Matrix Inversion

If
L- lower triangular matrix
U- upper triangular matrix
Then,
[A]{X}={B} can be decomposed into two matrices [L] and [U] such that
[L][U]=[A]
[L][U]{X}={B}
Similar to first phase of Gauss elimination, consider
[U]{X}={D}
[L]{D}={B}
[L]{D}={B} is used to generate an intermediate vector {D} by
forward substitution
Then, [U]{X}={D} is used to get {X} by back substitution.


Example:
Use LU decomposition to determine the matrix inverse for the following
system. Check your results by verifying that | || | | | I A A =
1
.
Matrix Inversion
5 . 21 5
5 . 61 2 6 3
27 2 10
3 2 1
3 2 1
3 2 1
= + +
= +
= +
x x x
x x x
x x x

Solution:

First, we compute the LU decomposition. The coefficient a
21
is eliminated by
multiplying row 1 by f
21
= 3/10 = 0.3 and subtracting the result from row
2. a
31
is eliminated by multiplying row 1 by f
31
= 1/10 = 0.1 and subtracting
the result from row 3. The factors f
21
and f
31
can be stored in a
21
and a
31
.

(
(

1 . 5 8 . 0 1 . 0
7 . 1 4 . 5 3 . 0
1 2 10


a
32
is eliminated by multiplying row 2 by f
32
= 0.8/(5.4) = 0.148 and subtracting
the result from row 3. The factor f
32
can be stored in a
32
.

(
(
(

352 . 5 148 . 0 1 . 0
7 . 1 4 . 5 3 . 0
1 2 10


Therefore, the LU decomposition is

Matrix Inversion
(
(
(

=
1 1481 . 0 1 . 0
0 1 3 . 0
0 0 1
] [L

(
(
(

=
352 . 5 0 0
7 . 1 4 . 5 0
1 2 10
] [U


The first column of the inverse can be computed by using [L]{D} = {B}

(
(
(

0
0
1
1 148 . 0 1 . 0
0 1 3 . 0
0 0 1
3
2
1
d
d
d


This can be solved for d
1
= 1, d
2
= 0.3, and d
3
= 0.055. Then, we can
implement back substitution

(
(
(

055 . 0
3 . 0
1
352 . 5 0 0
7 . 1 4 . 5 0
1 2 10
3
2
1
x
x
x


to yield the first column of the inverse

=
0104 . 0
059 . 0
111 . 0
} {X

Matrix Inversion

For the second column use {B}
T
= {0 1 0} which gives
{D}
T
= {0 1 0.148}. Then, Back substitution gives
{X}
T
= {0.038 0.176 0.028}.


For the third column use {B}
T
= {0 0 1} which gives
{D}
T
= {0 0 1}. Back substitution then gives
{X}
T
= {0.0069 0.059 0.187}.

Therefore, the matrix inverse is

(
(
(

187 . 0 028 . 0 0104 . 0

059 . 0 176 . 0 059 . 0
0069 . 0 038 . 0 111 . 0
] [
1
A


We can verify that this is correct by multiplying [A][A]
1
to yield the identity
matrix. For example, using MATLAB,
>> A=[10 2 -1;-3 -6 2;1 1 5];

>> AI=[0.111 0.038 0.0069;
-0.059 -0.176 0.059;
-0.0104 0.028 0.187];
Matrix Inversion

>> A*AI

ans =
1.0000 -0.0000 -0.0000
0.0000 1.0000 -0.0000
-0.0000 0.0000 1.0000