On accelerate overrelaxation methods for rank deficient linear systems (Q2489451)

The authors discuss the solution of systems of linear equations \(Ax=b\) with a complex rank deficient matrix \(A\). A least squares solution of minimal norm is computed. In a first step the system is augmented to a block \(4 \times 4\) consistent system. Then a subproper accelerated overrelaxation (AOR) splitting of the augmented coefficient matrix is used [see also \textit{H. Tian}, Appl. Math. Comput.~140, No.~2--3, 485--499 (2003; Zbl 1029.65039)]. A new subproper AOR splitting is proposed and a necessary and sufficient condition for the semiconvergence of the subproper AOR iteration matrix is given. The method is illustrated by a numerical example.