Geodesic
The influence of isolated largest eigenvalues on the numerical convergence of the CG method
Zapiski Nauchnykh Seminarov POMI, Computational methods and algorithms. Part XIII, Tome 248 (1998), pp. 5-16 
A. Yu. Yeremin; I. E. Kaporin. The influence of isolated largest eigenvalues on the numerical convergence of the CG method. Zapiski Nauchnykh Seminarov POMI, Computational methods and algorithms. Part XIII, Tome 248 (1998), pp. 5-16. http://geodesic.mathdoc.fr/item/ZNSL_1998_248_a0/
@article{ZNSL_1998_248_a0,
     author = {A. Yu. Yeremin and I. E. Kaporin},
     title = {The influence of isolated largest eigenvalues on the numerical convergence of the {CG~method}},
     journal = {Zapiski Nauchnykh Seminarov POMI},
     pages = {5--16},
     year = {1998},
     volume = {248},
     language = {ru},
     url = {http://geodesic.mathdoc.fr/item/ZNSL_1998_248_a0/}
}
TY  - JOUR
AU  - A. Yu. Yeremin
AU  - I. E. Kaporin
TI  - The influence of isolated largest eigenvalues on the numerical convergence of the CG method
JO  - Zapiski Nauchnykh Seminarov POMI
PY  - 1998
SP  - 5
EP  - 16
VL  - 248
UR  - http://geodesic.mathdoc.fr/item/ZNSL_1998_248_a0/
LA  - ru
ID  - ZNSL_1998_248_a0
ER  - 
%0 Journal Article
%A A. Yu. Yeremin
%A I. E. Kaporin
%T The influence of isolated largest eigenvalues on the numerical convergence of the CG method
%J Zapiski Nauchnykh Seminarov POMI
%D 1998
%P 5-16
%V 248
%U http://geodesic.mathdoc.fr/item/ZNSL_1998_248_a0/
%G ru
%F ZNSL_1998_248_a0

Voir la notice du chapitre de livre provenant de la source Math-Net.Ru

This paper considers the dependence of the convergence history of the CG method on largest eigenvalues of a symmetric positive definite matrix. It is demonstrated that, in solving ill-conditioned linear systems, the reproduction of largest eigenvalues can be so intensive that large eigenvalues cannot be treated as isolated. On the other hand, since the moment at which the smallest isolated eigenvalues start to govern the numerical convergence of the CG method, the character of convergence mainly depends on the smallest Ritz values.