Original title: Exact Combinatorial Optimization with Temporo-Attentional Graph Neural Networks
Authors: Mehdi Seyfi, Amin Banitalebi-Dehkordi, Zirui Zhou, Yong Zhang
This article explores the integration of machine learning algorithms into combinatorial optimization, a field that seeks to find the best solution within a given set of variables and constraints. Over the years, significant progress has been made in both research and industry. Inspired by the success of deep learning, researchers have recently turned to machine learning models to enhance existing combinatorial optimization solvers. Specifically, this paper focuses on two crucial aspects of machine learning algorithms in this context: temporal characteristics and attention. By incorporating temporal information and bipartite graph attention into the branch-and-bound (B&B) algorithm for variable selection, the authors argue that the solver’s performance can be improved. To support their claims, they provide intuitions and numerical results from various standard datasets commonly used in the literature and competitions. The code for their implementation is also made available.
Original article: https://arxiv.org/abs/2311.13843