Preconditioned GAOR methods for solving weighted linear least squares problems (Q2519716): Difference between revisions

The paper is interesting and well written. In order to solve the weighted linear least squares problem, one has to solve a linear system \(Hy=f\). To solve the linear system, there are the three iterative methods: Jacobi, Gauss-Seidel and SOR (successive overrelaxation), then AOR (accelerated overrelaxation) method and finally GAOR (generalized AOR) method. In order to decrease the spectral radius of the iteration matrix, and to increase the convergence rate, the authors propose to precondition the linear system, namely to take \(H^{\sim}y=f^{\sim}\); hence they introduce the preconditioned GAOR method. Two kinds of preconditioning are proposed, each one containing three preconditioners. The convergence rates of the proposed six preconditioned GAOR methods are better than the rate of the original GAOR method, whenever this is convergent. A numerical example confirms the theoretical results.