Towards learning-enhanced solvers for multi-parametric optimization
Abstract
Neural networks (NNs) have been employed in various ways to accelerate the solution of multi-parametric optimization problems (OPs), which prominently arise in model predictive control (MPC). Early approaches focused on approximating the parametric solution (e.g., the optimal control law) directly by NNs. More recently, learning-based techniques have been used to emulate or enhance solver iterations. For the direct approximation of the parametric solution, it has been demonstrated that exploiting structural properties of the underlying OP facilitates the design of suitable NN architectures. In this paper, we show that structural analysis can also be leveraged to systematically integrate learning into classical optimization algorithms, thereby enhancing their performance. We formalize our concept for the case of multi-parametric quadratic programs (mpQPs), which frequently arise in linear-quadratic MPC, as a representative application. More precisely, we demonstrate that tailored NNs enable a parameter-dependent fine-tuning of the step sizes in a standard dual projected gradient scheme. The design of the NNs is guided by the observation that optimal step sizes, which permit obtaining the exact optimal solution in finitely many iterations, are piecewise constant functions over the parameter space (e.g., the state space in MPC). We establish key theoretical properties of the proposed approach and illustrate its effectiveness in a numerical case study.
Related
- Comparative Performance Analysis of Optimization Solvers for Linear Support Vector Machines: Impact of Preconditioning and Dimensionality An Empirical Study Using the Benchopt Framework
- Student Psychology-Based Optimization Algorithm Enhanced by Teaching–Learning Mechanism for Global Optimization and Real-World Issues
- Generation of Custom Solvers in Rust for Convex Optimization
- Soliton Profiles in 1D: Enhanced Neural-Network Solvers and Constrained Deep Ritz Methods
- Computing Lower Bounds on the Nonnegative Rank via Non-Convex Optimization Solvers
BibTeX
@misc{teichrib2026towards,
title = {{Towards learning-enhanced solvers for multi-parametric optimization}},
author = {Dieter Teichrib and Fabian König and Moritz Schulze Darup},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.16520},
eprint = {2610.16520},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.16520},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - Towards learning-enhanced solvers for multi-parametric optimization
AU - Dieter Teichrib
AU - Fabian König
AU - Moritz Schulze Darup
PY - 2026
DA - 2026/10/12
PB - arXiv
DO - 10.48550/arxiv.2610.16520
UR - https://arxiv.org/abs/2610.16520
AB - Neural networks (NNs) have been employed in various ways to accelerate the solution of multi-parametric optimization problems (OPs), which prominently arise in model predictive control (MPC). Early approaches focused on approximating the parametric solution (e.g., the optimal control law) directly by NNs. More recently, learning-based techniques have been used to emulate or enhance solver iterations. For the direct approximation of the parametric solution, it has been demonstrated that exploiting structural properties of the underlying OP facilitates the design of suitable NN architectures. In this paper, we show that structural analysis can also be leveraged to systematically integrate learning into classical optimization algorithms, thereby enhancing their performance. We formalize our concept for the case of multi-parametric quadratic programs (mpQPs), which frequently arise in linear-quadratic MPC, as a representative application. More precisely, we demonstrate that tailored NNs enable a parameter-dependent fine-tuning of the step sizes in a standard dual projected gradient scheme. The design of the NNs is guided by the observation that optimal step sizes, which permit obtaining the exact optimal solution in finitely many iterations, are piecewise constant functions over the parameter space (e.g., the state space in MPC). We establish key theoretical properties of the proposed approach and illustrate its effectiveness in a numerical case study.
ER -