Global Journal of Computer Science and Technology: C

Software & Data Engineering

Volume 15 Issue 5 Version 1.0 Year 2015
Type: Double Blind Peer Reviewed International Research Journal
Publisher: Global Journals Inc. (USA)
Online ISSN: 0975-4172 & Print ISSN: 0975-4350

System of Linear Equations, Guassian Elimination

By Suriya Gharib, Syeda Roshana Ali, Rabia Khan , Nargis Munir
& Memoona Khanam
Fatima Jinnah Women University, Pakistan
Abstract- In this paper linear equations are discussed in detail along with elimination method.
Guassian elimination and Guass Jordan schemes are carried out to solve the linear system of
equation. This paper comprises of matrix introduction, and the direct methods for linear equations.
The goal of this research was to analyze different elimination techniques of linear equations and
measure the performance of Guassian elimination and Guass Jordan method, in order to find their
relative importance and advantage in the field of symbolic and numeric computation. The purpose of
this research is to revise an introductory concept of linear equations, matrix theorey and forms of
Guassian elimination through which the performance of Guass Jordan and Guassian elimination can
be measured.
Keywords : direct, indirect, backward stage, forward stage.
GJCST-C Classification : G.1.6

SystemofLinearEquationsGuassianElimination
Strictly as per the compliance and regulations of:

© 2015. Suriya Gharib, Syeda Roshana Ali, Rabia Khan, Nargis Munir & Memoona khanam. This is a research/review paper,
distributed under the terms of the Creative Commons Attribution-Noncommercial 3.0 Unported License http://creative
commons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction inany medium, provided the
original work is properly cited.
 System of Linear Equations, Guassian
Elimination
Suriya Gharib α, Syeda Roshana Ali σ, Rabia Khan ρ, Nargis Munir Ѡ & Memoona khanam¥

Abstract- In this paper linear equations are discussed in detail II. History of Guassian Elimination[2]
along with elimination method. Guassian elimination and
Guass Jordan schemes are carried out to solve the linear Classic books on the History of Mathematics,

2015
system of equation. This paper comprises of matrix as well as recent studies on this subject, place the
introduction, and the direct methods for linear equations. The

Year
origins of Guassian Elimination in a variety of ancient
goal of this research was to analyze different elimination texts from different places and times: China, Greece,
techniques of linear equations and measure the performance
Rome, India, medieval Arabic countries, and European
of Guassian elimination and Guass Jordan method, in order to 23
find their relative importance and advantage in the field of Renaissance. However, it is not exact to say that these
ancient texts describe what we understand today as the

( C ) Volume XV Issue V Version I

symbolic and numeric computation. The purpose of this
research is to revise an introductory concept of linear method of Guassian Elimination, since these texts
equations, matrix theorey and forms of Guassian elimination mainly present some specific problems that are solved
through which the performance of Guass Jordan and in a way that is accepted as Guassian Elimination,
Guassian elimination can be measured. butthey do not include any explicit statement of the set
Keywords: direct, indirect, backward stage, forward of rules that constitute the method of GE. The
stage. schoolbook elimination period corresponds to the
development of GE essentially as it is presented
I. Introduction incurrent high school textbooks. This period started with

A
system of equation is a set or collection of Isaac Newton who lectured on Algebraas it appeared in
equations solved together. Collection of linear Renaissance texts.
equations is termed as system of linear equations. Isaac Newton established first the rules of
They are often based on same set of variables. Various Gaussian elimination as they are still presented in
methods have been evolved to solve the linear current high school textbooks. Carl Friedrich Gauss
equations but there is no best method yet proposed for developed efficient methods for solving normal

Global Journal of C omp uter S cience and T echnology

solving system of linear equations[1]. Various methods equations, i.e., the special type of linear equations that
are proposed by different mathematicians based on the may arise in solutions of least square problems, via
speed and accuracy. However speed is an important Gaussian elimination of linear equations via hand
factor for solving linear equations where volume of computations.
computation is so large. Linear equation methods are
divided into two categories. Direct and Indirect. Each III. Guassian Elimination
category comprises of several elimination methods used Guassian elimination is the standard method for
for solving equations. this paper deals with Guassian solving linear equations. As it is a ubiquitous algorithm
elimination method, a direct method for solving system and plays a fundamental role in scientific computation.
of linear equations. An introductory portion of Guass Guassian elimination is a tool for obtaining the solution
Jordan elimination is also carried out in order to analyze of equations, to compute the determinant, for deducing
the performance of both methods. Indirect methods are rank of coefficient matrix. However Guassian
basically iterative methods and these methods have an Elimination depends more on matrix analysis and
advantage in a sense that they require fewer computation. It emphasis on block pivoting, methods of
multiplication steps for large computations. Iterative iteration and a means to improve the computed solution
methods can be implemented in smaller programs and quality. It involves two stages forward and backward
are fast enough. With study of system of linear equation stage.
one must be familiar with matrix theory that howdifferent Forward stage: Unknowns are eliminated in this stage by
operations are performed on a desired matrix to manipulation of equations and constitute an echelon
calculate the result. form.
Author α σ ρ Ѡ ¥ : Department of computer science Fatima Jinnah Backward stage: it is related with back substitution
Women University, RWP. e-mails: [email protected],
process on the reduced upper triangular method
[email protected], [email protected],
[email protected], [email protected] resulting in a solution of equation.

© 20 15 Global Journals Inc. (US)

 System of Linear Equations, Guassian Elimination

a) Steps to Solve Guassian Elimination 𝑎𝑎11 𝑎𝑎12 𝑎𝑎13 … 𝑎𝑎1𝑛𝑛 𝑎𝑎1,𝑛𝑛+1

Guassian Eliminationis systematic application of ⎡ 0 0 𝑎𝑎23 … 𝑎𝑎2𝑛𝑛 𝑎𝑎2,𝑛𝑛+1 ⎤
elementary row operations in system of equations [2]. It
⎢
⎢ 0 0 𝑎𝑎33 … 𝑎𝑎3𝑛𝑛 �� 𝑎𝑎3,𝑛𝑛+1 ⎥⎥
converts the linear system of equations to upper ⎢ : : : ⎥
triangular form, from which solution of equation is ⎣ 0 0 𝑎𝑎3𝑛𝑛 … 𝑎𝑎𝑛𝑛𝑛𝑛 𝑎𝑎𝑛𝑛,𝑛𝑛+1 ⎦
determined. Guassian elimination is summarized in the Final result after the row operation may result in
above mentioned steps[3]: above form:
i. Augmented matrix must be written for the system of 𝑎𝑎11 𝑎𝑎12 𝑎𝑎13 … 𝑎𝑎1𝑛𝑛 𝑎𝑎1,𝑛𝑛+1
linear equations.. ⎡ 0 0 𝑎𝑎23 … 𝑎𝑎2𝑛𝑛 𝑎𝑎2,𝑛𝑛+1 ⎤
ii. Transform A to upper triangular form using row
⎢ 0
⎢ 0 𝑎𝑎33 … 𝑎𝑎3𝑛𝑛 �� 𝑎𝑎3,𝑛𝑛+1 ⎥⎥
2015

operations on{A/b}. diagonal elements may not be ⎢ : : : ⎥

zero. ⎣ 0 0 0 … 𝑎𝑎𝑛𝑛𝑛𝑛 𝑎𝑎𝑛𝑛,𝑛𝑛+1 ⎦
Year

iii. Use back substitution for finding the solution of

problem. b) Sequential Algorithm – Gauss Elimination Method[4]
24
-Consider the system of linear equations with Input : Given Matrix a[1 : n, 1: n+1]
involving n variables[3]. Output : x[1 : n]
( C ) Volume XV Issue V Version I

a11x1 + a12x2 + … + a1nxn = a1,n+1 1. for k = 1 to n-1

a21x1 + a22x2 + … + a2nxn = a2,n+1 2. for i = k+1 to n
a31x1 + a32x2 + … + a3nxn = a3,n+1 3. u = aik/akk
……………… 4. for j = k to n+1
an1x1 + an3x2 + … + annxn = an,n+1 5. aij = aij – u * akj
Where aij and ai,j+1are constants, xi’s are 6. next j
variables. The system becomes equal to: 7. next i
AX=B 8. next k
a11 a12 a13 … a1n 𝑥𝑥1 9. xn = an,n+1/ann
⎡a21 a22 a23 … a2n ⎤ ⎡𝑥𝑥2 ⎤ 𝑎𝑎1,𝑛𝑛+1
⎢ a a a … a ⎥ ⎢𝑥𝑥 ⎥ 𝑎𝑎 2,𝑛𝑛+1 10. for i = n to 1 step -1
⎢ 31 32 33 3n ⎥ ⎢ 3 ⎥=� 𝑎𝑎 3,𝑛𝑛 +1 �
⎢: ∶ ∶⎥ ⎢ : ⎥ 11. sum = 0
Global Journal of C omp uter S cience and T echnology

:
𝑎𝑎 𝑛𝑛 ,𝑛𝑛 +1
⎣ a1n a2n a3n … ann ⎦ ⎣𝑥𝑥𝑛𝑛 ⎦ 12. for j = i+1 to n
13. sum = sum + aij * xj
Step 1: Store the coefficients in an augmented matrix. 14. next j
The superscript on aij means that this is the first time that 15. xi = ( ai,n+1 - sum )/aii
a number is stored in location (i, j).
16. next i
𝑎𝑎11 𝑎𝑎12 𝑎𝑎13 … 𝑎𝑎1𝑛𝑛 𝑎𝑎1,𝑛𝑛+1
17. end
⎡ 𝑎𝑎21 𝑎𝑎22 𝑎𝑎23 … 𝑎𝑎2𝑛𝑛 𝑎𝑎2,𝑛𝑛+1 ⎤
⎢ 𝑎𝑎
⎢ 31 𝑎𝑎32 𝑎𝑎33 … 𝑎𝑎3𝑛𝑛 �� 𝑎𝑎3,𝑛𝑛+1 ⎥⎥ c) Guassian Elimination Through Partial Pivoting
⎢ : : : ⎥ In actual computational practice, it is necessary
⎣ 𝑎𝑎1𝑛𝑛 𝑎𝑎2𝑛𝑛 𝑎𝑎3𝑛𝑛 … 𝑎𝑎𝑛𝑛𝑛𝑛 𝑎𝑎𝑛𝑛,𝑛𝑛+1 ⎦ to permute the rows of the matrix A (equivalently, the
Step 2 : If necessary, shift rows so that a11 ≠ 0 , equations of the system Ax = b) for obtaining a reliable
theneliminate x1 in row2 through n .The new elements algorithm.
are written aij to indicate that this is thesecond time that The permutations are performed on line as GE
a number has been stored in the matrix atlocation ( i , j ). proceeds and several permutation (or pivoting). Partial
pivoting involves the following steps:
𝑎𝑎11 𝑎𝑎12 𝑎𝑎13 … 𝑎𝑎1𝑛𝑛 𝑎𝑎1,𝑛𝑛+1
⎡ 0 𝑎𝑎22 𝑎𝑎23 … 𝑎𝑎2𝑛𝑛 𝑎𝑎2,𝑛𝑛+1 ⎤ Step 1: Select the equation having the larger 1st
⎢ 0 𝑎𝑎 𝑎𝑎 … 𝑎𝑎 � 𝑎𝑎 ⎥ coefficient in system of equation and place it at the 1st
⎢ 32 33 3𝑛𝑛 � 3,𝑛𝑛+1 ⎥
: : ⎥ entity of matrix.
⎢ :
⎣ 0 𝑎𝑎2𝑛𝑛 𝑎𝑎3𝑛𝑛 … 𝑎𝑎𝑛𝑛𝑛𝑛 𝑎𝑎𝑛𝑛,𝑛𝑛+1 ⎦
2 3 −1 1 6 4 1 9
Step 3 : New elements are written aij indicate that this is
�2 5 1 3� > � 2 5 1 3 �
thethird time that a number has been stored in the
6 4 1 9 2 3 −1 1
matrix at location ( i , j ) .

© 2015
1 Global Journals Inc. (US)
 System of Linear Equations, Guassian Elimination

Step 2 : Now perform the elementary row operations properties of the vertices has been a major research in
and convert the matrix into upper triangular form by the last decades. Also work on optimal elimination tree
using the pivot equation. The resultant matrix after was carried out, which proved of no importance in
operations may result in the form: sparsity preserving elimination trees, looking towards
6 4 1 9 optimal elimination trees could result in non linear fill.
� 0 5 1 3 � There has been minimum use of tree related graphs
0 0 −1 1 elimination outside the sparse Guassian Elimination.
Cholesky factor is described in terms of different set of
Step 3 : Make the equation equal to the number of vertices: sets of predecessor and successor, chain
variables and determine the solution of equation. elimination and elimination sets. The model of Guassian
d) LU Factorization Guassian Elimination Elimination gives a precise description of interaction

2015
LU factorization is the most important between master, sub problems which are hidden in
mathematical concept used in Guassian Elimination formulation of dynamic programming. In case of solving

Year
method. It plays a key role in the implementation of GE blocked linear equations with PDS matrices, proposed
in modern computers, and, finally, it is essential to model of computation is a straightforward extension of
facilitate the rounding error analysis of the algorithm. LU Guassian Elimination (point wise). 25
factorization method is performed in three steps [5]: Transformation associated with elimination of

( C ) Volume XV Issue V Version I

i. A=LU, compute the LU factorization. vertex is simply block elimination using submatrix of
block diagonal as block pivot.
ii. For y solve the lower triangular matrix as Ly=b by One application regarding computational model
using forward substitution method(i.e, start by is in the context of solution of asymmetric blocked
computing the first unknown as y1=b1 from the first structural system of linear system of equations which
equation, after that compute the second unknown demonstrates an indirect use in process of solution,
using the value of previous variable and so on. rules of assignments of columns to block for block
iii. Compute x for the upper triangular matrix using the elimination process. These rules provide a new concept
relation Ux=y by using backward substitution of pivoting. Consider zero subcolumn in original data
method. and non zero in partially reduced matrix. Computed
Consider the following matrix of the form subcolumn remains in the column space of some
Ax=b : subcolumn in the original data. Substituting one of these
subcolumn to the considered subcolumn is appealing.
𝑎𝑎11 𝑎𝑎12 𝑎𝑎13 𝑎𝑎14 𝑥𝑥1 𝑏𝑏1
It is complementary to sparse preserving elimination.

Global Journal of C omp uter S cience and T echnology

𝑎𝑎21 𝑎𝑎22 𝑎𝑎23 𝑎𝑎24 𝑥𝑥2 𝑏𝑏2
A=�𝑎𝑎 𝑎𝑎 𝑎𝑎 𝑎𝑎 � , 𝒙𝒙 = �𝑥𝑥 �,b=� � Cholesky Factorization method is motivated by solution
31 32 33 34 3 𝑏𝑏3
of so called normal equations that come from
𝑎𝑎41 𝑎𝑎42 𝑎𝑎43 𝑎𝑎44 𝑥𝑥4 𝑏𝑏4 linearalized KKT system in the context of Newton
Transform the matrix A into lower and upper method.
triangular matrix for the further computations. Let lower While considering interior point for solving large
triangular matrix be L and upper triangular matrix be U. scale block problems, from a numerical point of view to
solve linear system of equations is of great
consideration. Smallest height elimination trees tend to
𝑎𝑎11 0 0 0
have maximum number of leaves. Block Cholesky
𝑎𝑎21 𝑎𝑎22 0 0
L=� � includes block LU factorizations, the coefficients of
𝑎𝑎31 𝑎𝑎32 𝑎𝑎33 0
submatrix that correspond to the leaves is the original
𝑎𝑎41 𝑎𝑎42 𝑎𝑎43 𝑎𝑎44
data. Incomplete factorization consists of factorization
𝑎𝑎11 𝑎𝑎12 𝑎𝑎13 𝑎𝑎14 from leaves up to the level where data is to be
0 𝑎𝑎22 𝑎𝑎23 𝑎𝑎24 transformed several times by preceding block
U= � 0 0 𝑎𝑎33 𝑎𝑎34 � eliminations.
0 0 0 𝑎𝑎44
From these upper and lower triangular matrix V. Algorithms
perform the computations for both equations listed in a) Guassian Elimination[7]
step ii. Calculate the values of unknown. 1. {begin Reduction to Triangular form}
IV. Guassian Elimination as for i= 1 to N- 1 do
Computational Pardigm[6] f o r k = i + 1 t o N do
Sparse Guassian Elimination was studied in the aki=fl(aki/a11)
early 70’s. For the vertex elimination on the undirected for j = i + 1 to N+ 1 do
graphs a graph model was proposed. The structural akj= fl(akj – aki* aij)

© 20 15 Global Journals Inc. (US)

 System of Linear Equations, Guassian Elimination

2. { begin back-Substitution} it has been concluded that Guassian Elimination is

xN= fl(aN,N+1/ aNN) faster than the other elimination methods and it is used
in various scientific fields where large number of
for i= N-1 downto 1 do
computations are performed by elimination of variables.
for j = N downto i + 1do Our future directions are to use and develop the simple
ai,N+1 = fl(a1,N+1-aij*xj) and efficient method for non linear system of equations.
xi=fl(a1,N+1/aij)
References Références Referencias
b) Guass Jordan
1. Noreen Jamil,“ direct and indirect solvers for linear
1. {begin Reduction to Diagonal Form} system equations”, june 2012.
fori = 1 to Ndo 2. Froilán M. Dopico, “Alan Turing and the origins of
2015

fork = 1 to N(except i ) do modern Gaussian elimination”, Instituto de Ciencias

Matemáticas CSIC-UAM-UC3M-UCM, December
Year

forj= i + 1 to N+1 do
14, 2012.
aki = fl(aki/aij) 3. Yadanar Mon, Lai lai win kyi, “performance
26
for j= i+1 to N+1 do comparison of guassian elimination and guass
akj = fl(akj-akj*aij) Jordan”, International Journal of Computer &
( C ) Volume XV Issue V Version I

Communication Engineering Research (IJCCER),

2. {beginSolving Diagonal System}
Information Technology Department, Volume 2,02-
fori = 1 to N do
march-2014.
VI. Performance Comparison of Guassian 4. Luke Smith and Joan Powell: “An Alternative
Elimination with Guass Jordan Method to Gauss-Jordan Elimination: Minimizin g
Fraction Arithmetic”, The Mathematics Educator,
Guassian elimination and Guass Jordan 2011
methods are compared and analyzed on the basis of 5. J. F. Grcar, “How ordinary elimination became
execution time explained in the following table Gaussian elimination”. Historia Math.,
No of Time of Time of Guass 38(2):163_218, 2011.
variables Guassian Jordan 6. Etienne Loute, “guassian elimination as
Elimination (millisecond) computational paradigm”, August 2003.
(millisecond) 7. Nae Kaun Tsao, Ditroit Michigan, “On the
Equivalence of Gaussian Eliminatio and Gauss-
Global Journal of C omp uter S cience and T echnology

2 14 25
3 16 31 Jordan Reduction in Solving Linear Equations”,
4 20 36 February 1989.
5 26 39
6 29 56
7 46 76
Comparison through execution time
From the above mentioned results its clear that
Guassian elimination is more faster than Guass Jordan
method. Therefore, an efficient technique for solving
linear system of equations, determining the values of
unknowns in less time and less complicated procedure.
VII. Conclusion
There are different direct and indirect methods
which are used to compute the linear system of
equations. Guassian Elimination is a type of direct
method used to calculate the unknown variables. Many
scientific and engineering domains of computation may
take the form of linear equations. The equations in this
field may contain large number of variables and hence it
is important to solve these equations in an efficient
manner. This paper comprises of Guassian Elimination
method an efficient method to solve these equations.
Although the comparison on the basis of execution time
is carried out along with the Guass Jordan method and

© 2015
1 Global Journals Inc. (US)