Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Fusing Backdoors, Machine Learning, and Optimization for Large-Scale Parametric Mixed-Integer Programs

About

Large-scale optimization problems are often solved repeatedly under similar structural conditions, leading to substantial computational overhead. This occurs in applications such as power systems, transportation, and supply chain networks, where the underlying structure is fixed while parameters frequently vary under perturbations. This paper proposes a Learning to Optimize (LTO) framework that accelerates the solution of large-scale general mixed-integer problems by leveraging the concept of a backdoor, i.e., a subset of variables that drive most of the computational complexity. The proposed BIPC framework consists of three phases. Phase I is an identification procedure that discovers a backdoor for a set of instances in the distribution. Phase II uses supervised learning to develop machine learning models that, given an instance, predict values for bounded-domain backdoor variables and intervals for wide-domain backdoor variables. These predictions define a reduced optimization problem where the predictions constrain the backdoor variables, while the other variables remain free. Phase III optimizes this reduced problem and, if necessary, applies a correction step to restore feasibility or the optimality guarantees. Experiments on real-world, large-scale problems show substantial reductions in solution time with only a limited loss in solution quality. The framework enables organizations to solve large-scale optimization problems efficiently in the presence of frequent perturbations, such as unexpected events, demand fluctuations, or operational changes. Because these changes affect parameters rather than the problem structure, BIPC can quickly provide high-quality, feasible solutions, offering a practical approach to integrating machine learning into existing optimization pipelines.

El Mehdi Er Raqabi, Pascal Van Hentenryck• 2026

Related benchmarks

TaskDatasetResultRank
Large-scale OptimizationMMCNP Hard
PG Mean5
4
Large-scale OptimizationMMCNP Very-Hard
PG Mean (%)0.04
4
Large-scale OptimizationSLAP Hard
PG Mean1.1
4
Large-scale OptimizationSLAP Very-Hard
Performance Gap (PG) Mean8
4
Large-scale OptimizationCOURSE Very-Hard
Performance Gap (Mean)5
4
Large-scale OptimizationOCP Hard
PG Mean (%)7
4
Large-scale OptimizationOCP Very-Hard
PG Mean0.8
4
Large-scale OptimizationCOURSE Hard
PG Mean0.4
4
Large-scale OptimizationOFP Hard
PG Mean (%)9
3
Large-scale OptimizationOFP Very-Hard
Performance Gap (PG) Mean0.13
3
Showing 10 of 10 rows

Other info

Follow for update