Block $LU$ factorization is stable for block matrices whose inverses are block diagonally dominant
Zapiski Nauchnykh Seminarov POMI, Computational methods and algorithms. Part XVI, Tome 296 (2003), pp. 15-26

Voir la notice de l'article provenant de la source Math-Net.Ru

Let $A\in M_n(\mathbb C)$ and let its inverse $B=A^{-1}$ be represented as an $m\times m$ block matrix that is block diagonally dominant either by rows or by columns w.r.t. a certain matrix norm. We show that $A$ possesses a block $LU$ factorization w.r.t. the partitioning defined by $B$, and the growth factor for $A$ in this factorization is bounded above by $1+\sigma$,where $\sigma=\max_{1\le i\le m}\sigma_i$ and the $\sigma_i$, $0\le\sigma_i\le1$, are the row (column) block dominance factors of $B$. Further, the off-diagonal blocks of $A$ (and of its block Schur complements) satisfy the relations $$ \|A_{ji}A_{ii}^{-1}\|\le\sigma_j, \qquad j\ne i. $$
@article{ZNSL_2003_296_a1,
     author = {A. George and Kh. D. Ikramov},
     title = {Block $LU$ factorization is stable for block matrices whose inverses are block diagonally dominant},
     journal = {Zapiski Nauchnykh Seminarov POMI},
     pages = {15--26},
     publisher = {mathdoc},
     volume = {296},
     year = {2003},
     language = {ru},
     url = {http://geodesic.mathdoc.fr/item/ZNSL_2003_296_a1/}
}
TY  - JOUR
AU  - A. George
AU  - Kh. D. Ikramov
TI  - Block $LU$ factorization is stable for block matrices whose inverses are block diagonally dominant
JO  - Zapiski Nauchnykh Seminarov POMI
PY  - 2003
SP  - 15
EP  - 26
VL  - 296
PB  - mathdoc
UR  - http://geodesic.mathdoc.fr/item/ZNSL_2003_296_a1/
LA  - ru
ID  - ZNSL_2003_296_a1
ER  - 
%0 Journal Article
%A A. George
%A Kh. D. Ikramov
%T Block $LU$ factorization is stable for block matrices whose inverses are block diagonally dominant
%J Zapiski Nauchnykh Seminarov POMI
%D 2003
%P 15-26
%V 296
%I mathdoc
%U http://geodesic.mathdoc.fr/item/ZNSL_2003_296_a1/
%G ru
%F ZNSL_2003_296_a1
A. George; Kh. D. Ikramov. Block $LU$ factorization is stable for block matrices whose inverses are block diagonally dominant. Zapiski Nauchnykh Seminarov POMI, Computational methods and algorithms. Part XVI, Tome 296 (2003), pp. 15-26. http://geodesic.mathdoc.fr/item/ZNSL_2003_296_a1/