Comparisons of Commercial MIP Solvers and an Adaptive Memory (Tabu Search) Procedure for a Class of 0-1 Integer Programming Problems

Lars Magnus Hvattum, Arne Løkketangen, Fred Glover


The Boolean optimization problem (BOOP) is a highly useful formulation that embraces a variety of 0-1 integer programming problems, including weighted versions of covering, partitioning and maximum satisfiability problems. Several years ago Hvattum, Løkketangen and Glover (2006) introduced an adaptive memory (tabu search) method for BOOP which proved effective compared to competing approaches. However, in the intervening years, major advances have taken place in exact solvers for integer programming problems, leading to widely publicized successes by the leading commercial solvers XPRESS, CPLEX and GUROBI. The implicit message is that an alternative methodology for any broad class of IP problems such as Boolean Optimization Problems would now be dominated by the newer versions of these leading solvers.

We test this hypothesis by performing new computational experiments comparing the tabu search method for the BOOP class against XPRESS, CPLEX and GUROBI, and documenting improvements provided by the exact codes. The outcomes are somewhat surprising.


zero-one integer programming; Boolean optimization; commercial solvers; metaheuristics

Full Text:


Algorithmic Operations Research. ISSN: 1718-3235