A new bound-and-reduce approach of nonconvex quadratic programming problems

For the nonconvex quadratic programming problem, a new linear programming relaxation bound-and-reduce algorithm is proposed and its convergence is proved. In this algorithm, a new hyper-rectangle partition technique and a new linear programming relaxation tactics are used. At the same time, the hyper-rectangular reduction method is used to raise its convergent speed. The numerical results demonstrate the effectiveness and feasibility of the proposed algorithm.