Solutions to the Final Exam

FM 5001 Preparation for Financial Mathematics

1. (a) Two quadtratic forms Q and Q0 are said to be equivalent if (assuming Q, Q0 :

Rn → R) ∃ an invertible linear transformation L : Rn → Rn such that Q0 = Q ◦ L.

(b) The image of a linear transformation T : Rn → Rk is the subset of Rk

{T (v)| v ∈ Rn } .

(c) A function f : X → Y is onto if f (X) = Y .

(d) Let Q be a positive semidefinite quadtratic form, and let B be its polarization.
Then the Cauchy-Schwartz inequality states that ∀v, ω ∈ Rn (assuming Q : Rn → R)
p p
|B(v, ω)| ≤ Q(v) Q(w).

(e) The dimension of a subspace S of Rn is the size of any basis of S, i.e. the
number of vectors in any basis of S.

(f ) A matrix K ∈ Rn×n is orthogonal if and only if K T K = KK T = I.

2. (a) True
(b) True
(c) False
(d) False
(e) False
(f ) False
(g) True
(h) True

3. (a) The number of subsets that have exactly 4 elements is

 
8 8! 8·7·6·5
= = = 70.
4 4!4! 4·3·2·1
(b)
   
2 7 0 3 −1 −2 6 −3
R ↔R3 −R1
 3 −7 2 7  −−1−−→  3 −7 2 7  −−→
−1 −2 6 −3 2 7 0 3
   
1 2 −6 3 1 2 −6 3 1
R3
−3R1 →R2
 3 −7 2 7 −−−−−−→ 0 −13
  20 −2  −3−→
−2R1 →R3
2 7 0 3 0 3 12 −3

1
    
1 2 −6 3 1 2 −6 3
R2 ↔R3 13R2 →R3
 0 −13 20 −2  − −−−→  0 1 4 −1  −−−− −−→
−2R2 →R1
0 1 4 −1 0 −13 20 −2
   
1 0 −14 5 1 1 0 −14 5
−4R1 →R2
 0 1 4 −1  −72
→ 0 1 4 −1  −−−− −−→
5 14R1 →R3
0 0 72 −15 0 0 1 − 24
35
   
1 0 0 6 35
− C →C4
1 0 0 0
 0 1 0 − 1  −−6−−1−−→  0 1 0 0 .
6 1
5 C2 →C4
0 0 1 − 24 6
5 0 0 1 0
24
C3 →C4

(c)

dim(ker(C)) = number of zero columns in the fully canonical form of C = 1.

dim(im(C)) = number of nonzero columns in the fully canonical form of C

= number of nonzero rows in the fully canonical form of C
= 3.

(d)  √
 √

 2 0  
2 1 √0 3 1
MMT =  1
√1
= .
0 1 2 1 3
0 2
The characteristic polynomial of M M T is
 
3−λ 1
det = (3 − λ)2 − 1
1 3−λ
= (3 − λ − 1)(3 − λ + 1)
= (2 − λ)(4 − λ).

Therefore, the eigenvalues of M M T are λ1 = 2 and λ2 = 4.

The 2-eigenvectors of M M T are the nonzero vectors in ker(M M T − 2I).
 
T 1 1
MM = ,
1 1

and     
1 1 a 0
= =⇒ a + b = 0 =⇒ b = −a.
1 1 b 0
Thus,   
T 1
ker(M M − 2I) = a a∈R .
−1

2
A unit basis vector of ker(M M T − 2I) is, therefore,
 
1 1
ω1 = √ .
2 −1

The 4-eigenvectors of M M T are the nonzero vectors in ker(M M T − 4I).

 
T −1 1
MM = ,
1 −1

and     
−1 1 a 0
= =⇒ a = b.
1 −1 b 0
Thus,   
T 1
ker(M M − 4I) = a a∈R .
1
A unit basis vector of ker(M M T − 4I) is, therefore,
 
1 1
ω2 = √ .
2 1
Let  
1 1 1
K = [ω1 ω2 ] = √ .
2 −1 1
K is a rotation and
  √   √ 
1 1 1 2 1 √0 1 2 1
KM = √ = .
2 −1 1 0 1 2 −1 0 1

Let us denote the rows of KM by r1 and r2 . We can easily check that r1 and r2 are
orthogonal.
 √ 
r1 · r2 = 1, 2, 1 · (−1, 0, 1) = −1 + 1 = 0 =⇒ r1 ⊥ r2 .

(e)
KM L = D =⇒ KM = DL−1 .
D is a matrix of the form  
d1 0 0
D= .
0 d2 0
Thus,

r1 = d1 × 1st row of L−1

r2 = d2 × 2nd row of L−1 .

3
L−1 is orthogonal, so its rows have length 1. Therefore, the above equations imply
that
√
d1 = |r1 | = 1 + 2 + 1 = 2
√ √
d2 = |r2 | = 1 + 1 = 2,

and  
2 √0 0
D= .
0 2 0
Also,
√ !
1 2 1
1st row of L−1 = , ,
2 2 2
 
nd −1 1 1
2 row of L = − √ , 0, √ .
2 2
√

Since we know that 1, − 2, 1 is orthogonal to the rows of L−1 and its length is 2,
we can unitize it and use it as the third row of L−1 . Thus,
 √ 
1 2 1
2 2 2
−1 √1
L−1 = √ 0 .
 
2 √ 2
1 2 1
2
− 2 2

Since L is orthogonal,
 1

− √12 1

T √2 2
√
L = L−1 = 2
0 − 2
.
 
2 2
1 √1 1
2 2 2

It is easily checked that

 
2 √0 0
KM L = D = .
0 2 0

(f )
D = A ⊕ B =⇒ D−1 = A−1 ⊕ B −1 .
       
3 2 −1 1 5 −2 1 5 −2 5 −2
A= =⇒ A = = = .
7 5 det A −7 3 1 −7 3 −7 3
       
1 2 −1 1 −3 −2 1 −3 −2 3 2
B= =⇒ B = = = .
−1 −3 det B 1 1 −1 1 1 −1 −1
Therefore,  
5 −2 0 0
 −7 3 0 0 
D−1 = A−1 ⊕ B −1 =
 0
.
0 3 2 
0 0 −1 −1

4
(g) A matrix is diagonalizable iff for any root of the characteristic polynomial, its
multiplicity is equal to the dimension of its eigenspace. An example of a nondiago-
nalizable 3 × 3 matrix is  
2 1 0
A =  0 2 1 .
0 0 2

(h)
 
2−λ 1 0  
2 − λ 1
det  0 2−λ 1  = (2 − λ) det = (2 − λ)3 .
0 2−λ
0 0 2−λ

The eigenvalue of A is λ = 2 and its multiplicity is 3. Let us now find the dimension
of the 2-eigenspace of A.  
0 1 0
A − 2I = 0 0
 1 ,
0 0 0
and     
0 1 0 a 0
 0 0 1   b  =  0  =⇒ b = 0 and c = 0.
0 0 0 c 0
Therefore,   
 1 
ker(A − 2I) =  0  a a ∈ R .
0
 

Clearly, the dimension of the 2-eigenspace is 1. The multiplicity of the eigenvalue 2 is

not equal to the dimension of the 2-eigenspace, verifying that A is not diagonalizable.

(i)  
3 2 0 0
 7 5 0 0 
D =A⊕B =
 0
.
0 1 2 
0 0 −1 −3
D is diagonalizable iff A and B are diaginalizable. Let us check whether A and B
are diagonalizable.
 
3−λ 2
det(A − λI) = det = (3 − λ)(5 − λ) − 14 = λ2 − 8λ + 1.
7 5−λ

Thus,
√ √
8− 64 − 4 √ 8 + 64 − 4 √
det(A−λI) = 0 =⇒ λ1 = = 4− 15 and λ2 = = 4+ 15.
2 2

5
Now, let us check the dimensions of the corresponding eigenspaces.
 √ 
−1 + 15 2√
A − λ1 I = ,
7 1 + 15

and
 √      √
−1 + 15 2√ a 0 a(−1 + 15)√ + 2b = 0
= =⇒
7 1 + 15 b 0 7a + (1 + 15)b = 0

From the first equation we have

√
1− 15
b= a,
2
and substituting in the second equation, we obtain
√ √
(1 + 15)(1 − 15)
7a + a = 0 =⇒ 7a − 7a = 0 =⇒ 0 · a = 0,
2
which is true for any a ∈ R. Thus,
  
1√
ker(A − λ1 I) = 1− 15 a a∈R ,
2

and the λ1 -eigenspace of A is 1-dimensional.

 √ 
−1 − 15 2√
A − λ2 I = ,
7 1 − 15

and
 √      √
−1 − 15 2√ a 0 a(−1 − 15)√ + 2b = 0
= =⇒
7 1 − 15 b 0 7a + (1 − 15)b = 0

From the first equation we have

√
1+ 15
b= a,
2
and substituting in the second equation, we obtain
√ √
(1 − 15)(1 + 15)
7a + a = 0 =⇒ 7a − 7a = 0 =⇒ 0 · a = 0,
2
which is true for any a ∈ R. Thus,
  
1√
ker(A − λ2 I) = 1+ 15 a a∈R ,
2

and the λ2 -eigenspace of A is 1-dimensional. Therefore, A is diagonalizable.

6
  
1−λ 2
det(B − λI) = det = (1 − λ)(−3 − λ) + 2 = λ2 + 2λ − 1.
−1 −3 − λ
Thus,
√ √
−2 − 4+4 √ −2 − 4+4 √
det(B−λI) = 0 =⇒ λ1 = = −1− 2 and λ2 = = −1+ 2.
2 2
Now, let us check the dimensions of the corresponding eigenspaces.
 √ 
2+ 2 2√
B − λ1 I = ,
−1 −2 + 2
and
 √      √
2+ 2 2√ a 0 a(2 + 2) + √
2b = 0
= =⇒
−1 −2 + 2 b 0 −a + (−2 + 2)b = 0
From the first equation we have
√
−2 − 2
b= a,
2
and substituting in the second equation, we obtain
√ √
(−2 + 2)(−2 − 2)
−a + a = 0 =⇒ −a + a = 0 =⇒ 0 · a = 0,
2
which is true for any a ∈ R. Thus,
  
1√
ker(B − λ1 I) = −2− 2 a a∈R ,
2

and the λ1 -eigenspace of B is 1-dimensional.

 √ 
2− 2 2√
B − λ2 I = ,
−1 −2 − 2
and
 √      √
2− 2 2√ a 0 a(2 − 2) + √
2b = 0
= =⇒
−1 −2 − 2 b 0 −a + (−2 − 2)b = 0
From the first equation we have
√
−2 + 2
b= a,
2
and substituting in the second equation, we obtain
√ √
(−2 − 2)(−2 + 2)
−a + a = 0 =⇒ −a + a = 0 =⇒ 0 · a = 0,
2
7
which is true for any a ∈ R. Thus,
  
1√
ker(B − λ2 I) = −2+ 2 a a∈R ,
2

and the λ2 -eigenspace of B is 1-dimensional. Therefore, B is diagonalizable.

Since both A and B are diagonalizable over the real numbers, we can conclude that
D is also diagonalizable over the real numbers.

(j) Let 

 −1, if 0 ≤ ω < 0.25
1, if 0.25 ≤ ω < 0.5

X(ω) =
 −1,
 if 0.5 ≤ ω < 0.75
1, if 0.75 ≤ ω ≤ 1


and 
−1, if 0 ≤ ω < 0.5
Y (ω) =
1, if 0.5 ≤ ω ≤ 1.
X and Y are piecewise constant and independent random variables.

(k)

Partition of X = P = {[0, 0.25), [0.25, 0.5), [0.5, 0.75), [0.75, 1]}

Partition of Y = Q = {[0, 0.5), [0.5, 1]}

It is easily seen that

E[X|P ] = 0
and
E[X|Q] = 0.