Global solution of mixed-integer quadratic programs through quadratic convex reformulation

We review the quadratic convex reformulation approach for quadratic programs with integer variables. We also show the recent extensions to quadratically constrained programs and to the case of mixed-integer variables. In all these extensions, the global framework is the same: in a preprocessing step, we compute a tight equivalent reformulation of the original quadratic program that we deduce from the solution of its SDP relaxation. The equivalent reformulation is easier to solve because its continuous relaxation is a convex problem. Then, we solve the equivalent reformulation by standard BB.