Abstract
The Generalized Conjugate Gradient (GCG) method of Concus and Golub [1] and Widlund [2] and its adjusted form (AGCG) are considered for the numerical solution of the least squares problems. Some numerical comparisons with the SOR [3] and CG [4] method are also given indicate that the GCG method is always better than the SOR [3] and that the adjusted GCG (AGCG) scheme is preferable to the CG, GCG and SOR methods.
Zusammenfassung
Die verallgemeinerte Methode der konjugierten Gradienten (GCG) von Concus und Golub [1] und Widlund [2] und ihre angepaßte Form (AGCG) werden bei der numerischen Lösung von Least-Squares-Problemen untersucht. Einige numerische Vergleiche mit dem SOR-Verfahren [3] und dem CG-Verfahren [4] werden gezogen und deuten an, daß die GCG-Methode stets besser als die SOR-Methode ist und daß das AGCG-Schema allen anderen erwähnten Methoden vorzuziehen ist.
References
Concus, P., Golub, G. H.: A generalized conjugate gradient method for nonsymmetric systems of linear equations. Lecture Notes in Economics and Mathematical Systems 134 (Glowinski, R., Lions, J. L., eds.), pp. 56–65. New York: Springer-Verlag 1976.
Widlund, O.: A Lanczos method for a class of nonsymmetric systems of linear equations. SIAM Numer. Anal.15, 801–812 (1978).
Markham, T. L., Neumann, M., Plemmons, R. J.: Convergence of a direct-iterative method for large-scale least-squares problems. LAA69, 155–167 (1985).
Freund, R.: A note on two block SOR method for sparse least squares problems. LAA88/89, 211–221 (1987).
Evans, D. J., Li, C.: The theoretical aspects of the generalized conjugate gradient method applied to the least squares problems (in preparation).
Peters, G., Wilkinson, J. H.: The least squares problems and pseudo-inverses. Comput. J.13, 309–316 (1970).
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Evans, D.J., Li, C. Numerical aspects of the generalized CG-method applied to least squares problems. Computing 41, 171–178 (1989). https://doi.org/10.1007/BF02238742
Received:
Revised:
Issue Date:
DOI: https://doi.org/10.1007/BF02238742