Iterative regularization based on singular analysis (Q2563469)

A method for solving ill conditioned linear systems \(Ax=b\) is proposed. This method is based on the conjugate gradient method. The basic difference is that also approximations of the singular values of matrix \(A\) are computed in a relatively fast way. Thus the numerical rank of matrix \(A\) and the necessary number of iterations can be assessed. The method is effective especially for matrices of low rank.