A fast global algorithm for multi-linearly constrained separable binary quadratic program
Guangtai Qu,Shaoze Li,Zhibin Deng,Cheng Lu
DOI: https://doi.org/10.3934/jimo.2024132
2024-10-10
Journal of Industrial and Management Optimization
Abstract:The multi-linearly constrained separable binary quadratic programming problem (MLSBQP) is a direct extension of the classical singly-linearly constrained separable binary quadratic programming problem (SLSBQP). In the literature, several algorithms have been proposed for SLSBQP which can solve problem instances with millions of variables globally. However, for the case of MLSBQP, the current state-of-the-art solver can only solve problems with up to 40000 binary variables within a reasonable computational time. In this paper, we propose an extremely efficient branch-and-bound algorithm for solving MLSBQP. The main feature of the proposed algorithm is that an efficient dual coordinate ascent algorithm is applied to obtain lower bounds, and the complexity of each inner loop of the dual coordinate ascent algorithm is . Our numerical experiments show that the proposed branch-and-bound algorithm can solve MLSBQP with up to millions of variables within only several minutes, and performs much faster than Gurobi when the number of linear constraints is small.
engineering, multidisciplinary,operations research & management science,mathematics, interdisciplinary applications