A bidiagonalization algorithm for solving large and sparse ill-posed systems of linear equations (Q1111337)

A bidiagonalization (Lanczos type) method is used for solving regularized least squares problems of the form \(\| b-Ax\|^ 2_ 2+\mu^ 2\| x\|^ 2_ 2\) or \(\| b-Ax\|^ 2_ 2+\mu^ 2\| Bx\|^ 2_ 2\to Min.\) with a large sparse \(m\times n\)-matrix A and some (n-p)\(\times n\)-matrix B. The regularization parameter is determined using cross-validation. Author's conclusion is that the suggested method is more flexible and robust than the conjugate gradient method.