Research map / Learning theory and algorithms / Combinatorial and search optimization

L2P-MIP: Learning to Presolve for Mixed Integer Programming

ICLR 2024 · Learning theory and algorithms › Combinatorial and search optimization · approach: Learning branch-and-cut for MILP

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Learns instance-specific presolving decisions for mixed integer programming to simplify instances before branch-and-bound.

Problem

Presolving techniques in MIP solvers are usually set by default regardless of the input instance, and little work has studied instance-specific presolve configuration before branch-and-bound.

Approach

Learns to select presolving techniques per instance (L2P-MIP), replacing fixed solver defaults with a learned, instance-aware presolver for MIP pre-processing before B&B.

Closest papers by meaning

• Learning to Branch in Mixed Integer Programming (AAAI 2016 · similarity 0.89)
• Generalization Guarantees for Learning Score-Based Branch-and-Cut Policies in Integer Programming (NeurIPS 2025 · similarity 0.87)
• Apollo-MILP: An Alternating Prediction-Correction Neural Solving Framework for Mixed-Integer Linear Programming (ICLR 2025 · similarity 0.87)
• Predicting Lagrangian Multipliers for Mixed Integer Linear Programs (ICML 2024 · similarity 0.86)
• RoME: Domain-Robust Mixture-of-Experts for MILP Solution Prediction across Domains (NeurIPS 2025 · similarity 0.86)
• Multi-Stage Predict+Optimize for (Mixed Integer) Linear Programs (NeurIPS 2024 · similarity 0.85)