |
| use end-to-end QP-based mapping instead of the second-best framework | deep Koopman model · VDP training time | wall-clock time: 2.6x speedup | Improved |
| use end-to-end QP-based mapping instead of the second-best framework | deep Koopman model · TSR training time | wall-clock time: 1.9x speedup | Improved |
| use QP-based mapping instead of alternating LS training (EDMD-DL) | deep Koopman model · WT training time | wall-clock time: 0.23±0.27 -> 0.56±0.02 | Hurt |
| use QP-based mapping instead of second-best baseline in MPC | Koopman MPC · CartPole | final setpoint error: 61.8% reduction | Improved |
| use QP-based mapping instead of second-best baseline in MPC | Koopman MPC · VDP | final setpoint error: 27.3% reduction | Improved |
| use QP-based mapping instead of second-best baseline in MPC | Koopman MPC · TSR | final setpoint error: 36.4% reduction | Improved |
| use QP-based mapping instead of LS, end-to-end, or independent linear model in MPC | Koopman MPC · TSR | control success: baselines fail; QPKO 0.51±0.39 | Improved |
Closest papers by meaning
- Differentiation Through Black-Box Quadratic Programming Solvers (NeurIPS 2025 · similarity 0.85)
- Deep-Koopman-KANDy: Dictionary Discovery for Deep-Koopman Operators with Kolmogorov-Arnold Networks for Dynamics (NeurIPS 2026 · similarity 0.85)
- Efficient Parametric SVD of Koopman Operator for Stochastic Dynamical Systems (NeurIPS 2025 · similarity 0.85)
- ResKoopNet: Learning Koopman Representations for Complex Dynamics with Spectral Residuals (ICML 2025 · similarity 0.84)
- Deep Distributed Optimization for Large-Scale Quadratic Programming (ICLR 2025 · similarity 0.84)
- DiLQR: Differentiable Iterative Linear Quadratic Regulator via Implicit Differentiation (ICML 2025 · similarity 0.83)
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.
