H OMEWORK 5

120202: ESM4A - N UMERICAL M ETHODS

Spring 2016

Prof. Dr. Lars Linsen

School of Engineering and Science
Jacobs University

Due: Friday, March 11, 2016, at 7pm

(in the mailbox labeled “Linsen” in the entrance hall of Research I).

Problem 10: Iterative Methods for Solving Linear Equation Systems. (3+3+3+6=15 points)
Let A = L + D + U be a non-singular matrix with L being a lower triangular, D being a diagonal, and
U being an upper triangular matrix.
(a) Show that the Jacobi iteration for solving Ax = b can be written as

xk+1 = D−1 (b − (L + U )xk ) .

(b) Show that the Gauss-Seidel iteration for solving Ax = b can be written as

xk+1 = (D + L)−1 (b − U xk ) .

(c) Derive how the Successive Overrrelation iteration for solving Ax = b with weight w can be re-
written in a similar form to the one in (b).

(d) Show that the Jacobi method converges for any non-singular, strictly diagonally dominant matrix
A (for any starting point x0 ).

Problem 11: Jacobi Method. (6+7+2=15 points)

Given the system of linear equations

5x1 + 2x2 − 2x3 = 10

x1 + 6x2 − 2x3 = −6
x2 + 4x3 = 8

(a) Given start point x0 = 0, execute the first two steps of a Jacobi iteration, i. e. compute x1 and x2 .
(b) Given start point x0 = 0, execute the first two steps of a Gauss-Seidel iteration, i. e. compute x1
and x2 .

(c) Will the Jacobi and the Gauss-Seidel method converge? Explain your answer!