On AMG methods with F-smoothing based on Chebyshev polynomials and their relation to AMGr
Electronic transactions on numerical analysis, Tome 45 (2016), pp. 146-159.

Voir la notice de l'article provenant de la source Electronic Library of Mathematics

Summary: MacLachlan, Manteuffel, and McCormick [Numer. Linear Algebra Appl., 13 (2006), pp. 599 -- 620] introduced a new algebraic multigrid method, the so-called reduction-based algebraic multigrid method (AMGr). Different from typical multigrid methods, the smoother of the AMGr method is acting only on the fine-grid points. To analyze the AMGr method, different constants and parameters are used. Here, we further analyze the AMGr method. We show that the parameter used by MacLachlan et al. has another important property. We show that it is closely related to the root of a Chebyshev polynomial. This fact also explains the good performance of AMGr. By examining this relation with Chebyshev polynomials, we extend the concept of the AMGr method. We consider algebraic multigrid methods with fine-grid smoothers and AMG methods that are based on polynomial smoothing. We also establish bounds for the error propagation operator. The bound is minimal if Chebyshev polynomials are chosen. If more than one smoothing step is used, the error bound is smaller than the bound given for the AMGr method. For only one smoothing step, the polynomial-based AMG with Chebyshev polynomials coincides with the AMGr method. In this case, our convergence analysis gives some new explanation of the high performance of the AMGr method as well as the parameters used in the AMGr method.
Classification : 65F10, 65F20
Keywords: AMG, amgr, Chebyshev polynomials
@article{ETNA_2016__45__a18,
     author = {Gossler, Florian and Nabben, Reinhard},
     title = {On {AMG} methods with {F-smoothing} based on {Chebyshev} polynomials and their relation to {AMGr}},
     journal = {Electronic transactions on numerical analysis},
     pages = {146--159},
     publisher = {mathdoc},
     volume = {45},
     year = {2016},
     language = {en},
     url = {http://geodesic.mathdoc.fr/item/ETNA_2016__45__a18/}
}
TY  - JOUR
AU  - Gossler, Florian
AU  - Nabben, Reinhard
TI  - On AMG methods with F-smoothing based on Chebyshev polynomials and their relation to AMGr
JO  - Electronic transactions on numerical analysis
PY  - 2016
SP  - 146
EP  - 159
VL  - 45
PB  - mathdoc
UR  - http://geodesic.mathdoc.fr/item/ETNA_2016__45__a18/
LA  - en
ID  - ETNA_2016__45__a18
ER  - 
%0 Journal Article
%A Gossler, Florian
%A Nabben, Reinhard
%T On AMG methods with F-smoothing based on Chebyshev polynomials and their relation to AMGr
%J Electronic transactions on numerical analysis
%D 2016
%P 146-159
%V 45
%I mathdoc
%U http://geodesic.mathdoc.fr/item/ETNA_2016__45__a18/
%G en
%F ETNA_2016__45__a18
Gossler, Florian; Nabben, Reinhard. On AMG methods with F-smoothing based on Chebyshev polynomials and their relation to AMGr. Electronic transactions on numerical analysis, Tome 45 (2016), pp. 146-159. http://geodesic.mathdoc.fr/item/ETNA_2016__45__a18/