Some results on the regularization of LSQR for large-scale discrete ill-posed problems (Q1708059)

The authors derive bounds for the \(2\)-norm distance between the \(k\)-dimensional Krylov subspace of the LSQR algorithm and the \(k\)-dimensional right singular space. Therefore they show that the algorithm has better regularizing effects for severely and moderately ill-posed problems than for mildly ill-posed ones. Numerical experiments illustrating these regularizing properties of LSQR algorithm are also presented.