On Kaczmarz's projection iteration as a direct solver for linear least squares problems

In this paper we construct and theoretically analyze a class of direct projection algorithms for the numerical solution of linear least squares problems. These algorithms are obtained by adding supplementary directions for projection, constructed as linear combinations of the initial system rows and columns, in Kaczmarz and Extended Kaczmarz iterative methods. The above ideas are extended to the block row and column versions of the previously mentioned methods. The developed algorithms are then compared with other direct projection-based methods by the application to problems arising in multibody elasticity. (C) 2011 Elsevier Inc. All rights reserved.