You are on page 1of 13

ITERATIVES

 
METHODS
Jacobi Method
 Suppose we are trying to solve a system of linear equation Mx = b.
If we assume that the diagonal entries are non-zero (true if the
matrix M is positive definite), then we may rewrite this equation as:

Dx + Moffx = b

Where:
D is the diagonal matrix containing the diagonal entries of M and Moff
contains the off-diagonal entries of M. Because all the entries of the
diagonal matrix are non-zero, the inverse is simply the diagonal matrix
whose diagonal entries are the reciprocals of the corresponding entries
of D.
Thus, we may bring the off-diagonal entries to the right hand side
and multiply by D-1:

x = D-1(b - Moffx)

You will recall from the class on iteration, we now have an


equation of the form x = f(x), except in this case, the argument is
a vector, and thus, one method of solving such a problem is to
start with an initial vector x0.
Gauss-Seidel
 For the Jacobi method applied to a system of n linear
equations defined by Mx = b, we start with an initial vector x0
and iterate the function:
 f(x) = D-1(b − Moffx) To begin, let xk = (xk, i), D = (di, j), b = (bi),
Mk, : represent the kth row of M.
 Because the matrix D-1 is a diagonal matrix, in essence, we are
calculating the following n operations:
xk + 1,1 = d1,1-1 ⋅ (b1 − Moff;1,: ⋅ xk)
xk + 1,2 = d2,2-1 ⋅ (b2 − Moff;2,: ⋅ xk)

xk + 1,n = dn,n-1 ⋅ (bn − Moff;n,: ⋅ xk)


 Consider the last entry: even though we have calculated
almost all the new entries of xk + 1, we are still using xk to
calculate the last entry. If the sequence is converging.
 Thus, the Gauss-Seidel method is a refinement of the Jacobi
method which may be stated as follows:
 To calculate xk + 1, first let xk + 1 = xk and then update this vector
with each calculation:
xk + 1,1 = d1,1-1 ⋅ (b1 − Moff;1,: ⋅ xk + 1)
xk + 1,2 = d2,2-1 ⋅ (b2 − Moff;2,: ⋅ xk + 1)


xk + 1,n = dn,n-1 ⋅ (bn − Moff;n,: ⋅ xk + 1)
Gauss-Seidel Relaxation
 The Gauss-Seidel method is a technique for solving the
equations of the linear system of equations Ax = b one at a
time in sequence, and uses previously computed results as
soon as they are available,

bi   
(k ) ( k 1)
(k ) j  i aij x j j i aij x j
xi 
aii
 There are two important characteristics of the Gauss-Seidel
method should be noted. Firstly, the computations appear to be
serial. Since each component of the new iterate depends upon
all previously computed components, the updates cannot be
done simultaneously as in the Jacobi method. Secondly, the
new iterate depends upon the order in which the equations are
examined. If this ordering is changed, the components of the
new iterates (and not just their order) will also change.
 In terms of matrices, the definition of the Gauss-Seidel
method can be expressed as

(k )
xi  ( D  L) 1 (Ux ( k 1)  b)

 where the matrices D, -L and -U represent the diagonal,


strictly lower triangular, and strictly upper triangular parts of,
respectively.
 The Gauss-Seidel method is applicable to strictly diagonally
dominant, or symmetric positive definite matrices .
Example.

Figure 5. The effect of freezing the boundary on several levels


of a surface.
Convergence of Iteratives
Methods
 If A is symmetric and positive definite, the Gauss-Seidel
method converges.

If A is symmetric and the matrix is the form:


 a11  a12 ....  a1n 
 a a22 ....  a2 n 
 21
D  L U   . . . 
 . .
. . 
 a  a2 n .... ann 
 1n

And is positive definite, the Jacobi method is convergent.


 If A is symmetric and positive definite, the relaxation method
converges if and only if 0 < w <2.
 If w < 1 the method is named subrelajation and if w > 1,
Overrelaxation.
 If A is symmetric, positive definite and tridiagonal, the optimal
value of w for the convergence of the relaxation method is:

2
w
1  1  Pj 2
where:
Pj: The spectral radius of the matrix Jacobi iteration method.
Bibliography
 http://
www.ana.iusiani.ulpgc.es/metodos_numericos/document/apun
tes/Parte_4.pdf
 http://www.google.com.co/imgres?
imgurl=http://www.unavarra.es/personal/victor_dominguez/po
rtMatlab.jpg&imgrefurl=http://www.unavarra.es/personal/vict
or_dominguez/libroMatlab.htm&usg=__g0H6VfxnJXinSq932
hKACQW-
4mE=&h=423&w=596&sz=61&hl=es&start=19&itbs=1&tbni
d=qcEPT90K9KGOJM:&tbnh=96&tbnw=135&prev=/images
%3Fq%3DGauss-Seidel%2Bcon%2Brelajaci%25C3%25B3n
%26hl%3Des%26sa%3DG%26gbv%3D2%26tbs%3Disch:1

You might also like