Ijetr021210

International Journal of Engineering and Technical Research (IJETR)
ISSN: 2321-0869, Volume-2, Issue-2, February 2014
7 www.erpublication.org

Abstract— The term "iterative method" refers to a wide
range of techniques which use successive approximations to
obtain more accurate solutions .In this paper an attempt to
solve systems of linear equations of the form AX=b, where A is a
known square and positive definite matrix. We dealt with two
iterative methods namely stationary(Jacobi, Gauss-Seidel, SOR)
and non-stationary (Conjugate Gradient, Preconditioned
Conjugate Gradient).To achieve the required solutions more
quickly we shall demonstrate algorithms for each of these
methods .Then using Matlab language these algorithms are
transformed and then used as iterative methods for solving these
linear systems of linear equations. Moreover we compare the
results and outputs of the various methods of solutions of a
numerical example. The result of this paper the using of
non-stationary methods is more accurate than stationary
methods. These methods are recommended for similar situations
which are arise in many important settings such as finite
differences, finite element methods for solving partial
differential equations and structural and circuit analysis .
Index Terms—Matlab language, iterative method, Jacobi.
I. STATIONARY ITERATIVE METHODS
The stationary methods we deal with are Jacobi iteration
method, Gauss-Seidel iteration method and SOR iteration
method.
A. Jacobi iteration method
The Jacobi method is a method in linear algebra for
determining the solutions of square systems of linear
equations. It is one of the stationary iterative methods where
the number of iterations is equal to the number of variables.
Usually the Jacobi method is based on solving for every
variable xi of the vector of variables =(x1,x2,….,xn) ,
locally with respect to the other variables .One iteration of the
method corresponds to solve every variable once .The
resulting method is easy to understand and implement, but the
convergence with respect to the iteration parameter k is slow.
B. Description of Jacobi’s method:-
Consider a square system of n linear equations in n variables
AX=b (1.1)
where the coefficients matrix known A is
A=
The column matrix of unknown variables to be determined X
is
and the column matrix of known constants b is
Manuscript received February 01, 2014.
PROF. D. A. GISMALLA, Dept. of Mathematics , College of Arts and
Sciences, Ranyah Branch,TAIF University ,Ranyah(Zip Code :21975) ,
TAIF, KINDOM OF SAUDI ARAIBA, Moblie No. 00966 551593472 .
The system of linear equations can be rewritten
(D+R)X=b (1.2)
where A=D+R , D =
is the diagonal matrix D of A and
Therefore, if the inverse exists and Eqn.(1.2) can be
written as
X= (b-RX) (1.3)
The Jacobi method is an iterative technique based on solving
the left hand side of this expression for X using a previous
value for X on the right hand side .Hence , Eqn.(1.3) can be
rewritten in the following iterative form after k iterations as
= (b-R ) ,k=1,2,…. (1.4)
Rewriting Eqn.(1.4) in matrix form . We get by equating
corresponding entries on both sides
i=1,…,n ,k=1,2,…. (1.5)
We observe that Eqn. (1.4) can be rewritten in the form
k=1,2,……, (1.6)
where T= (-L-U) and C = b
or equivalently as in the matrix form of the Jacobi iterative
method
= +D-1
b k=1,2,….
where T= (-L-U) and C = b
In general the stopping criterion of an iterative method is
to iterate until
<
for some prescribed tolerance .For this purpose, any
convenient norm can be used and usually the norm , i.e.
< ,
This means that if is an approximation to ,then
the absolute error is and the relative error is
provided that This implies that a
vector can approximate X to t significant digits if t is the
largest non negative integer for which <5*
C. The Matlab Program for JACOBI
The Matlab Program for Jacobi’s Method with it’s
Command Window is shown in Fig.(1.1)
D. Gauss-Seidel iteration method
The Gauss-Seidel method is like the Jacobi method,
except that it uses updated values as soon as they are available
.In general, if the Jacobi method converges, the Gauss-Seidel
Matlab Software for Iterative Methods and
Algorithms to Solve a Linear System
PROF. D. A. GISMALLA

Matlab Software for Iterative Methods and Algorithms to Solve a Linear System
method will converge faster than the Jacobi method, though
still relatively slowly.
Now if We consider again the system in Eqn.(1.1)
AX=b (1.1)
and if We decompose A into a lower triangular component
L* and a strictly upper triangular component U. Moreover if
D is the diagonal component of A and L is the strictly lower
component of A then
A= L*+U
Where
L* =L+D
Therefore the given system of linear equations in Eqn.(1.1)
can be rewritten as:
(L* +U)X=b
Using this we get
L*X=b-UX (1.7)
The Gauss-Seidel method is an iterative technique that solves
the left hand side of this expression for X, using previous
values for X on the right hand side. Using Eqn.(1.7)iteratively
,it can be written as:
= (b-U ), k=1,2,…. , (1.8)
provided that exists. Substituting for L* from above we
get
-1
(b)-(D+L)-1
U ,
k=1,2,…., (1.9)
where T = -( -1
)U and C= -1
b
Therefore Gauss-Seidel technique has the form
=T +C, k=1,2,…. (1.10)
However ,by taking advantage of the triangular forms
of D,L,U and as in Jacobi's
method ,the elements of can be computed sequentially
by forward substitution using the following form of Eqn.
(1.5) above:
= (1.11)
i=1,2,…,n , k=1,2,3,……
The computation of uses only the previous elements of
that have already been computed in advanced to
iteration k.This means that ,unlike the Jacobi method, only
one storage vector is required as elements can be over
rewritten as they are computed ,which can considered as an
advantageous for large problems .The computation for each
element cannot be done in parallel . Furthermore, the values at
each iteration are dependent on the order of the original
equations.
E. The Matlab Program for Gauss-Seidel iteration
method
The Matlab Program for Gauss-Seidel Method with it’s
Command Window is shown in the Fig.(1.2)
F. SOR iteration method (Successive Over Relaxation)
SOR method is devised by applying an extrapolation w to
the Gauss-Seidel method .This extrapolation takes the form of
a weighted average between the previous iteration and the
computed Gauss-Seidel iteration successively.
If w>0 is a constant, the system of linear equations in
Eqn.(1.1) can be written as
(D+wL)X=wb-[wU+(w-1)D] X (1.12)
Using this Eqn.(1.12) the SOR iteration method is given by
k=1,2 ,3… (1.13)
provided exists .

Now , let Tw =
and Cw= w
Then the SOR technique has the iterative form
for a constant w> 0,
k=1,2,…, (1.14)
By Eqn. (1.14) it is obvious that the method of SOR
is an iterative technique that solves the left hand side of this
expression for when the previous values for on the
right hand side are computed .Analytically ,this may be
written as:
k=1,2 ,3… (1.15)
By taking advantage of the triangular form of
(D+ wL) it can proved that .
Using this, the elements of can be computed sequentially
using forward substitution as in Eqn.(1.3) in section 1.1and
Eqn.(1.10) in section 1.2 above , i.e.
( ) ,
i=1,2,…,n, k=0,1,…. (1.16)
The choice of the relaxation factor w is not necessarily easy
and depends upon the properties of the coefficient matrix .For
positive –definite matrices it can be proved that 0<w<2will
lead to convergence, but we are generally interested in faster
convergence rather than just convergence.
G. The Matlab Program for SOR iteration method
The Matlab Program for SOR Method with it it’s
Command Window is shown in Fig.(1.3)
II. THE CONJUGATE GRADIENT METHOD
It is used to solve the system of linear equations
Ax = b (2.1)
for the vector x where the known n-by-n matrix A is
symmetric (i.e. AT
= A), positive definite (i.e. xT
Ax > 0 for all
non-zero vectors x in Rn
), and real, and b is known as well.
We denote the unique solution of this system by .
A. The conjugate gradient method as a direct method
We say that two non-zero vectors u and v are
conjugate (with respect to A) if
(2.2)
Since A is symmetric and positive definite, the left-hand side
defines an inner product
Two vectors are conjugate if they are orthogonal with respect
to this inner product. Being conjugate is a symmetric relation:
if u is conjugate to v, then v is conjugate to u. (Note: This
notion of conjugate is not related to the notion of complex
conjugate.)
Suppose that { } is a sequence of n mutually conjugate
directions. Then the{ } form a basis of Rn
, so we can expand
the solution of Ax = b in this basis:
and we see that
The coefficients are given by
(because are mutually conjugate)
This result is perhaps most transparent by considering the
inner product defined above.
This gives the following method for solving the equation Ax =
b: find a sequence of n conjugate directions, and then
compute the coefficients .
B. The conjugate gradient method as an iterative method
The direct algorithm was then modified to obtain an
algorithm which only requires storage of the last two residue

vectors and the last search direction, and only one matrix
vector multiplication.
The algorithm is detailed below for solving Ax = b
where A is a real, symmetric, positive-definite matrix. The
input vector x0 can be an approximate initial solution or 0.
repeat
if is sufficiently small then exit loop
end repeat
The result is
This is the most commonly used algorithm. The same formula
for is also used in the Fletcher–Reeves nonlinear
conjugate gradient method.
C. The code in Matlab for the Conjugate
Iterative Algorithm in Fiq.(1.4) & Fiq.(1.5)
Alternatively another Matlab Code for Conjugate
Gradient Algorithm is in Fiq.(1.5)
D. Convergence properties of the conjugate gradient
method
The conjugate gradient method can theoretically be viewed
as a direct method, as it produces the exact solution after a
finite number of iterations, which is not larger than the size of
the matrix, in the absence of round-off error. However, the
conjugate gradient method is unstable with respect to even
small perturbations, e.g., most directions are not in practice
conjugate, and the exact solution is never obtained.
Fortunately, the conjugate gradient method can be used as an
iterative method as it provides monotonically improving
approximations to the exact solution, which may reach
the required tolerance after a relatively small (compared to the
problem size) number of iterations. The improvement is
typically linear and its speed is determined by the condition
number of the system matrix A: the larger is , the
slower the improvement.
If is large, preconditioning is used to
replace the original system
with so that
gets smaller than , see below
E. The preconditioned conjugate gradient method
Preconditioning is necessary to ensure fast convergence
of the conjugate gradient method. The preconditioned
conjugate gradient method takes the following form:
repeat
if rk+1 is sufficiently small then exit loop end if
end repeat
The result is
The above formulation is equivalent to applying the
conjugate gradient method without preconditioning to the
system
where
The preconditioned matrix M has to be symmetric
positive-definite and fixed, i.e., cannot change from iteration
to iteration. If any of these assumptions on the preconditioned
is violated, the behavior of the preconditioned conjugate
gradient method may become unpredictable.
An example of a commonly used preconditioned is the
incomplete Cholesky factorization.
F. The code in Matlab for the Preconditioned Conjugate
Gradient Algorithm in Fig(1.6)
G. The flexible preconditioned conjugate gradient
method
In numerically challenging applications
sophisticated preconditioners are used, which may
lead to variable preconditioning, changing between
iterations. Even if the preconditioned matrix is
symmetric positive-definite on every iteration, the
fact that it may change makes the arguments above invalid,
and in practical tests leads to a significant slowdown of the
convergence of the algorithm presented above. Using the
Polak–Ribière formula
instead of the Fletcher–Reeves formula
may dramatically improve the convergence in this case. This
version of the preconditioned conjugate
gradient method can be called flexible, as it allows

for variable preconditioning. The implementation of
the flexible version requires storing an extra vector.
For a fixed preconditioned, so both
formulas for are equivalent in exact arithmetic, i.e.,
without the round-off error.
The mathematical explanation of the better
convergence behavior of the method with the Polak–Ribière
formula is that the method is locally optimal in this case, in
particular, it converges not slower than the locally optimal
steepest descent method.
III. CONCLUSION
When We ran each Matlab Code program and testified
with a numerical example We observe that
the non-stationary algorithms converge faster than the
stationary algorithms.
The reader can look and compare at the output result
programs in Figures denoted above ((these Figures are
Fig.(1) , Fig(2) , ….,Fig(5) and Fig(6) )) to ensure that
non-stationary algorithms especially
the Preconditioned Conjugate Algorithm approximates the
solution accurately to five decimal
places with the number of iterations N equals 5 which less
compared to the other Algorithms tackled in this
paper.
The paper also analysis the Convergence properties of the
conjugate gradient method and shows there are two methods,
the direct or the iterative conjugate.
Further, better convergence behavior can be achieved
when using Polak–Ribière formula which
converges locally optimal not slower than the locally
optimal steepest descent methed.
Furthermore, these Algorithms can are
recommended for similar situations which are arise in many
important settings such as finite differences,
finite element methods for solving partial differential
equations and structural and circuit analysis.
ACKNOWLEDGMENT
First, I would like to thank the department of Mathematics,
Tabuk University Saudi Arabia to allow me teach the Course
of Mathematica 2010 that give me the KNOWLEDGE of
what the symbolic languages are. Second, I would like to
thank the department of Mathematics, Gezira University
Sudan to teach the Course of Numerical Analysis applied with
Matlab Language during the years 2011 & 2012 that give me

the KNOWLEDGE to differentiate between the symbolic
languages and the processing languages
REFERENCES
[1] David G. lay ,linear algebra and it's applications,
2nd
ed,AddisonWesleyPublishing, Company, (August 1999).
[2]Jonathan Richard Shewchuk, An Introduction to the ConjugateGradient
Method, Without the Agonizing Pain, School of Computer Science
Carnegie Mellon University ,(August 1994).
[3]KendeleA.Atkinson, Introduction to Numerical Analysi, 2nd
ed, John
Wilev and Sons,(1988).
[4]Luenberger,David G., inear and Non linear Programming, 2nd
ed, Addison
Wesley ,Reading MA,(1984).
[5]Magnus R. Hestenes and Eduard, Methods of Conjugate Gradients for
Solving Linear, Systems, the National Bureau of Standards ,(1952).
[6]Michael Jay Holst B.S. ,Software for Solving Linear Systems with
Conjugate Gradient Methods, Colorado State University,(1987).
[7]Ortega ,J.M. , Numerical Analysis ,a second course, Academic press,
NewYork,(1972).
[8]R.BarrettM.Berry and T.Fchan and J.Demmel and J.Donato and
V.Eijkhout and R.Pozo and C.Romine and H.Vander Vorst ,
Template for the Solution of Linear Systems Building Blocks for
Iterative Methods, 2nd
,Siam,(1994).
[9]Richard L.Burden, J.DouglasFaire Numerical Analysis, 7th
ed,
wadsworth group, (2001).
[10] Ssacson,E. and H.B.Keller, Analysis of Numerical Methods, John
Wiley&Sons,New York (1996)
PROF. D. A. GISMALLA, Dept. of Mathematics , College of Arts and
Sciences, Ranyah Branch,TAIF University ,Ranyah(Zip Code :21975),
TAIF, KINDOM OF SAUDI ARAIBA, Moblie No. 00966 551593472 .

Ijetr021210

More Related Content

What's hot (19)

Viewers also liked (15)

Similar to Ijetr021210 (20)

More from Engineering Research Publication (20)

Recently uploaded (20)

Ijetr021210