Gauss Elimination: Somnath Roy

You might also like

Download as ppt, pdf, or txt
Download as ppt, pdf, or txt
You are on page 1of 18

Gauss Elimination

Somnath Roy
Department of Mechanical Engineering

1
Gauss elimination
Earliest documented methodology for solving system of linear equations.
Credit- Chinese mathematicians (109 AD), Sir Isaac Newton (1670), Gauss (1910)
formalized the notations. Later used in computer programs.

2
a11 x1  a12 x2  ......  a1n xn  b1 (1)
a21 x1  a22 x2  ......  a2 n xn  b2 (2)
.
an1 x1  an 2 x2  ......  ann xn  bn (n)

Forward elimination: subtract (1)*a21/a11 from (2), (1)*a31/a11 from (3) and so on
a11 x1  a12 x2  ......  a1n xn  b1 (1)
 x2  ......  a2 n xn  b2
a22 (2)
.
an 2 x2  ......  ann
 xn  bn (n)
a1 j ai1
where aij  aij  This is valid if the pivot
a11
a11≠0
subtract (2’)*a31/a’22 from (3’), (2’)*a41/a’22 from (3’) and so on

a11 x1  a12 x2  ............  a1n xn  b1 (1)


 x2  .............  a2 n xn  b2
a22 (2)
 x3  ...  a3n xn  b3
a33 (3)
.
a2 j ai2
an3 x3  ...  ann
 xn  bn (n ) where aij  aij  ,  0
a22
a22

After n-1 steps


a11 x1  a12 x2  ............  a1n xn  b1 (1)
 x2  .............  a2 n xn  b2
a22 (2)
 x3  ...  a3n xn  b3
a33 (3)
.
a ( n 1) nn xn  b( n 1) n (n n 1 )
( n 1) ( n  2)
a ( n  2) ( n 1) n a ( n  2) n ( n 1)
where a nn a nn  ( n  2)
, a( n  2) ( n 1)( n 1)  0
a ( n 1)( n 1)
Back substitution: subtract eqn (n)*ai-1,n/an,n from eqn (i-1) [ if an,n≠0]
a*11 x1  a*12 x2  .........  a *1( n 1) xn 1  0  b *1 (1)
a*22 x2  ...........  a *2( n 1) xn 1  0  b *2 (2)
a 33 x3  ..  a
* *
x
3( n 1) n 1 0b *
3 (3)
.
.
a*( n 1)( n 1) xn 1  0  b* n 1 (n-1)
a nn xn  b
* *
n (n)

b*1
x1  (1)
a*11
After similar n-1 steps b*2
a*11 x1  b*1 (1) x2  (2)
a*22
a *
x b
22 2
*
2 (2) b*3
x3  (3)
a x b
*
33 3
*
3 (3) a*33 These steps are valid if a*ii≠0
.  .
. .
a*( n 1)( n 1) xn 1  b*n 1 (n-1) b*n 1
xn 1  (n-1)
a*( n 1)( n 1)
a*nn xn  b*n (n)
b*n
xn  (n)
a*nn
Restriction on success of Gauss-Elimination
This method fails if any of the pivots is zero at any step.

Equations may rearranged to get non zero pivot. However, this may not work in
following cases

A pivot becomes zero during forward elimination if the equation (row) is linear
combination of previous equations.

A pivot becomes zero during back substitution if the equation (row) is linear
combination of subsequent equations also.

6
Success of Gauss-Elimination
Gauss Elimination will be successful for equations (rows/columns) with no linear
dependence.

For full rank matrices or matrices with non-zero determinants.

Infinite solutions can not be calculated using Gauss-Elimination

7
Example-1 forward elimination
2w  3x  4 y  5 z  14 (12 )
2w  3x  4 y  5 z  14 (1)
w  2x  3y  z  3 (2) 3.5 x  y  1.5 z  4 (22 )
3w  x  y  3 z  0 (3) 8 y  9 z  17 (32 )
w x  2y  z  5 (4)  y / 7  9 / 7 z  10 / 7 (4 2 )

2w  3x  4 y  5 z  14 (11 ) 2 w  3 x  4 y  5 z  14 (13 )
3.5 x  y  1.5 z  4 (21 ) 3.5 x  y  1.5 z  4 (23 )
3.5 x  7 y  10.5 z  21 (31 ) 8 y  9 z  17 (33 )
0.5 x  1.5 z  2 (41 ) (9 / 7  9 / 56) z  10 / 7 +17/56 (43 )

8
Example-1 (contd.)- back substitution
2 w  3x  4 y  5 z  14 (13 )
3.5 x  y  1.5 z  4 (23 )
8 y  9 z  17 (33 )
9 / 8 z  9 / 8 (43 )
2w  3 x  4 y  9 (1 ) 4 2 w  2 (16 )

3.5 x  y  2.5 (24 ) 3.5 x  3.5 (26 )

8 y  8 (34 ) 8 y  8 (36 )

9 / 8 z  9 / 8 (4 4 ) 9 / 8 z  9 / 8 (46 )

2 w  3x  5 (15 )  w  1
3.5 x  3.5 5
(2 )  x  1
  
8 y  8 (35 )   
 y  1
9 / 8 z  9 / 8 (45 )  z  1

9
Cost of computation
As this is used in computer programs for large matrix, it is important to estimate the floating
point operations during this process.

Consider a matrix (A)nxn: Associated b-vector has n elements also.

Forward elimination:
First step does not operate on first row. Therefore, total number of calculations on matrix
components=n2-n.
Next step takes (n-1) 2-(n-1). The last step takes 22-2 steps

10
Cost of computation
Total no. of operations in forward substitution=
n2-n+ (n-1) 2-(n-1)+…. + 22-2=n(n+1)(2n+1)/6 – n(n+1)/2=(n3-n)/3

In back substitution, The number of operations are small=1+2+…n.

Total number of operations is of the order of n3

11
Example-2 forward substitution
x  2 y  4 z  3 (1 )
x  2 y  4 z  3 (1)
5 y  13 z  13 (2)
x  2 y  z  2 (2)
5z  5 (3)
3x  y  z  4 (3)

x  2 y  4 z  3 (1 ) Backward substitution


5 z  5 (2) will follow
5 y  13z  13 (3)

12
Example-3 break of elimination steps
x  2 y  4 z  3 (1 )
x  2 y  4 z  3 (1)
 y  5 z  5 (2)
x  y  z  2 (2)
02 (3)!!!!!!
2 x  3 y  3z  1 (3)

x  2 y  4 z  3 (1 )
 y  5 z  5 (2) The process breaks here. Row permutation will not
 y  5z  7 (3) work

13
Matrix operations during Gauss elimination
x  2 y  4 z  3 (1) 1 2 4   x   3
   
x  2y  z  2 (2)  1 2 1   y    2 
3x  y  z  4 (3) 3 1 1   z   4 

1 2 4   x   3
x  2 y  4 z  3 (1 )    
 0 0 5   y    5 
5 z  5 (2)
0 5 13   z  13 
5 y  13z  13 (3)

1 2 4   x  3
x  2 y  4 z  3 (1 )    
 0 5 13   y   13 
5 y  13z  13 (2)
0 0 5   z   5 
5z  5 (3)
Matrix operations during row combinations
Multiplying row 1 by l and subtracting from row 2 is represented by
multiplication with matrix
 1 0 0
 l 1 0 
 
 0 0 1 

 1 0 0   a1 a2 a3   a1 a2 a3 
 l 1 0   b b2  
b3   b1  la1 b2  la2 b3  la3 
  1
 0 0 1   c1 c2 c3   c1 c2 c3 
 1 0 0 1 0 0 
Inverse of  l 1 0 is  l 1 0
   
 0 0 1  0 0 1 

As again addition of row 1 multiplied by l with row 2 brings back the


original matrix.
Matrix operations during row permutations
Interchanging row 3 and row 2 is obtained by multiplication with permutation
matrix
1 0 0 
0 0 1 
 
 0 1 0 

1 0 0   a1 a2 a3   a1 a2 a3 
0 0 1  b b2 b3    c1 c2 c3 
  1
 0 1 0   c1 c2 c3   b1 b2 b3 

Inverse of permutation matrix is the matrix itself


Matrix operations during Gauss elimination (ex 2)
1 2 4   x  3
   
 1 2 1   y    2 
3 1 1   z   4 

 1 2 4   x   3  1 2 4   1 0 0  1 2 4 
   
  0 0 5   y    5   0 0 5    1 1 0  1 2 1 
    
   
 3 1 1   z   4   3 1 1   0 0 1  3 1 1 

1 2 4  1 0 0  1 2 4   1 0 0   1 0 0  1 2 4   1 0 0  1 2 4 
1 2 4   x   3 0 0 5    0 1 0  0 0 5    0 1 0   1 1 0  1 2 1    1 1 0  1 2 1 
   
  0 0 5   y    5             
 0 5 13   z  13  0 5 13   3 0 1  3 1 1   3 0 1   0 0 1  3 1 1   3 0 1  3 1 1 

1 2 4   x  3 1 2 4  1 0 0  1 2 4  1 0 0   1 0 0  1 2 4   1 0 0  1 2 4 
    0 5 13   0 0 1  0 0 5   0 0 1   1 1 0  1 2 1    3 0 1  1 2 1 
  0 5 13   y   13             
 0 0 5   z   5  0 0 5  0 1 0  0 5 13  0 1 0   3 0 1  3 1 1   1 1 0  3 1 1 

You might also like