(HW2 LS EE571) Akhtar Rasool

EE 571 Linear Systems Home Work 2 Page 1 / 28
EE 571 Linear Systems Spring 2014

Home Work # 2

Due Date: 19-03-2014
Submitted By: Akhtar Rasool
Student ID: 17197

Faculty of Engineering and Natural Sciences,
Sabanci University, Istanbul
Turkey


Contents Page No.
Question # 01 3
Question # 02 3
Part (i) 3
Part (ii) 4
Part (iii) 6
Part (iv) 6
Question # 03 6
Question # 04 8
Question # 05 9
Question # 06 10
Part (a) 10
Part (b) 10
Part (c) 11
Question # 07 11
Part (a) 11
Part (b) 13
Part (c) 13
Question # 08 14
Question # 09 22
Question # 10 23
Part (a) 25
Part (b) 25
Part (c) 26


Question # 01
Let S be a vector space and let F S. Prove that F is a subspace of S if for all x, y F and for all
scalars u, , ux + y F.
Solution-Proof:
To show that F is a subspace of S, we need to prove that:
(a) If x, y F , then xy is in F.
(b) If o is any real number and x is any vector in F, then ox is in F.
Thus let
F
1
= u
1
x +
1
y and F
2
= u
2
x +
2
y
be vectors in F. Then
F
1
+F
2
= (u
1
x +
1
y) +(u
2
x +
2
y) = (u
1
+u
2
)x +(
1
+
2
)y
which is in F.
Also, if c is a scalar, then
cm
1
= c(u
1
x +
1
y) = (cu
1
)x + (c
1
)y
is in F.
Hence F is a subspace of S.
Question # 02
Let A = _
1 1
1 4
1 4
4
2
2
1 1 u
_
i. What is basis of the column span of A.
ii. Is this an orthonormal basis? If not determine an orthonormal basis for the column span of
A.
iii. Determine the QR factorization of the matrix of A.
iv. Verify your result in part (iii) using MATLAB.
Solution (i)-Basis of the column span of A: The reduced Echelon form of A is determined as follows;
= _
1 1
u S
u S
4
6
2
u u 4
_ , R
2
R
1
, R
3
R
1
, R
4
R
1

= _
1 1
u 1
u S
4
6S
2
u u 4
_ , by _
1
S
] R
2


= _
1 u
u 1
u u
14S
6S
4
u u 4
_ , by R
1
+R
2
, R
3
SR
2

= _
1 u
u 1
u u
14S
6S
1
u u 4
_ , by _
1
4
] R
3

= _
1 u
u 1
u u
u
u
1
u u u
_ , by R
1

14
S
R
3
, R
2
+
6
S
R
3
, R
4
+4R
3

The original vectors corresponding to all non-zero columns are a basis.
x
1
= |1 1 1 1]
t
, x
2
= |1 4 4 1]
t
, x
3
= |4 2 2 ]
t

Solution (ii):
A set of vectors form an orthonormal set if all vectors in the set are mutually orthonogonal and all of unit
length.
For orthogonality:
u
|
u
j
= (| j)
u
1
u
2
= __
1
1
1
1
_ _
1
4
4
1
__ = 1 +4 +4 1 = 6 u
Since the above one is not satisfied so this set is not an orthonormal set.
Now transform to an orthonormal basis:
S = Spon __
1
1
1
1
_ , _
1
4
4
1
_ , _
4
2
2
u
__
Let us call the three vectors above x
1
, x
2
and x
3
. First, we will compute the norm of the vector x
1
;
x
1
= 1 +1 +1 +1 = 4 = 2
Thus the first vector in the orthonormal basis is;
u
1
=
x
1
x
1
= _
12
12
12
12
_
Next we use x
2
and x
3
to find vectors orthogonal to u
1
.

y
2
= x
2
pruj
u
1
x
2
= x
2
(x
2
. u
1
)u
1
= _
1
4
4
1
_ __
1
4
4
1
_ _
12
12
12
12
___
12
12
12
12
_
= _
1
4
4
1
_ 4_
12
12
12
12
_ = _
S
2
2
1
_

y
3
= x
3
pruj
u
1
x
3
= x
3
(x
3
. u
1
)u
1
= _
4
2
2
u
_ __
4
2
2
u
_ _
12
12
12
12
___
12
12
12
12
_
= _
4
2
2
u
_ 2_
12
12
12
12
_ = _
S
S
1
1
_
Now y
2
and y
3
span a subspace of H
4
which is orthogonal to u
1
. We will repeat the above procedure on
the set {y
2
, y
3
]. First, we compute the norm of the vector y
2
:
y
2
= 2S4 + 2S4 + 2S4 + 2S4 =
1u
2
= S
Thus the second vector in the orthonormal basis is;
u
2
=
y
2
y
2
= _
12
12
12
12
_
Next we use y
3
to find a vector orthogonal to u
2
.
z
3
= y
3
pruj
u
2
y
3
= y
3
(y
3
. u
2
)u
2
= _
S
S
1
1
_ __
S
S
1
1
_ _
12
12
12
12
___
12
12
12
12
_
= _
S
S
1
1
_ 2_
12
12
12
12
_ = _
2
2
2
2
_
Finally, we need to compute the norm of z
3
:
z
3
= 4 +4 +4 +4 = 16 = 4
Thus the final vector in the orthonormal basis is;

u
3
=
z
3
z
3
= _
12
12
12
12
_
Thus the orthonormal basis is;
__
12
12
12
12
_ , _
12
12
12
12
_ , _
12
12
12
12
__
Solution-part(iii): QR factorization:
S = Spon __
1
1
1
1
_ , _
1
4
4
1
_ , _
4
2
2
u
__
Q = _
12 12 12
12 12 12
12
12
12
12
12
12
_ _
u.S u.S u.S
u.S u.S u.S
u.S
u.S
u.S
u.S
u.S
u.S
_
R = Q
t
A = _
12 12 12 12
12 12 12 12
12 12 12 12
_ _
1 1
1 4
1 4
4
2
2
1 1 u
_
R = _
2 S 2
u S 2
u u 4
_
Solution-part(iv):QR factorization by MATLAB:
|Q, R] = qr(A) = _
2 S 2
u S 2
u u 4
_
Gives the same result just the signs are opposite with all the elements. R factorization of an m n
matrix A as the product of an m m matrix Q and m n matrix R = |r
|j
], where r
|j
= if | > j.
Question # 03 (Problem 3.8 on page 80 from Chens Book)
Find the general solution of
_
1 2
1

3 4
2 2
1
_ x = _
3
2
1
_
How many parameters do you have?
Solution:
Let

A = _
1 2
u 1
u u
S 4
2 2
u 1
_ onJ y = _
S
2
1
_
Also let
A = _
1 2
u 1
u u
S 4
2 2
u 1
_ = |a
1
a
2
a
3
a
4
]
Since 2(a
2
) a
3
= a
1
so a
2
, a
3
and a
4
are linearly independent and rank(A) = 3.
Here : y = 2(a
2
) a
3
a
4

i.e; rank(|A, y]) = 3
So this y lies in the range space of A and a solution x exists in Ax = y.
Now we will find the solution, x as below by using Gaussian Elimination Method.
The augmented form is:
_
1 2 S
u 1 2
u u u
4 S
2 2
1 1
_ = _
1 2 S
u 1 2
u u u
4 S
2 2
1 1
_ , (1)R
2

= _
1 2 S
u 1 2
u u u
4 S
2 2
1 1
_ , R
1
2R
2

Now writing the linear combination gives the following equations;
x
1
x
3
+8x
4
= 7, x
2
+ 2x
3
2x
4
= 2, x
4
= 1
By backward substitution and letting x
3
= u , the values are found to be;
x
4
= 1, x
3
= u, x
2
= u, x
1
= 1
So
x
P
= |1, , , 1]
t

is the required solution.
As we know that
Nu|||ty (A) = n R(A) = 4 3 = 1
Which means the dimension of the null space of A is 1, the number of parameters in the general solution
will be 1.
Now we will find basis of the null space of A, for this let;
u = _
1 2
u 1
u u
S 4
2 2
u 1
_
The Reduced Echelon form is found to be;

= u = _
1 u
u 1
u u
1 8
2 2
u 1
_ , (1)R
2
& R
1
2R
2

= u = _
1 u
u 1
u u
1 u
2 u
u 1
_ , R
1
8R
3
& R
2
+2R
3

Here x
3
is a free variable and x
1
, x
2
onJ x
4
are basis variable.
By solving ux = with x
3
= 1;
_
1 u
u 1
u u
1 u
2 u
u 1
_ _
x
1
x
2
x
3
x
4
_ = _
u
u
u
_
x
1
x
3
= u , x
2
+ 2x
3
= u onJ x
4
= u
We get ;
x
4
= u, x
3
= 1, x
2
= 2 onJ x
1
= 1
So basis of the null space of A is;
F
1
= |1 2 1 ]
t

Thus the general soution of Ax = y can be expressed as;
x = x
P
+ uF
1
= _
1
1
_ + u_
1
2
1
_ , or ony rcol u
u is the only parameter.
Find the solution in the following that has the smallest Euclidean norm.
_
1 2
1

3 4
2 2
1
_ x = _
3
2
1
_
Solution:
x = x
P
+ uF
1
= _
1
1
_ +u_
1
2
1
_ , or ony rcol u (I)

Now nurm of x is;
x = (o 1)
2
+(2o)
2
+o
2
+1 = 6o
2
2o +2
The EucliJcon norm of x is;

0x
0u
=
= 12u 2 = = u =
1

Substituting u =
1
in (I) yields;
x = _
5
13
1
1
_
Has the smallest EucliJcon norm.
Question # 05
Let A
nn
have n distinct eigenvalues 2
|
with associated eigenvectors F
|
. Show that {F
|
]
|=1
n
is linearly
independent.
Solution-Proof: Suppose that that {F
|
]
|=1
n
is linearly independent. By reordering the F
|
s if necessary,
there exists some k such that 1 k n and coefficients o
1
, o
2
, o
k
such that;
= o
k
=1
(I)
And o
u i. Let z
be the eigenvalue associated to F

|
for each i. Then
= (A z
k
I)u
= (A z
k
I) o
F
|
k
=1
, rom (I)
= o
(A z
k
I)F
|
k
=1
, using Sum or k :olucs
= o
(AI
z
k
IF
|
)
k
=1
, rigt multiplicotion o F
|

= o
(AF
|
z
k
F
|
)
k
=1
, multiplicoti:c iJcntity
= o
(z
F
|
z
k
F
|
)
k
=1
, AF
|
= 2
|
F
|

= o
(z
z
k
)
k
=1
, toking F
|
common rom tc rigt

= o
(z
z
k
)
k
=1
, sincc kt tcrm gi:cs u solution crc
For all i < k, z
z
k
u, since the F
|
s were pre-asumed to have distinct eigenvalues. Therefore
o
(z
z
k
) i < k
Now we have obtained a linear dependence relation
= o
(z
z
k
)F
|
k-1
=1

nvolving only the vectorsF
1
, F
2
, , F
k-1
where all the coefficients are nonzero. We can repeat this
process (the next step is to multiply the matrix A 2
k-1
I, etc.) until we arrive at an equation.
= o. F
1
, or somc o u
This implies F
1
= , which is a contradiction. Therefore F
1
, F
2
, , F
n
are linearly independent.
Question # 06
Let A
nn
= uu
T
where u and u are n hy 1 column vectors. Determine the followings:
(a) Eigenvalues of A
(b) Eigenvectors of A
(c) Rank of A
Solution (a & b)-Eigenvalues and Eigenvectors of A: Let
u = _
u
11
u
21
u
n1
_ onJ u = _
:
11
:
21
:
n1
_
A
nn
= uu
T
= __
u
11
u
21
u
n1
_ _
:
11
:
21
:
n1
_
t
_ = __
u
11
u
21
u
n1
_ |
:
11
:
21 :
n1
]_ = _
u
11
:
11
u
11
:
21 u
11
:
n1
u
21
:
11
u
21
:
21

u
21
:
n1
u
n1
:
11
u
n1
:
21 u
n1
:
n1
_
Let A
1,1
= u
11
:
11
, A
1,2
= u
11
:
21
etc.
A
nn
=
A
1,1
A
1,2
A
1,n
A
2,1
A
2,2

A
2,n
A
n,1
A
n,2
A
n,n

If we multiply A with n rows and n columns by a vector X, the result will be another vector w = AX, aslo
with n rows and one column. That is;

_
x
1
x
2
x
n
_ is mapped to _
w
1
w
2
w
n
_ =
A
1,1
A
1,2
A
1,n
A
2,1
A
2,2

A
2,n
A
n,1
A
n,2
A
n,n
_
x
1
x
2
x
n
_

Where for each index i;
w
|
= A
|,1
x
1
+A
|,2
x
1
++A
|,n
x
n
= A
|,j
X
j
n
j=1

In general if X
j
are not all zero, the vectors X and AX will not be parallel. When they are parallel (i.e;
when there is some real number z such that AX = 2X) we say that X is an eigenvector of A. In that case,
the scale factor z is said to be the eigenvalue corresponding to that eigenvector.
Solution Part(c): Rank of matrix A:
Since here A is a n by n matrix. So we can find its rank by the statement below.
If A is an n n matrix, then rank(A) = n iff A is row equivalent to I
n
.
For the proof of this statement,
If rank(A) = n then A is row equivalent to a matrix B in reduced row echelon form, and rank(B) =
n.
Since rank(B) = n , we conclude that B has no zero rows. For this, as we know that;
If A be an n n matrix in reduced row echelon form and A I
n
,then A has a row consisting entirely
of zeros.
This statement implies that B = I
n
.
Hence A is row equivalent to I
n
.
Conversely, if A is row equivalent to I
n
, then
rank(A) = rank(I
n
) = n
Question # 07
Let A
nn
be a Hermitian-symmetric matrix (i.e. A
T
= A where represents complex conjugate
operation). Prove the following statements.
(a) All eigenvalues of matrix A are real.
(b) Eigenvectors corresponding to different eigenvalues are orthogonal. Then we can normalize the
length of all eigenvectors to unity in order to construct an orthonormal basis {F
|
]
|=1
n
.
(c) The model matrix of A, namely F = |u
1
, u
2
, , u
n
], is unitary (t. e. F
-1
= F
T
).
Solution-Proof (a) Let A be a Hermitian Matrix. Then by definition;
(A
)
t
= A (|)

Let z be an eigenvalue of A.
Let u be an eigenvector corresponding to the eigenvalue z.
By definition of eigenvector;
Au = 2u
Pre-multiplying both sides by (u)
t
,
(u)
t
Au = (u)
t
2u
= (u)
t
Au = 2(u)
t
u (ii)
Firstly note that both (u)
t
Au and (u)
t
u are 1 1 matrices.
Now from (ii) , consider Left side and taking conjugate transpose, we get;
((u)
t
Au
)
t
= (u)
t
(A
)
t
((u)
t
)
t
(iii)
As A is Hermitian, and ((u)
t
)
t
= u, by double conjugate transpose is itself, it follows that;
(u)
t
(A
)
t
((u)
t
)
t
= (u)
t
Au
Substituting this result in (iii), we get;
((u)
t
Au
)
t
= (u)
t
Au
That is (u)
t
Au is also Hermitian. By product with conjugate transpose Matrix is Hermitian, (u)
t
u is
Hermitian.
So both (u)
t
Au and (u)
t
u are Hermitian1 1 matrices. Now suppose that we have for some o, b C
(u)
t
Au = |a]
(u)
t
u = |h]
Note that b u as a definition: eigenvector is non-zero.
By definition of a Hermtian matrix:
o = o onJ h = b

Where o denotes the complex conjugate of o.
By complex number equals conjugate iff wholly real, it follows that o, b H , that is, are real.
From equation (ii),
|o] = z|b]
= o = zb
= z =
o
b
, b u

Hence z , being a quotient of real numbers, is real.
Solution-Proof (b) Suppose p is an eigenvalue of A with eigenvector u and z is an eigenvalue of A with
eigenvector :. Then
(u. u) = (u). u ossocioti:c propcrty
= (Au). u , sincc Au = u
= u. (A
)
t
u, sincc A = (A
)
t

= u. Au
= u. (2u), sincc Au = 2u
= 2
(u. u)
= 2(u. u), sincc 2
= 2
u. u = = 2,
So if the eigenvalues are different then u. u = .
Hence proved..!
Suppose u
1
is an eigenvector corresponding to the eigenvalue z
1
of A. Then
Au
1
= 2
1
u
1

So
Au
1
= 2
1
u
1
= |2
1
|u
1
(i) (using Norm Axioms)
But A preserves lengths because A is here constructing orthonormal set, so

Au
1
= u
1
(ii)
Substituting (ii) in (i), we get;
u
1
= |2
1
|u
1

1 = |2
1
|
Hence proved..!
Solution-Proof (c):
Here
I = | u
1
u
2
u
n]
Suppose that
Fu
1
Fu
2
= u
1
u
2
or c:cry u
1
, u
2
orc Herm|t|an

Then
Iu
1
u
2
= u
1
u
2
= Fu
1
Fu
2
= (u
2
)
T
(F
)
T
Fu
1
= (F
)
T
Fu
1
u
2

By using if u
1
, u
2
are Hermitian then their dot product is given by;
u
1
u
2
= (u
2
)
T
u
1

So that;
((F
)
T
F I)u
1
u
2
=
In particular, this holds when;
u
2
= ((F
)
T
F I)u
1

So that;
((F
)
T
F I)u
1
((F
)
T
F I)u
1
=
When;
((F
)
T
F I)u
1
=
So,
((F
)
T
F I) =
= (F
)
T
F = I
(F
)
T
= F
-1

Hence proved..!
Question # 08 (Problem 3.13 on page 81 Chens Book)
Find Jordan-form representations of the following matrices:
A
1
= _
1 4 1
2
3
_ , A
2
= _
1
1
2 4 3
_ , A
3
= _
1 1
1
2
_ , A
4
= _
4 3
2 1
25 2
_
Note that all except A
4
can be diagonalized.
Solution:
(a) ]urdan Furm Reprexentat|un u A
1
:
A
1
= _
1 4 1u
u 2 u
u u S
_
The characteristic polynomial of A
1
is ;

1
(z) = Jct(zI A
1
) = _
z 1 4 1u
u z 2 u
u u z S
_ = (z 1)(z 2)(z S)
Since
1
(z) is a diagonal matrix so the eigenvalues of A
1
are 1, 2, S. They all are distinct so the [orJon
Form rcprcscntotion of A
1
will be diagonal.
The cigcn:cctor associated wtih z = 1 is any non-zero solution of
(A
1
1I)q
1
= _
u 4 1u
u 1 u
u u 2
_ q
1
= u
= _
u 4 1u
u 1 u
u u 2
_ _
q
11
q
12
q
13
_ = _
u
u
u
_
Here we set q
11
= c, now the above system can be written as;
4q
12
+1uq
13
= u, q
12
= u, 2q
13
= u = q
13
= u, q
12
= u
Hence
q
1
= (
q
11
q
12
q
13)
t
= (c u u)
t
= c(1 u u)
t

Where c is any arbitrary number. We may also write as follows;
q
1
= (1 u u)
t

(A
1
2I)q
2
= _
1 4 1u
u u u
u u 1
_ q
2
= u
= _
1 4 1u
u u u
u u 1
_ _
q
21
q
22
q
23
_ = _
u
u
u
_
Here we set q
22
q
21
+4q
22
+ 1uq
23
= u, 2q
23
= u
= q
23
= u, q
22
= c, q
21
= 4c
Hence
q
2
= (
q
21
q
22
q
23)
t
= (4c c u)
t
= c(4 1 u)
t

q
2
= (4 1 u)
t

The cigcn:cctor associated wtih z = S is any non-zero solution of
(A
1
SI)q
3
= _
2 4 1u
u 1 u
u u u
_ q
3
= u

= _
2 4 1u
u 1 u
u u u
_ _
q
31
q
32
q
33
_ = _
u
u
u
_
Here we set q
33
2q
31
+ 4q
32
+ 1uq
33
= u, q
32
= u
= q
33
= c, q
32
= u, q
31
= Sc
Hence
q
3
= (
q
31
q
32
q
33)
t
= (Sc u c)
t
= c(S u 1)
t

q
3
= (S u 1)
t

Now we have;
Q = |
q
1
q
2
q
3
] = _
1 4 S
u 1 u
u u 1
_
|| = _
1 4 S
u 1 u
u u 1
_ = 1
1 u
u 1
= 1
-1
=
AJ]()
||
=
1
1

1 u
u 1

u u
u 1

u 1
u u
4 S
u 1

1 S
u 1

1 4
u u
4 S
1 u

1 S
u u

1 4
u 1
t
= _
1 u u
4 1 u
S u 1
_
t
= _
1 4 S
u 1 u
u u 1
_
Thus the [orJon Form rcprcscntotion of A
1
with respect to {
q
1
q
2
q
3
] is;
A
`
1
=
-1
A
1
= _
1 4 S
u 1 u
u u 1
_ _
1 4 1u
u 2 u
u u S
_ _
1 4 S
u 1 u
u u 1
_ = _
1 4 S
u 1 u
u u 1
_ _
1 8 1S
u 2 u
u u S
_
= A
1
= Q
-1
A
1
Q = _
1
2
3
_

(b) ]urdan Furm Reprexentat|un u A
2
:
A
2
= _
u 1 u
u u 1
2 4 S
_
2
is ;
2
(z) = Jct(zI A
2
) = _
z 1 u
u z 1
2 4 z +S
_ = z
z 1
4 z +S
+
u 1
2 z +S
= z(z
2
+Sz +4) +(2)

=
2
(z) = Jct(zI A
2
) = z
3
+Sz
2
+4z +2 = (z +1)(z
2
+ 2z +2)
=
2
(z) = Jct(zI A
2
) = (z +1)(z
2
+2z +1 +1) = (z +1)|(z + 1)
2
(1)]
=
2
(z) = Jct(zI A
2
) = (z +1)|(z +1)
2
]
2
] = (z + 1)(z + 1 ])(z + 1 +])
Thus the eigenvalues of A
2
are 1, 1 +], 1 ]. They all are distinct so the [orJon
Form rcprcscntotion of A
2
will be diagonal.
The cigcn:cctor associated wtih z = 1 is any non-zero solution of;
(A
2
+1I)q
1
= _
1 1 u
u 1 1
2 4 2
_ q
1
= u
= _
1 1 u
u 1 1
2 4 2
_ _
q
11
q
12
q
13
_ = _
u
u
u
_
Here we set q
13
q
11
+q
12
= u, q
12
+q
13
= u, 2q
11
4q
12
2q
13
= u = q
13
= c, q
12
= c, q
11
= c
Hence
q
1
= (
q
11
q
12
q
13)
t
= (c c c)
t
= c(1 1 1)
t

q
1
= (1 1 1)
t

The cigcn:cctor associated wtih z = 1 +] is any non-zero solution of;
(A
2
(1 + ])I)q
2
= _
1 ] 1 u
u 1 ] 1
2 4 2 ]
_ q
2
= u
= _
1 ] 1 u
u 1 ] 1
2 4 2 ]
_ _
q
21
q
22
q
23
_ = _
u
u
u
_
Here we set q
21
(1 ])q
21
+ q
22
= u; (1 ])q
22
+q
23
= u; 2q
21
4q
22
+ (2 ])q
23
= u
= q
21
= c, q
22
= (1 +])c, q
23
= 2]c,
Hence
q
2
= (
q
21
q
22
q
23)
t
= (c (1 +])c 2]c)
t
= c(1 1 +] 2])
t

q
2
= (1 1 +j 2j)
t

The cigcn:cctor associated wtih z = 1 ] is any non-zero solution of

(A
2
(1 ])I)q
3
= _
1 + ] 1 u
u 1 +] 1
2 4 2 +]
_ q
3
= u
= _
1 +] 1 u
u 1 + ] 1
2 4 2 +]
_ _
q
31
q
32
q
33
_ = _
u
u
u
_
Here we set q
31
= c , now the above system can be written as;
(1 + ])q
31
+q
32
= u, (1 +])q
32
+ q
33
= u, 2q
31
4q
32
+(2 + ])q
33
= u
= q
31
= c , q
32
= (1 ])c, q
33
= 2]c
Hence
q
3
= (
q
31
q
32
q
33)
t
= (c (1 ])c 2]c)
t
= c(1 1 ] 2])
t

q
3
= (1 1 j 2j)
t

Now we have;
Q = |
q
1
q
2
q
3
] = _
1 1 1
1 1 + j 1 j
1 2j 2j
_
|| = _
1 1 1
1 1 +] 1 ]
1 2] 2]
_ = 1_
1 +] 1 ]
2] 2]
_ +1_
1 1
2] 2]
_ + 1_
1 1
1 +] 1 ]
_ = 2]
-1
=
AJ]()
||
=
1
2]
_
1 + ] 1 ]
2] 2]
_ _
1 1 ]
1 2]
_ _
1 1 + ]
1 2]
_
_
1 1
2] 2]
_ _
1 1
1 2]
_ _
1 1
1 2]
_
_
1 1
1 + ] 1 ]
_ _
1 1
1 1 ]
_ _
1 1
1 1 + ]
_
t

=
1
2]
_
2] 2 2] +2 2] 1 ] 2] +1 ]
2] 2] 2] 1 2] +1
1 ] +1 ] 1 + ] 1 1 +] +1
_
t
=
1
2]
_
4] 1 + ] 1 +]
4] 2] 1 1 +2]
2] ] ]
_
t

=
-1
=
AJ]()
||
=
1
2]
_
4] 4] 2]
1 + ] 2] 1 ]
1 + ] 1 +2] ]
_ =
2 2 1
1
2
]
2
1
]
2

1
2
1
2
+
]
2
1 +
]
2

1
2

1
with respect to {
q
1
q
2
q
3] is;

A
`
2
=
-1
A
2
=
2 2 1
1
2
]
2
1
]
2

1
2
1
2
+
]
2
1 +
]
2

1
2
_
u 1 u
u u 1
2 4 S
_ _
1 1 1
1 1 +] 1 ]
1 2] 2]
_
= A
`
2
=
-1
A
2
=
2 2 1
1
2
]
2
1
]
2

1
2
1
2
+
]
2
1 +
]
2

1
2
_
1 1 +] 1 ]
1 2] 2]
1 2 + 2] 2 2]
_
= A
2
= Q
-1
A
2
Q = _
1
1 +j
1 j
_
(c) ]urdan Furm Reprexentat|un u A
3
:
A
3
= _
1 u 1
u 1 u
u u 2
_
3
is ;
3
(z) = Jct(zI A
3
) = _
z 1 u 1
u z 1 u
u u z 2
_ = (z 1)(z 1)(z 2) = (z 1)
2
(z 2)
Since
3
(z) is a diagonal matrix so the eigenvalues of A
3
are 1, 1, 2. The eigenvalue 1 has multiplicity 2,
and nullity (A
3
I) = S ronk(A
3
I) = S 1 = 2 .
The A
3
has two linearly independent cigcn:cctors associated wtih z = 1.
(A
3
1I)q
1
= _
u u 1
u u u
u u 1
_ q
1
= u
= _
u u 1
u u u
u u 1
_ _
q
11
q
12
q
13
_ = _
u
u
u
_
So,
q
1
= (1 )
t

Similarly,
q
2
= ( 1 )
t

(A
3
2I)q
3
= _
1 u 1
u 1 u
u u u
_ q
3
= u

= _
1 u 1
u 1 u
u u u
_ _
q
31
q
32
q
33
_ = _
u
u
u
_
So,
q
3
= (1 1)
t

Now we have;
Q = |
q
1
q
2
q
3
] = _
1 1
1
1
_
|| = _
1 u 1
u 1 u
u u 1
_ = 1
1 u
u 1
= 1
-1
=
AJ]()
||
=
1
1

1 u
u 1

u u
u 1

u 1
u u
u 1
u 1

1 1
u 1

1 u
u u
u 1
1 u

1 1
u u

1 u
u 1
t
= _
1 u u
u 1 u
1 u 1
_
t
= _
1 u 1
u 1 u
u u 1
_
1
with respect to {
q
1
q
2
q
3
] is;
A
`
3
=
-1
A
3
= _
1 u 1
u 1 u
u u 1
_ _
1 u 1
u 1 u
u u 2
_ _
1 u 1
u 1 u
u u 1
_ = _
1 u 1
u 1 u
u u 1
_ _
1 u 2
u 1 u
u u 2
_
= A
3
= Q
-1
A
3
Q = _
1
1
2
_
(d) ]urdan Furm Reprexentat|un u A
4
:
A
4
= _
u 4 S
u 2u 16
u 2S 2u
_
4
is ;
4
(z) = Jct(zI A
4
) = _
z 4 S
u z 2u 16
u 2S z +2u
_ = z
3

Clearly A
4
has only one disctinct eigenvalue u with multiplicity 3,
Nullity (A
4
uI) = S 2 = 1
Thus A
4
has only one independent eigenvector associated with u ;
A
4
F
1
=
= _
u 4 S
u 2u 16
u 2S 2u
_ _
I
11
I
12
I
13
_ = _
u
u
u
_

Here we set I
11
= 1, I
13
= u now the above system can be written as;
4I
12
+SI
13
= u = I
12
= u
Hence
F
1
= (F
11
F
12
F
13
)
t
= (1 )
t

We can compute generalized eigenvectors of A
4
from equations below;
A
4
F
2
= F
1

= _
u 4 S
u 2u 16
u 2S 2u
_ _
I
21
I
22
I
23
_ = _
1
u
u
_
4I
22
+ SI
23
= 1, 2uI
22
+ 16I
23
= u
Solving both of the above equations gives;
F
2
= ( 4 5)
t

Now,
A
4
F
3
= F
2

= _
u 4 S
u 2u 16
u 2S 2u
_ _
I
31
I
32
I
33
_ = _
u
4
S
_
4I
22
+ SI
23
= u, 2uI
22
+ 16I
23
= 4
Solving both of the above equations gives;
F
3
= ( 3 4)
t

Now we have;
Q = |F
1
F
2
F
3
] = _
1
4 3
5 4
_
|| = _
1 u u
u 4 S
u S 4
_ = 1
4 S
S 4
= 1
-1
=
AJ]()
||
=
1
1
4 S
S 4

u S
u 4

u 4
u S
u u
S 4

1 u
u 4

1 u
u S
u u
4 S

1 u
u S

1 u
u 4
t
= _
1 u u
u 4 S
u S 4
_
t
= _
1 u u
u 4 S
u S 4
_
Thus the represcntotion of A
4
with respect to the basis {F
1
F
2
F
3
] is ;

A
`
4
=
-1
A
4
= _
1 u u
u 4 S
u S 4
_ _
u 4 S
u 2u 16
u 2S 2u
_ _
1 u u
u 4 S
u S 4
_ = _
1 u u
u 4 S
u S 4
_ _
u 1 u
u u 4
u u S
_
= A
4
= Q
-1
A
4
Q = _
1
1

_
Given
A = _
1 1 u
u u 1
u u 1
_
Find A
1
, A
13
and e
At
.
Solution:
The characteristic polynomial of A is;
(2) = det(2I A) = _
2 1 1
2 1
2 1
_ = (2 1)
2 1
2 1

= (2) = det(2I A) = (2 1)(2)(2 1) = (2 1)(2
2
2) = 2(2 1)
2

Let
(z) = [
0
+ [
1
z +[
2
z
2

on the spectrum of A, we have
(i) Computation of A
1
:
(z) = (z)
(u) = (u) , (u)
10
= [
0
= [
0
= u
(1) = (1) , (1)
10
= [
0
+[
1
+[
2
= [
1
+[
2
= 1
i(1)
=
i(1)
, 1u(1)
10
= [
1
+2[
2
= [
1
+2[
2
= 1u
Solving above equations we have;
[
0
= u, [
1
= 8 onJ [
2
= 9
A
1
= 8A +9A
2
= 8 _
1 1 u
u u 1
u u 1
_ + 9_
1 1 1
u u 1
u u 1
_ = _
1 1 9
1
1
_
(ii) Computation of A
13
:
(z) = (z)
(u) = (u) , (u)
103
= [
0
= [
0
= u

(1) = (1) , (1)
103
= [
0
+ [
1
+[
2
= [
1
+[
2
= 1
i(1)
=
i(1)
, 1uS(1)
102
= [
1
+2[
2
= [
1
+ 2[
2
= 1uS
[
0
= u, [
1
= 1u1 onJ [
2
= 1u2
A
13
= 1u1A +1u2A
2
= 1u1_
1 1 u
u u 1
u u 1
_ +1u2 _
1 1 1
u u 1
u u 1
_ = _
1 1 12
1
1
_

(iii) Computation of e
At
:
(z) = (z)
(u) = (u) , e
= [
0
= [
0
= 1
(1) = (1) , e
t
= [
0
+ [
1
+[
2
= [
1
+[
2
= e
t
1
i(1)
=
i(1)
, te
t
= [
1
+ 2[
2
= [
1
+ 2[
2
= te
t

[
0
= 1, [
1
= 2e
t
te
t
2 onJ [
2
= te
t
e
t
+ 1
e
At
= [
0
I +[
1
A +[
2
A
2

= e
At
= _
1 u u
u 1 u
u u 1
_ +(2e
t
te
t
2 ) _
1 1 u
u u 1
u u 1
_ +(te
t
e
t
+1 ) _
1 1 1
u u 1
u u 1
_
= e
At
= _
e
t
e
t
1 te
t
e
t
+1
1 e
t
1
e
t
_
Question # 10
(a) Show that L
p
norm of a vector x, x
p
= | |x
|
|
p n
|=1
]
1p
satisfies all the axioms of a norm for
p 1. (Hint: Use Minkowski Inequality when it is necessary)
(b) Show that L
p
norm of a vector x becomes x
= max
|=1,,n
|x
|
|when we set p = in the above
expression.
(c) Determine the L
1
, L
2
, Fruhen|uux and L
norms of the following matrices:

A
1
= j
4 3
3
[ , A
2
= j
1 2
3
[ , A
3
= j
1 2
1
[
Solution:
(a) Show that L
p
norm of a vector x, x
p
= | |x
|
|
p n
|=1
]
1p
satisfies all the axioms of a norm for
p 1. (Hint: Use Minkowski Inequality when it is necessary)


Solution-Proof (a): A norm on a vector space F is a map . : F H such that for any x, y F and any
u K, where K is a field.
1. x , and x = x =
2. ux = |u|x
3. x +y x +y (Triangular Inequality)
Proof (a-1):
It is obvious as
p 1
so
x ur p 1 and x
p
=
_|x
|
|
p
n
|=1
_
1p
= ( uxed ur | and un|y |)
|x
|
|
p
n
|=1
=
x = , for each x
Proof (a-2):
ux
p
= _|ux
|
|
p
n
|=1
_
1p
= _u
p
|x
|
|
p
n
|=1
_
1p
= _u
p
|x
|
|
p
n
|=1
_
1p
= u_|x
|
|
p
n
|=1
_
1p
= ux
p

Proof (a-3):
The case p = 1 is obtained by taking summation on,
|x + y| |x| +|y|
If 1 < p < , define q from the equation
1
p
+
1
q
= 1.
Then using Holders inequality, we get
|x
j
+y
j
|
p
n
j=1
= |x
j
+ y
j
|
p-1
|x
j
+y
j
|
n
j=1

Using absolute property of inequality;
|x
j
+ y
j
|
p-1
|x
j
|
n
j=1
+ |x
j
+ y
j
|
p-1
|y
j
|
n
j=1

Using Holders inequality,

_|x
j
+y
j
|
(p-1)q
n
j=1
_
1q
_|x
j
|
p
n
j=1
_
1p
+_|x
j
+ y
j
|
(p-1)q
n
j=1
_
1q
_|y
j
|
p
n
j=1
_
1p

Dividing the inequality by [ |x
j
+y
j
|
(p-1)q
n
j=1

1q
and
using that (p 1)q = p and (1
1
q
=
1
p
), we get n
_|x
j
+y
j
|
p
n
j=1
_
1p
_|x
j
|
p
n
j=1
_
1
p
+ _|y
j
|
p
n
j=1
_
1p

i.e;
x +y
p
x
p
+y
p

Hence proved..!
(b) Show that L
p
norm of a vector x becomes x
= max
|=1,,n
|x
|
|when we set p = in the above
expression.
Solution-Proof (b): Since
|x
k
| _|x
j
|
p
j=1
_
1
p
= x
p

k N , p 1
We have
x
x
p

Thus in particular;
x
ltm
p
|nx
p
(|)
On the other had, we know that
x
p
= _|x
j
|
p-q
j=1
|x
j
|
q
_
1
p

x
p-q
p
_|x
j
|
q
j=1
_
1
p


x
1-
q
p
. x
q
q
p

q < p where we used |x
]
| x
p
] N.
Therefore we arrived at;
ltm
p
xupx
p
ltm
p
xup_x
1-
q
p
. x
q
q
p
_
= ltm
p
xupx
p
ltm
p
xup_x
1-
q
p
. x
q
q
p
_ = x
. 1 (||)
From (i) and (ii), we conclude that;
= ltm
p
xupx
p
x
ltm
p
|nx
p

= lim
p
x
p
exists and equals x

(c) Determine the L
1
, L
2
, Fruhen|uux and L
norms of the following matrices:

A
1
= j
4 3
3
[ , A
2
= j
1 2
3
[ , A
3
= j
1 2
1
[
Solution (c):
Formulas to Evaluae Norms are as under;
1) L
1
Nurm: A
1
= xupAx
1
= max
j
|a
|j
|
m
|=1
= max||argext cu|umn ahxu|ute Sum]
2) L
2
Nurm: A
2
= xup
A
2
=1
Ax
2
= 2
max
= the max|mum e|genua|ue u (A
t
A)
3) Fruhen|uux Nurm: A = _ |a
|j
|
2
j |

4) L
Nurm: A
= xupAx
= max
|
|a
|j
|
n
j=1
= max||argext Ruw Ahxu|ute Sum]
Solution - Computation of Norms (c-1):
A
1
= j
4 S
S 6
[
A
1
1
= max((|4| +|S|), (|S| +|6|)) = max(7,9) = 9
A
1
= max((|4| + |S|), (|S| +|6|)) = max(7,9) = 9

A
1
t
= j
4 S
S 6
[
A
1
t
A
1
= j
4 S
S 6
[ j
4 S
S 6
[ = j
2S Su
Su 4S
[
Eigcn:olucs = |zI A
1
t
A
1
| = j
z 2S Su
Su z 4S
[ = (z 2S)(z 4S) 9uu = z
2
7uz + 22S
Eigcn:olucs = |zI A
1
t
A
1
| = (z SS)
2
(1u1u)
2
= (z SS 1u1u)(z SS +1u1u)

Eigcn:olucs = (S.S772, 66.6228)
A
1
2
= z
mux
= tc moximum cigcn:oluc o (A
t
A) = 66.6228 = 8. 1 23
Fruhen|uux Norm:
_
|o
]
|
2
]
= 4
2
+ S
2
+S
2
+ 6
2
= 7u = 8. 3
A
2
= j
1 2
3
[,
A
2
1
= max((|1| +|S|), (|2| +|u|)) = max(4,2) = 4
A
2
= max((|1| +|2|), (|S| +|u|)) = max(S,S) = 3

A
2
t
= j
1 S
2 u
[
A
2
t
A
2
= j
1 S
2 u
[ j
1 2
S u
[ = j
1u 2
2 4
[
Eigcn:olucs = |zI A
2
t
A
2
| = j
z 1u 2
2 z 4
[ = (z 1u)(z 4) 4 = z
2
14z +S6
Eigcn:olucs = |zI A
2
t
A
2
| = (z 7)
2
(1S)
2
= (z 7 1S)(z 7 +1S)
Eigcn:olucs = (7 +1S, 7 1S) = (S.S944,1u.6uS6)
A
2
2
= z
mux
t
A) = 1u.6uS6 = 3. 25
Fruhen|uux Norm:
_
|o
]
|
2
]
= 1
2
+ 2
2
+S
2
+ u
2
= 14 = 3. 7 417
A
3
= j
1 2
u 1
[
A
3
1
= max((|1| +|u|), (|2| +|1|)) = max(1,S) = 3
A
3
= max((|1| +|2|), (|u| +|1|)) = max(S,1) = 3

A
3
t
= j
1 u
2 1
[
A
3
t
A
3
= j
1 u
2 1
[ j
1 2
u 1
[ = j
1 2
2 S
[
Eigcn:olucs = |zI A
3
t
A
3
| = j
z 1 2
2 z S
[ = (z 1)(z S) 4 = z
2
6z +1
Eigcn:olucs = |zI A
3
t
A
3
| = (z S)
2
(22)
2
= (z S 22)(z S + 22)

Eigcn:olucs = (S.8284, u.1716)
A
3
2
= z
mux
t
A) = S.8284 = 2. 4142
Fruhen|uux Norm:
_
|o
]
|
2
]
= 1
2
+ 2
2
+ u
2
+1
2
= 6 = 2. 4495

(HW2 LS EE571) Akhtar Rasool

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(HW2 LS EE571) Akhtar Rasool

Uploaded by

Copyright:

Available Formats

EE 571 Linear Systems Home Work 2 Page 1 / 28

EE 571 Linear Systems Spring 2014

_ , or ony rcol u (I)

be the eigenvalue associated to F

norms of the following matrices:

norms of the following matrices:

= max((|4| + |S|), (|S| +|6|)) = max(7,9) = 9

= max((|1| +|2|), (|S| +|u|)) = max(S,S) = 3

= max((|1| +|2|), (|u| +|1|)) = max(S,1) = 3

You might also like