A numerically stable reduced-gradient type algorithm for solving large-scale linearly constrained minimization problems

In this paper the authors present a reduced-gradient type algorithm for solving large-scale linearly constrained minimization problems. During each iteration of the algorithm linearly systems are solved using a preconditioned conjugate-gradient scheme. The preconditioning scheme uses orthogonal transformations, thus providing numerical stability. The total storage used by the algorithm may be predicted before beginning the calculations. The authors present some numerical experiments which confirm the reliability of the algorithm.