An implicit shift bidiagonalization algorithm for ill-posed systems (Q1347073)

Iterative methods based on Lanczos bidiagonalization (LBD) with full reorthogonalization are considered for solving large scale discrete ill- posed linear least squares problems of the form \(\min_ x \| Ax - b \|_ 2\). Section 1 discusses the use of generalized cross-validation to estimate the optimal regularization of an LBD solution when the noise level is unknown. Section 4 studies the implicitly restarted LBD method. Section 5 explores the use of implicit restarts in the context of determining regularization parameters for the LBD, and presents some numerical examples to demonstrate their efficiency.