Author(s) :
Matthias Bollhöfer
,
Youcef Saad
Preprint series :
University of Minnesota Supercomputing Institute Research Report UMSI 2000/39
MSC 2000
- 65F05 Direct methods for linear systems and matrix inversion
-
65F10 Iterative methods for linear systems
-
65F50 Sparse matrices
Abstract :
This paper discusses the relations between a broad class of
incomplete LU factorization techniques and factorized sparse
approximate inverse techniques based on computing
triangular matrices Z,W such that ZT AW is approximately
diagonal. We will show that most of these approaches are essentially
equivalent to approximately inverting the triangular factors obtained
from a modified incomplete LU factorization of the initial system.
Keywords :
sparse matrices, ILU, modified ILU, sparse approximate inverse, AINV