The convergence behaviour of preconditioned CG and CG-S in the presence of rounding errors
Incomplete Choleski decompositions and modified versions thereof are quite effective preconditioners for the conjugate gradients method. It is explained why Modified Incomplete Choleski may lead to more iteration steps, in some cases, whereas it should theoretically be more effective than standard Incomplete Choleski. It is also shown, by carefully analyzing a numerical example, why the convergence behaviour of Conjugate Gradients-Squared can be, sometimes unacceptable, irregular.
Unable to display preview. Download preview PDF.
- Concus, P., Golub, G.H., O'Leary, D.P., A generalized conjugate gradient method for the numerical solution of elliptic partial differential equations, in: Sparse Matrix Computations (J.R. Bunch and D.J. Rose, eds), pp.309–332, Academic Press, New York, 1976Google Scholar
- Kaasschieter, E.F., The solution of non-symmetric linear systems by bi-conjugate gradients or conjugate gradients squared, Delft University of Technology, Faculty of Mathematics and Informatics, Report 86-21, 1986Google Scholar
- Paige, C.C., The computation of Eigenvalues and Eigenvectors of Verg Large Sparse Matrices, Ph.D.thesis, Univ. of London, 1971Google Scholar
- Parlett, B.N., Misconvergence in the Lanczos algorithm, Department of Mathematics, Univ.of California at Berkeley, Report PAM-404, 1987Google Scholar
- Van der Vorst, H.A., Preconditioning by incomplete decompositions, Ph.D.Thesis, Univ. of Utrecht, 1982Google Scholar