Automatic reformulation for general mixed 0–1 linear programs

The efficiency of solving mixed 0–1 linear programs by linear programming based branch-and-bound algorithms depends heavily on the formulation of the problem. In seeking a better formulation, automatic reformulation employs a range of different techniques for obtaining an equivalent formulation of a given pure or mixed integer programming problem, such that the new formulation has a tighter LP-relaxation than the original formulation. This paper surveys the use of automatic reformulation techniques for general mixed 0–1 linear programs.