On a regularization technique for Kovarik-like approximate orthogonalization algorithms (Q2919609)

This paper deals with four versions of Kovarik-like iterative orthogonalization algorithms for approximating the minimal norm solution of symmetric least squares problems. The authors previously observed in practical applications the so-called ``divergent behaviour'' of these algorithms, that is, a too big number of iterations can deteriorate dramatically the computed solution. In the present paper, the authors establish an almost optimal formula for the number of iterations necessary to obtain a sufficiently accurate approximation as well as to avoid the above mentioned troubles.