Perturbation bounds for the least squares problem (Q1364065)

The article investigates the effect of perturbations in \(A\) and \(b\) on the normal pseudosolution of the least squares problem \(Ax\simeq b\) without any constraints on the ranks of the original and the perturbed problem. The normal pseudosolution of the original problem is approximated by the orthogonal projection of the normal pseudosolution of the perturbed problem on the corresponding subspace of principal singular vectors. The error is estimated using singular value decomposition.