Algorithms for bound constrained quadratic programming problems (Q1122320)

The authors present an algorithm which combines standard active set strategies with the gradient projection method for the solution of quadratic programming problems subject to bounds. The algorithm of this paper requires the accurate solution of systems of linear equations. This implies that either the system matrices are of moderate size, or have special structure. It is shown, that if the quadratic objective function is bounded below on the feasible set then termination occurs at a stationary point in a finite number of iterations. The authors describe a set of test problems in which five parameters can be varied (size, number of active constraints at the starting point and at the solution, degeneracy of the solution). The numerical results show that the number of iterations and the execution time required by the gradient projection method are almost always less than those of the active set strategy.