Abstract: New class of preconditioners for iterative algorithms of sparse linear systems solution built using basis matrix method and incomplete decomposition methodology has been proposed. Algorithms with static and dynamic restriction set along with additional refinement procedures have been presented. Results of developed algorithms' testing carried out on matrices from Tim Davis Matrix Collection have been given. Basing on received results, matrix classes, applying proposed preconditioners on which resulted in iterative algorithms speed-up and/or accuracy increase, have been identified

Keywords: linear systems, sparse matrices, poor conditioned matrices, preconditioners, basis matrix method.

ACM Classification Keywords: H.4.2 Information Systems Applications: Types of Systems: Decision Support.

Link:

BUILDING PRECONDITIONERS USING BASIS MATRIX METHOD

Volodymyr Kudin, Vsevolod Bohaienko

http://www.foibg.com/ijicp/vol01/ijicp01-02-p09.pdf