Matrix Inversion and Eigenvalue
Matrix Inversion and Eigenvalue
VALUE PROBLEMS
CROUT’S METHOD
To find the inverse of the square matrix A. Decompose A in to
A=LU where L is a lower triangular matrix and U is an unit upper
triangular matrix. A-1=(LU)-1 =U-1 L-1 .
If L is lower triangular, then L-1 is also lower triangular. Also if U is
L
upper triangular, then U-1 is also upper triangular. SinceL L-1=I we can
find the lower triangular matrix L-1 such that L L-1=I .
Since U U-1 = I ,we can find the upper triangular matrix U-1 such that
U U-1 = I . Having known L-1 and U-1 , we get A-1 =U-1 L-1 .
Decomposition of A in to A= LU .
Let us consider
a11 a12 a13 l11 0 0 1 u12 u13
A a21 a22 a23 , L l21 l22 0 andU 0 1 u23
a31 a32 a33 l31 l32 l33 0 0 1
l11 0 0 1 u12 u13 a11 a12 a13
l l22 0 0 1 u23 a21 a22 a23
21
l31 l32 l33 0 0 1 a31 a32 a33
l11 l11u12 l11u13 a11 a12 a13
l l u l l u l u a a a
21 21 12 22 21 13 22 23 21 22 23
1 0 0 1 2 3
L
0 1 0 , U 0
1 4
2 2 2
0 0 1
1 2 5 1 0 0 4 3 5 / 2
A1 0 1 4 0 1 0 4 3 2
0 0 1 1 1 1/ 2 1 1 1/ 2
(A:B) where i=2,3.By this all elements in the first column of (A:B)
except a11 are made zero. Now (A:B) takes the form
a11 a12 a13 1
( A : B) 0 b22 b23 c
0 b32 b33 d
Now take the pivot b22 . Now considering b22 as pivot we will
make b32 in the second column of (A:B) as zero. Now multiply the
second row by b32 and add to the corresponding element of the
b22
3rd row. Now b32 is reduced to zero. Now (A:B) takes the form
a11 a12 a13 1
( A : B) 0 b22 b23 c
0 0 c33 e
Since the element a11 is zero we will interchange the first and the
second row. The reduced system is
1 2 0 0 1 0
0 1 1 1 0 0
3 1 4 0 0 1
By performing R3+(-3)R1,we get
1 2 0 0 1 0
0 1 1 1 0 0
0 7 4 0 3 1
By performing R3+7R2 , we get
1 2 0 0 1 0
0 1 1 1 0 0
0 0 3 7 3 1
7
x31
x11 2 x21 0 3
4
x21 x31 1 x21
3
3 x31 7 8
x11
3
x12 2 x22 1 x32 1
x22 x32 0 x22 1
3x32 3 x12 1
1
x33
x13 2 x23 0 3
1
x23 x33 0 x23
3
3 x33 1
2
x13
3
8 2
3 1
3
Hence A1
4
1
1
3 3
7 1
1
3 3
GAUSS JORDAN METHOD
This method is a modification of the above Gauss elimination
method. In this method, we consider the augmented matrix
(A:B) to be a11 a12 a13 1 0 0
( A : I ) a21 a22 a23 0 1 0
a31 a32 a33 0 0 1
The coefficient matrix A can be brought to a unit matrix by
making all the elements above and below the leading diagonal are
zeros and all the leading diagonal elements are 1’s. Then (A:I)
takes the form
1 0 0 a b c
( A : I ) 0 1 0 d e f
0 0 1 g h i
Now the inverse of A is
a b c
A1 d e f
g h i
EXAMPLE
Find the inverse of the matrix A using Gauss Jordan method where
2 1 1
A 3 2 3
1 4 9
SOLUTION:
Consider (A:I) to be 2 1 1 1 0 0
( A : I ) 3 2 3 0 1 0
1 4 9 0 0 1
Add multiples of first row to the other rows so that the elements
in the second row first column and the third row first column
become zero.
2 1 1 1 0 0 3
R2 R2 R1
0 1 3 3
0
2
1
2 2 2 1
3
R R3 R1
7 17 1 2
0 0 1
2 2 2
Make the element in the third row and the second column as
zero by using the element in the second row and the second
column. 2 1 1 1 0 0
1 3 3
0 1 0 R3 R3 7 R2
2 2 2
0 0 2 10 7 1
Make the element in the second row and the third column zero
by using the element in the third row and the third column .
2 1 1 1 0 0
1 17 3 3
0 0 6 R2 R2 R3
2 4 4 4
0 0 2 10 7 1
Make the element in the first row and third column to
zero by using the element in the third row and third
column.
7 1
2 6
1 0 2 2
0 1 17 3 1
0 6 R1 R1 R3
2 4 4 2
0 0 2 10 7 1
Make the element in the first row and second column to zero by
using the element in the second row and second column.
2 0 0 6 5 1
1 17 3 1
0 0 6 R1 R1 R3
2 4 4 2
0 0 2 10 7 1
Make all the diagonal elements in the coefficient matrix to unity.
Then (A:I) becomes,
5 1
3 R1
2 2 R
1 0 0 1
2
0 17 3
1 0 12 R2 R2 * 2
2 2
0 0 1
7 1 R3 R3 1
5
2 2 2
an1 an 2 ann
On expansion it gives an nth degree equation in λ called
the characteristic equation of the matrix A.Its roots λi ,
(i=1,2,………..n) are called the eigen values (or) latent roots
and corresponding to each eigen value, equation (2) will have non
zero solution X=[x1 x2 xn ]T which is known as the eigen
vector.
POWER METHOD
In many engineering problems it is required to compute the
numerically largest eigen value and the corresponding eigen
vector.In such cases, the following iterative method can be used. If
X1,X2,……………Xn be the eigen vectors corresponding to the eigen
values λ1 , λ2 , λn then an arbitrary column vector can be
written as, X k1 X1 k2 X 2 .................. kn X n .
Then
AX k1 ( AX 1 ) k2 ( AX 2 ) ................... kn ( AX n ).
AX k1 (1 X 1 ) k2 (2 X 2 ) ................... kn (n X n ).
A2 X k1 (12 X 1 ) k2 (2 2 X 2 ) ................... kn (n 2 X n ).
Then Ar X k11r X 1 k2 2 r X 2 .................. kn n r X n
2
r
n
r
A X 1 k1 X 1 k2 ................. kn
r r
1 1
1 2 ...................................... n .
Therefore, everytime we multiply the column vector by A, it
becomes nearer to the eigen vector X1 . Then we make the largest
component of the resulting column vector unity to avoid the
the factor k1 .
Thus start with a column vector X which is as near the solution
as possible and evaluate AX which is written as (1) X (1)
after normalisation.
.This gives the first approximation λ(1) to the eigen value
and X(1) to the eigen vector. Similarly we evaluate
AX(1) = λ(2) X(2) which gives the second approximation. We repeat
this process till X(r) - X(r-1) becomes negligible. Then λ(r) will be the
largest eigen value and X(r) ,the corresponding eigen vector.
This iterative procedure for finding the dominant eigen value of a
matrix is known as Rayleigh’s power method.
1
-1 -1
NOTE: Rewriting AX=λX as X= λA X, we have A X= X. If we use
this equation the above method yields the smallest eigen value.
15 4 3
A 10 12 6
20 4 2
SOLUTION
Let the initial approximation to the eigen vector be X=[1 1 1]T .Then
15 4 3 1 8 0.444
AX 10 12 6 1 4 18 0.222 18 X (1)
20 4 2 1 18 1
So,the first approximation to the eigen value is -18 and the
corresponding eigen vector is [-0.444 0.222 1]T .Now
15 4 3 0.444 1
AX (1) 10 12 6 0.222 10.548 .105 (2) X (2)
20 4 2 1 0.736
The second approximation to the eigen value is -10.548 and the
eigen vector is [1 -0.105 -0.736]T . Repeating the above process,
0.93
AX (2) 18.948 0.361 (3) X (3)
1
1
AX (3) 18.394 0.415 (4) X (4)
0.981
0.995
AX (4) 19.698 0.462 (5) X (5)
1
1
AX (5) 19.773 0.480 (6) X (6)
0.999
0.997
AX (6) 19.922 0.49 (7) X (7).
1
1
AX (7) 19.956 0.495 (8) X (8).
0.999
Since λ(7) = λ(8) and X(7) =X(8) approximately the dominant
eigen value and the corresponding eigen vector are given by
1 1
(8) X (8) 19.956 0.495 (or )20 0.5
0.999 1
PROBLEMS FOR PRACTICE:
Find the dominant eigen values and the corresponding eigen
vectors of the following matrices using power method.
1 1 0 0.5
(1) 1 2 1 ....................... ANS :3, 1
0 1 1 0.5
1 1 2 0.25
(2) 0 3 1 ........................ ANS :5, 0.5
0 0 5 1
JACOBI METHOD
Let A be a given real symmetric matrix. Its eigen values are real
and there exists a real orthogonal matrix B such that B-1AB is a
diagonal matrix D. Jacobi method consists of diagonalising A by
applying a series of orthogonal transformations B1,B2,………..Br such
that their product B satisfies the equation B-1AB=D. For this
purpose,we choose the numerically largest non-diagonal element aij
and form a 2*2 sub matrix,
aii aij
A1 whereaij a ji
a ji a jj
which can be easily diagonalised.
• Consider an orthogonal matrix, cos sin
B1
sin cos
This equation gives four values of Ө , but to get the least possible
rotation , we choose -π/4 ≤ Ө ≤ π/4 .Thus (1) reduces to a diagonal
matrix.
As a next step, the largest non diagonal element (in magnitude) in
the new rotated matrix is found and the above procedure is
repeated using the orthogonal matrix B2
In this way , a series of a series of such transformations are
performed so as to annihilate the non diagonal elements.After
making ‘r’ transformations we obtain
Br 11 Br 11...................B11 AB1...............Br 1 Br B 1 AB
As r→∞, B-1 AB approaches a diagonal matrix whose diagonal
elements are the eigen values of A. Also the corresponding
columns of B=B1B2B3…………Br are the eigen vectors of A.
EXAMPLE: Using jacobi’s method, find all the eigen values and
eigen vectors of the matrix 1 2 2
A 2 3 2
2 2 1
SOLUTION: Here the largest non diagonal elements is a13=a31=2.
Also a11=1,a33=1.
2a13 2* 2
tan 2 .
a11 a33 1 1 4
Then
1 1
0
cos 0 sin 2 2
B1 0 1 0 0 1 0 & B11 BT
sin 0 sin 1 1
0
2 2
Hence the eigen values of the given matrix are 5,1,-1 and the
corresponding eigen vectors are the columns of
1 1 1 1 1
1 1 0
0 2
2
2 2 2 2
2
1 1 1 1
B B1 B2 0 1 0 0 0
1 1
2 2
2 2
0 0 0 1 1 1 1
2 2
2
2
2
DISADVANTAGES:
The element annihilated by a transformation, may not
remain zero during the subsequent transformations.
This method can be used only for symmetric matrices.
PROBLEMS FOR PRACTICE:
(1).Obtain using Jacobi’s method,all the eigen values and eigen
vectors of the matrix
2.536, 0.006,1.469
1 1 0.5
0.531 0.721 0.444
(1). A 1 1 0.25 ...................... ANS :
0.461 0.686 0.562
0.5 0.25 2
0.71 0.094 0.698
4, 2, 6
1 1
5 0 1 2 0 2
(2). A 0 2 0 ............................. ANS :
0 1 0
1 0 5 1
1
0
2 2