Mixed integer solver tuning improved by one-pass large model selection
Nominate-then-Adjudicate: LLM-Assisted One-Pass and Per-Instance MIP Solver Tuning
Machine Learning
Summary
Optimizing many parameters for mixed-integer programming solvers can make them work better on specific problems but usually takes a long time for each case. The authors propose a two-step approach: first, choose a small set of good parameter options for all cases, then quickly pick the best one for each problem using a single call to a large pre-trained model. This quick selection method avoids costly testing on every problem and still improves solver performance significantly compared to standard settings and other fast tuning methods.
What this means in practice
- •For operations teams: Speed up tuning of optimization solvers on diverse large-scale problems by using a one-pass selection method requiring no additional solver runs.
- •For enterprise optimization software developers: Integrate rapid per-instance solver parameter selectors to improve solver outcomes in software that handles large, varied optimization workloads.$Commercial implications: This method enables embedding AI-assisted solver tuning in commercial optimization tools for improved efficiency and performance.
Authors
Yifan Zhao, Qianyang Zhao, Xinyu Guan, Kai Wei, Yuming Deng
Abstract
Mixed-integer programming solvers expose hundreds or even thousands of parameters whose best setting varies per instance, yet per-instance search is prohibitively expensive. We frame this as configuration for heterogeneous Big Data optimization workloads and propose a nominate-then-adjudicate pipeline built around two design choices. First, a fixed portfolio is sampled once and reused across instances, removing per-instance configuration search at inference; on the evaluated families, the resulting shortlist retains substantial oracle headroom, making selection within it the empirical focus. Second, the fixed 20-feature representation can map structurally different instances to the same vector, preventing a feature-only selector from distinguishing them even when they prefer different configurations. The foundation model receives no task-specific weight updates, and selection is one-pass: one model call and no solver calls at inference. The study represents structural Variety through distinct formulation families and decision Value through solver-performance outcomes; the one-call, zero-solve protocol specifies the decision interface and leaves end-to-end latency unmeasured. Among comparisons exceeding a 20% practical-difference threshold, the complete adjudicator records 538 wins and 185 losses against the top nominee, while its win rate against the default rises from 84.9% to 90.2%. Against per-instance configuration baselines under a matched one-pass protocol it leads the cluster-level methods and the best-performing evaluated per-instance baseline, ISAC+.