An accelerated successive orthogonal projections method for solving large-scale linear feasibility problems

AbstractWe consider linear feasibility problems in the “standard” form Ax = b, 1 ⩽ x ⩽ u. The successive orthogonal projections method may be used for solving this problem using sparse orthogonal factorizations techniques for computing the projections on Ax = b. We introduce an acceleration technique in order to speed up the (generally slow) convergence of the method. We present some numerical experiments