General scheme for solving linear algebraic problems by direct methods

Abstract

Assume that A is an m × n real matrix with m⩾ n and rank( A)= n. Let b be a real vector with m components and consider the problem of finding x=A †b , where A †=(A TA) −1A T . A general scheme for solving this problem is described. Many well-known and commonly used in practice direct methods can be found as special cases within the general scheme. This is illustrated by four examples. The general scheme can be used to study the common properties of the direct methods. The usefulness of this approach in the efforts to improve the efficiency of the direct methods for sparse matrices is demonstrated by formulating an algorithm that can be applied to any particular method belonging to the general scheme. This algorithm is implemented in several subroutines solving linear algebraic problems by different direct methods. Numerical results, obtained in a wide range of runs with these subroutines, are given.