TABLE OF CONTENTS One of the problems encountered most frequently in scientific computation is the solution of systems of simultaneous linear equations. This chapter covers the solution of linear systems by Gaussian elimination and the sensitivity of the solution to errors in the data and roundoff errors in the computation.
With matrix notation, a system of simultaneous linear equations is written
In the most frequent case, when there are as many equations as unknowns, A is a given square matrix of order n, b is a given column vector of n components, and x is an unknown column vector of n components.
Students of linear algebra learn that the solution to Ax = b can be written x = A ?1 b, where A ?1 is the inverse of A. However, in the vast majority of practical computational problems, it is unnecessary and inadvisable to actually compute A ?1. As an extreme but illustrative example, consider a system consisting of just one equation, such as
The best way to solve such a system is by division:
Use of the matrix inverse would lead to
The inverse requires more arithmetic a division and a multiplication instead of just a division and produces a less accurate answer. Similar considerations apply to systems of more than one equation. this is even true in the common situation where there are several systems of equations with the same matrix A
