An approximate inverse matrix technique for arrowhead matrices

From MaRDI portal





New techniques for computing the approximate inverse matrix to an arrowhead matrix are proposed. This approach is suitable for solving large linear systems on parallel and vector processors. The recursive scheme for finding the exact inverse matrix is based on LU-factorization of the original matrix without inverting the decomposition factors. To reduce the computational work the above method is modified into an approximate inverse arrow matrix algorithm, wherein a certain number of components of the inverse matrix, lying near to the main diagonal, are retained. In a next step this method is optimized by using a moving window shifted from bottom to top such that smaller numbers of elements are stored. Further explicit preconditioned iterative schemes like Richardson's and Chebychev's ones, based on the approximate inverse matrix techniques, are presented and applied to solving linear systems. Numerical results are also reported.




Cited in
(20)








This page was built for publication: An approximate inverse matrix technique for arrowhead matrices

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4256047)