Computing projections with LSQR (Q678216): Difference between revisions

LSQR uses the Golub-Kahan bidiagonalization process to solve sparse least squares problems with and without regularization. In some cases, projections of the right-hand side vector are needed, rather than the least squares solution itself. It is shown that projections as well as the least squares solutions may be obtained from the bidiagonalization as linear combinations of the theoretically orthogonal vectors.