Abstract:Mixed-Integer Linear Programming (MILP) is an optimization technique widely used in various fields. Primal heuristics, which reduce the search space of MILP, have enabled traditional solvers (e.g., Gurobi) to efficiently find high-quality solutions. However, traditional primal heuristics rely on expert knowledge, motivating the advent of machine learning (ML)-based primal heuristics that learn repetitive patterns in MILP. Nonetheless, existing ML-based primal heuristics do not guarantee solution feasibility (i.e., satisfying all constraints) and primarily focus on prediction for binary decision variables. When addressing MILP involving non-binary integer variables using ML-based approaches, feasibility issues can become even more pronounced. Since finding an optimal solution requires satisfying all constraints, addressing feasibility is critical. To overcome these limitations, we propose a novel reinforcement learning (RL)-based solver that interacts with MILP to find feasible solutions, rather than delegating sub-problems to traditional solvers. We design reward functions tailored for MILP, which enables the RL agent to learn relationships between decision variables and constraints. Additionally, to effectively model complex relationships among decision variables, we leverage a Transformer encoder-based graph neural network (GNN). Our experimental results demonstrate that the proposed method can solve MILP problems and find near-optimal solutions without delegating the remainder to traditional solvers. The proposed method provides a meaningful step forward as an initial study in solving MILP problems end-to-end based solely on ML.

Learning to Generate Scalable MILP Instances.

G2MILP: Learning to Generate Mixed-Integer Linear Programming Instances for MILP Solvers

A Deep Instance Generative Framework for MILP Solvers Under Limited Data Availability

MILP-FBGen: LP/MILP Instance Generation with Feasibility/Boundedness

MILP-StuDio: MILP Instance Generation via Block Structure Decomposition

ACM-MILP: Adaptive Constraint Modification Via Grouping and Selection for Hardness-Preserving MILP Instance Generation

A Multi-Agent Learning Framework for Mixed-Integer Linear Programming

Light-MILPopt: Solving Large-scale Mixed Integer Linear Programs with Lightweight Optimizer and Small-scale Training Dataset

Towards Foundation Models for Mixed Integer Linear Programming

Multi-task Representation Learning for Mixed Integer Linear Programming

Distributional MIPLIB: a Multi-Domain Library for Advancing ML-Guided MILP Methods

Manifold Interpolation for Large-Scale Multiobjective Optimization Via Generative Adversarial Networks

RL-MILP Solver: A Reinforcement Learning Approach for Solving Mixed-Integer Linear Programs with Graph Neural Networks

A multi-granularity clustering based evolutionary algorithm for large-scale sparse multi-objective optimization

Learning to Select Cutting Planes in Mixed Integer Linear Programming Solving

Scalable Multi-instance Learning

Learning optimal objective values for MILP

Mixed-Integer Linear Optimization via Learning-Based Two-Layer Large Neighborhood Search

Learning to optimize by multi-gradient for multi-objective optimization

Scalability in Manufacturing Systems: a Hybridized GA Approach.