Recursive interpolation algorithm: A formalism for solving systems of linear equations. II: Iterative methods (Q5961646): Difference between revisions

The solution of a system of linear equations can be reduced if the given matrix can be split into four block matrices such that the Schur complement of one diagonal block is easily obtained. This leads to the concept of RIA (recursive interpolation algorithms). Part I [ibid. 76, No. 1-2, 13-30 (1996)]\ is concerned with algebraic aspects of Schur complements and their construction. Part II considers iterative methods which produce the exact solution after at most \(n\) steps. [Of course, this is interesting for problems of medium size, but one needs other arguments for large scale problems with \(n>10000\).]\ In the framework of RIA among others, the Daniel method, the Axelsson-Vassilevski method, the CG-method, the CR-algorithm, the CRNE-algorithm, the PGCR-algorithm, iterations with PD-matrices and the simplifications for SPD-matrices, the TRIA(s)-algorithm and the scaled ABS algorithm are treated.