You are on page 1of 2

More on iterative solution of Ax  b

Iterative
Iterativesolution
solution

Jacobi’s
Jacobi’smethod
method

Gauss’-Seidel’s
Gauss’-Seidel’s
Approach
Approach
SOR
SORmethod
method

SOR method. Successive Over-Relaxation Method

Consider the basic framework for an iterative


solution of Ax  b through an example.
4 x1  3 x2  24
3x1  4 x2  x3  30
 x2  4 x3  24

Its solution is x  3 4  5 t .

We begin with a starting solution x ( 0 )   1 1 1 t .

Gauss-Seidel frames the equations in the following


way:
x1( k )  0.75 x2( k 1 )  6
x2( k )  0.75 x1( k )  0.25 x3( k 1 )  7.5
x3( k )  0.25 x2( k )  6
For Gauss’-Seidel,

1  i 1 ( k ) n
( k 1 ) 
xi( k )    aij x j   aij x j  bi 
aii  j 1 j i 1 

In SOR, we begin right about here. We take ( 1   )


units of the previous estimate of xi and take  units
of the currently available estimate to formulate the
iterative solution:

  i 1 n 
xi( k )  ( 1   ) xi( k 1 )     a x
ij j
(k )
  aij x(j k 1 )  bi 
aii  j 1 j i 1 

The value of  is usually less than 2.0. When   1 we


get the Gauss’-Seidel method; for 0    1, we get
Under-Relaxation Method. For 1    2 , we have the
Over-Relaxation Method.

If we take   1.5 , our SOR equations appear as

x1( k )  0.5 x1( k 1 ) 


1.5
4

 3 x2( k 1 )  24 
1.5
x2( k )  0.5 x2( k 1 )  ( 3 x1( k )  4 x2( k 1 )  x3( k 1 )  30 )
4
….

Assignment. Write a program that would iteratively


obtain SOR solutions given A and b .

You might also like