New variable neighbourhood search based 0-1 MIP heuristics

Abstract

In recent years many so-called matheuristics have been proposed for solving Mixed Integer Programming (MIP) problems. Though most of them are very efficient, they do not all theoretically converge to an optimal solution. In this paper we suggest two matheuristics, based on the variable neighbourhood decomposition search (VNDS), and we prove their convergence. Our approach is computationally competitive with the current state-of-the-art heuristics, and on a standard benchmark of 59 0-1 MIP instances, our best heuristic achieves similar solution quality to that of a recently published VNDS heuristic for 0-1 MIPs within a shorter execution time.

Keywords

• 0-1 mixed Integer Programming
• matheuristics
• variable neighbourhood search
• pseudo-cuts
• convergence