MATH2822: Mathematical Methods for Actuarial Science II

2024-2025 Sem 2
Revision Class 3 Suggested Solution

STA: Xinyue YU (Christina), xinyue yu@connect.hku.hk, April 3, 2025

1. Let
1 2 3 2 1 0  111 333 789
    
𝐴 = 3 2 1 , 𝐵 = 1 0 −1 , 𝐶 = 333 999 666 .
     
     
1 1 1 0 −1 −2 321 963 753
     
(a) Prove that 𝐴 and 𝐵 are row equivalent.
(b) Prove that 𝐴 and 𝐶 are not row equivalent.

Suggested solution:

(a) We find the RREF of 𝐴 and 𝐵 as follows.

1 2 3  1 1 2 3  1 0 −1
 − 4 𝒓2 →𝒓2
−𝒓1 +𝒓3 →𝒓3
   
 𝒓2 +𝒓3 →𝒓3
𝐴 −−−−−−−−−→ 0 −4 −8 −−−−−−−→ 0 1 2  −−−−−−−−−→ 0 1 2 
   
−3𝒓1 +𝒓2 →𝒓2     −2𝒓2 +𝒓1 →𝒓1  
0 −1 −2 0 −1 −2 0 0 0 
     

1 0 −1 1 0 −1 1 0 −1


𝒓1 ↔𝒓2
 −2𝒓 +𝒓 →𝒓   𝒓 +𝒓 →𝒓  
1 2 2  2 3 3
𝐵 −−−−−→ 2 1 0  −−−−−−−−−→ 0 1 2  −−−−−−−→ 0 1 2 
    
     
0 −1 −2 0 −1 −2 0 0 0 
     
Since the RREF of 𝐴 and 𝐵 are the same, they are row equivalent.
(b) Let 𝒂 1 , 𝒂 2 , 𝒂 3 and 𝒄 1 , 𝒄 2 , 𝒄 3 be the columns of 𝐴 and 𝐶 respectively. Observe that 3𝒄 1 = 𝒄 2 .
If 𝐴 and 𝐶 are row equivalent, then 3𝒂 1 = 𝒂 2 by the column correspondence property.
However, this is not the case since 3𝒂 1 = (3, 9, 3) 𝑇 ≠ 𝒂 2 . Thus, 𝐴 and 𝐶 are not row
equivalent.

1
2. Find the value(s) of 𝑡 such that the system of linear equations




 𝑥 + 𝑦 + 𝑡𝑧 = 1,



𝑥 + 𝑡𝑦 + 𝑧 = 1,



 𝑡𝑥 + 𝑦 + 𝑧 = 1



(a) has no solution;

(b) has a unique solution;
(c) has exactly 3 solutions;
(d) has infinitely many solutions.

Suggested solution:
1 1 𝑡 1
 
We first transform the augmented matrix 𝐴 = 1 𝑡 1 1 using EROs as follows.
 
 
 𝑡 1 1 1
 

1 1 𝑡 1 1 1 𝑡 1 
  −𝑡𝒓 +𝒓 →𝒓 
1 3 3
1 𝑡 1 1 −−−−−−−−−→ 0 𝑡 − 1 1 − 𝑡 0  (1)
   
  −𝒓1 +𝒓2 →𝒓2  
 𝑡 1 1 1 0 1 − 𝑡 1 − 𝑡 2 1 − 𝑡 
   

If 𝑡 ≠ 1, we further apply EROs as follows.

1 1 𝑡 1  1 1 1 𝑡 1 1 0 𝑡 + 1 1
 1−𝑡 𝒓3 →𝒓3
   
 −𝒓2 +𝒓3 →𝒓3
0 𝑡 − 1 1 − 𝑡 0  −−1−−−−−→ 0 1 −1 0 −−−−−−−−→ 0 1 −1 0 (2)
   
 −𝒓2 +𝒓1 →𝒓1
0 1 − 𝑡 1 − 𝑡 2 1 − 𝑡  𝑡 −1 𝒓2 →𝒓2
    
0 1 1 + 𝑡 1 0 0 2 + 𝑡 1
     

(a) When 𝑡 = −2, the matrix in (2) has a leading one in the last column, there is no solution.
1 0 0 ∗
 
(b) When 𝑡 ≠ 1, −2, we can further reduce the matrix in (2) to the RREF 0 1 0 ∗ . There
 
 
0 0 1 ∗
 
is a unique solution.
(c) It is impossible to have exactly 3 solutions.
1 1 1 0
 
(d) When 𝑡 = 1, the matrix 𝐴 is row equivalent to 0 0 0 0 by (1). There are infinitely
 
 
0 0 0 0
 
many solutions.

3. Prove that we can use Type II and Type III EROs to interchange two rows. In other words, a Type
I elementary matrix can be represented as a product of Type II and Type III elementary matrices.

2
 Suggested solution:
Suppose we want to interchange row 𝑖 and row 𝑗. To have more observations on the entries, we
focus on (𝑖, 1)th entry 𝑎 and ( 𝑗, 1)th entry 𝑏.

• 𝒓 𝑖 + 𝒓 𝑗 → 𝒓 𝑗 : the entries become 𝑎 and 𝑎 + 𝑏

• −𝒓 𝑗 + 𝒓 𝑖 → 𝒓 𝑖 : the entries become −𝑏 and 𝑎 + 𝑏
• 𝒓 𝑖 + 𝒓 𝑗 → 𝒓 𝑗 : the entries become −𝑏 and 𝑎
• −𝒓 𝑖 → 𝒓 𝑖 : the entries become 𝑏 and 𝑎

The two entries are interchanged. As the same holds for other columns, we have interchanged
row 𝑖 and row 𝑗 using this sequence of operations.

4. Let 𝐴 be an 𝑚 × 𝑛 matrix, and let 𝒃 be an 𝑚 × 1 column vector. Suppose 𝐴𝒙 = 𝒃 is consistent,

and 𝒙 = 𝒑 is a particular solution. Prove that 𝒖 is a solution to 𝐴𝒙 = 𝒃 if and only if 𝒖 = 𝒗 + 𝒑
for some 𝒗 which is a solution to the homogeneous equation 𝐴𝒙 = 0.
Suggested solution:
”If” part: Since 𝒖 = 𝒗 + 𝒑 with 𝐴𝒗 = 0 and 𝐴 𝒑 = 𝒃, we have

𝐴𝒖 = 𝐴(𝒗 + 𝒑) = 𝐴𝒗 + 𝐴 𝒑 = 0 + 𝒃 = 𝒃.

This shows 𝒖 is a solution to 𝐴𝒙 = 𝒃.

”Only if” part: Since 𝐴𝒖 = 𝒃 and 𝐴 𝒑 = 𝒃, by defining 𝒗 = 𝒖 − 𝒑, we have 𝒖 = 𝒗 + 𝒑 and

𝐴𝒗 = 𝐴(𝒖 − 𝒑) = 𝐴𝒖 − 𝐴 𝒑 = 𝒃 − 𝒃 = 0

This shows 𝒗 is a solution to 𝐴𝒙 = 0.

−3 −2 2 
 
5. Using elementary row operations, find the inverse of 𝐴 =  4 1 −4 .
 
 
−2 3 5 


3
 Suggested solution:

 1 −1 −2 1 1 0
 2𝒓 +𝒓 →𝒓 1 −1 −2 1 1 0

𝒓2 +𝒓1 →𝒓1

1 3 3
( 𝐴|𝐼) −−−−−−−→  4 1 −4 0 1 0 −−−−−−−−−→ 0 5 4 −4 −3 0
   
  −4𝒓1 +𝒓2 →𝒓2  
−2 3 5 0 0 1 0 1 1 2 2 1
 
1 −1 −2 1 1 0 1 0 −1 3 3 1 

𝒓2 ↔𝒓3 

 −5𝒓2 +𝒓3 →𝒓3 
−−−−−→ 0 1 1 2 2 1 −−−−−−−−−→ 0 1 1 2 2 1

  𝒓1 +𝒓2 →𝒓1  
0 5
 4 −4 −3 0 

0 0 −1 −14 −13 −5
 
1 0 −1 3 3 1 1 0 0 17 16 6 
−𝒓3 →𝒓3 
  −𝒓 +𝒓 →𝒓 
 3 2 2 
−−−−−−→ 0 1 1 2 2 1 −−−−−−−−→ 0 1 0 −12 −11 −4

  𝒓1 +𝒓3 →𝒓1  
0 0 1 14 13 5 0 0 1 14 13 5 
  

 17 16 6 

Then 𝐴 −1 = −12 −11 −4 .
 
 
 14 13 5 

6. Determine whether each of the following is true or false and briefly explain the reasons.

(a) Let 𝐴, 𝐵," 𝐶 and#𝐷 be invertible matrices of the same size. Then the block matrix defined
𝐴 𝐵
by 𝑀 = is invertible.
𝐶 𝐷
(b) If 𝑆1 and 𝑆2 are finite subsets of R𝑛 such that 𝑆1 is contained in Span 𝑆2 , then Span 𝑆1 is
contained in Span 𝑆2 .
(c) Some systems of linear equations have exactly two solutions.
(d) If a system of linear equations has more variables than equations, then it must have infinitely
many solutions.
(e) If 𝑇 : R𝑛 → R𝑛 and 𝑈 : R𝑛 → R𝑛 are linear transformations such that 𝑇 ◦ 𝑈 is the identity
transformation, then 𝑈 is the inverse of 𝑇.
(f) If 𝑇 : R𝑛 → R𝑛 is an invertible linear transformation, then 𝑇 𝑘 is the identity transformation
for some positive integer 𝑘, where 𝑇 𝑘 = 𝑇 ◦ 𝑇 ◦ · · · ◦ 𝑇 .
| {z }
𝑘 times

Suggested solution:

(a) False." Consider

# the 1 × 1 matrices 𝐴 = 𝐵 = 𝐶 = 𝐷 = [1] which are invertible. But
1 1
𝑀= is non-invertible.
1 1
(b) True. If 𝑆1 is contained in Span 𝑆2 , then the vectors in 𝑆1 can be represented as linear
combinations of vectors in 𝑆2 . Span 𝑆1 is the linear combinations of vectors in 𝑆1 , which
can also be represented by linear combinations of vectors in 𝑆2 .

4
 (c) False. A linear system can only have no solution, unique solution, and infinitely many
solutions.
h i
(d) False. If the last row of the RREF of ( 𝐴|𝒃) is 0 · · · 0 1 , then the linear system has
no solution.
(e) True. Suppose 𝐴 and 𝐵 are standard matrix of 𝑇 and 𝑈 respectively. Then 𝑇 (𝒙) = 𝐴𝒙
and 𝑈 (𝒙) = 𝐵𝒙 for any 𝒙 ∈ R𝑛 . Given 𝑇 ◦ 𝑈 is the identity transformation, we have
𝑇 ◦ 𝑈 (𝒙) = 𝐴(𝐵𝒙) = 𝐴𝐵𝒙 = 𝒙 for any 𝒙 ∈ R𝑛 , then 𝐴𝐵 = 𝐼, 𝐵 = 𝐴 −1 . Then
𝑈 (𝒙) = 𝐵𝒙 = 𝐴 −1 𝒙 = 𝑇 −1 (𝒙), 𝑈 is the inverse of 𝑇. (Remark: From 𝐴𝐵 = 𝐼, we know
that 𝐴 and 𝐵 are invertible, so do 𝑇 and 𝑈, hence 𝑇 and 𝑈 have inverse.)
" #
𝑥+𝑦
(f) False. Consider an invertible linear transformation defined on R2
by 𝑇 (𝑥, 𝑦) = .
𝑦
" # " # " # " #
1 2 𝑘 0
Then 𝑇 (0, 1) = , 𝑇 2 (0, 1) = 𝑇 (1, 1) = , and inductively, 𝑇 𝑘 (0, 1) = ≠ for
1 1 1 1
any positive integer 𝑘. Therefore, 𝑇 𝑘 cannot be the identity transformation.

7. Let 𝐴 be an 𝑚 × 𝑛 matrix with 𝑚 > 𝑛, prove that there does not exist an 𝑛 × 𝑚 matrix 𝐵 such that
𝐴𝐵 = 𝐼𝑚 .
Suggested solution:
We prove by contradiction. If there exists 𝐵 such that 𝐴𝐵 = 𝐼𝑚 , then rank(𝐴𝐵) = 𝑚. Note
that rank( 𝐴) ≤ 𝑛 < 𝑚 and all columns of 𝐴𝐵 are linear combinations of columns of 𝐴, then
rank( 𝐴𝐵) ≤ rank( 𝐴) < 𝑚. We arrive at a contradiction, so such 𝐵 does not exist.
Remark: We can prove rank( 𝐴𝐵) ≤ rank( 𝐴) by consider the column space of 𝐴𝐵 is a subspace
of the column space 𝐴, so the dimension cannot be larger. Column space will be learned in
Chapter 7.

8. For an 𝑚 × 𝑛 matrix 𝐴, prove that rank 𝐴 = rank 𝐴𝑇 .

Suggested solution:
 𝒂𝑇 
 1 h i
Let 𝐴 =  · · ·  , then 𝐴𝑇 = 𝒂 1 · · · 𝒂 𝑚 . Let 𝑟 = rank( 𝐴𝑇 ), then there exist 𝑟 row vectors of
 
 
 𝒂𝑇 
 𝑚
𝐴 such that they are linearly independent. Without loss of generality, we assume that 𝒂 1 , · · · , 𝒂𝑟
are linearly independent. Next, we claim that 𝐴𝒂 1 , · · · , 𝐴𝒂𝑟 are also linearly independent.
Prove the claim: Suppose that there exist 𝑐 1 , · · · , 𝑐𝑟 such that 𝑐 1 𝐴𝒂 1 + · · · + 𝑐𝑟 𝐴𝒂𝑟 = 0. It is
equivalent to 𝐴(𝑐 1 𝒂 1 + · · · + 𝑐𝑟 𝒂𝑟 ) := 𝐴𝒗 = 0, where 𝒗 = 𝑐 1 𝒂 1 + · · · + 𝑐𝑟 𝒂𝑟 . It also implies
that 𝒂𝑇𝑖 𝒗 = 0 for all 𝑖 = 1, · · · , 𝑟. Then 𝒗 𝑇 𝒗 = 0 and thus 𝒗 = 0. Since 𝒂 1 , · · · , 𝒂𝑟 are linearly
independent, we have 𝑐 1 = · · · = 𝑐𝑟 = 0 and thus 𝐴𝒂 1 , · · · , 𝐴𝒂𝑟 are linearly independent.
Note that 𝐴𝒂 1 , · · · , 𝐴𝒂𝑟 are linear combinations of columns of 𝐴, we have rank( 𝐴) ≥ 𝑟 =
rank( 𝐴𝑇 ). We do a similar analysis to 𝐴𝑇 , we have rank( 𝐴𝑇 ) ≥ rank( 𝐴). Since the two
inequalities hold simultaneously, we arrive at rank( 𝐴) = rank( 𝐴𝑇 ).

5
9. Compute the determinant of following matrices:
2 1 0 1 

 
−2 −3 −1 −2
(a) 
4
 4 −8 0 
 
 3 −7 −2 4 

2
 9 −3 0
0
 1 −4 0 2 0
 
 
(b)  3 −2
 8 −4 5
−6 2 −7 3 0
 
 
0 20 0 0

1 1 1 1 1 

1 2 3 4 5
 
 
(c) 1 3 5
 7 9 
1 4 7 10 13
 
 
1 5 9 13 17

2 1
 1 · · · 1
 
1 2
 1 · · · 1
(d) 𝐴 = 1 1
 2 · · · 1 is an 𝑛 × 𝑛 matrix
. . .. . . .
 .. ... . .. 

 
1 1
 1 · · · 2
1
 𝑥1 𝑥 12 · · · 𝑥 1𝑛−1 
𝑥 22 · · ·
 
1 𝑥2 𝑥 2𝑛−1 

(e) Vandermonde matrix is defined by 𝑉 = 1
 𝑥3 𝑥 32 · · · 𝑥 3𝑛−1 
. .. .. . . .. 
 .. . . . . 

𝑥 𝑛 𝑥 𝑛2 · · ·
 
1 𝑛−1
𝑥 𝑛 

Suggested solution:

6
(a)

2 1 0 1 −2𝒄 4 + 𝒄 1 → 𝒄 1 0 0 0 1
2 −1 −1
−2 −3 −1 −2 −𝒄 4 + 𝒄 2 → 𝒄 2 2 −1 −1 −2
= =− 4 4 −8
4 4 −8 0 4 4 −8 0
−5 −11 −2
3 −7 −2 4 −5 −11 −2 4
2𝒄 2 + 𝒄 1 → 𝒄 1
0 −1 0
−𝒄 2 + 𝒄 3 → 𝒄 3 12 −12
= − 12 4 −12 = −
−27 9
−27 −11 9
1
12 𝒓 1 → 𝒓 1
1
9 𝒓2 → 𝒓2 1 −1
= −108 = 216.
−3 1

(b) Expanding along some column or row with some entries are 0, we have

2 9 0 −3 0
2 9 0 −3
1 −4 0 2 0 2 9 −3
1 −4 0 2 2 −3
3 −2 8 −4 5 = 5 = −35 1 −4 2 = 70 = 490.
−6 2 −7 3 1 2
−6 2 −7 3 0 0 2 0
0 2 0 0
0 2 0 0 0

(c) We have

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
𝒓1 +𝒓3 →𝒓3 −2𝒓2 +𝒓3 →𝒓3
1 3 5 7 9 = 2 4 6 8 10 = 0 0 0 0 0 .
1 4 7 10 13 1 4 7 10 13 1 4 7 10 13
1 5 9 13 17 1 5 9 13 17 1 5 9 13 17

Since there is a zero row, the determinant is 0.

(d) Adding each row 2, row 3, · · · , row 𝑛 to row 1, we obtain

𝑛+1 𝑛+1 𝑛+1 ··· 𝑛+1 1 1 1 ··· 1

1 2 1 ··· 1 1 2 1 ··· 1
det 𝐴 = 1 1 2 ··· 1 = (𝑛 + 1) 1 1 2 · · · 1 .
.. .. .. .. .. .. .. .. . . ..
. . . . . . . . . .
1 1 1 ··· 2 1 1 1 ··· 2

7
 Next, we subtract the first row from each of the other rows. This implies

1 1 1 ··· 1
0 1 0 ··· 0
det 𝐴 = (𝑛 + 1) 0 0 1 · · · 0 = 𝑛 + 1,
.. .. .. . . ..
. . . . .
0 0 0 ··· 1

since the last matrix is a triangular matrix with determinant of 1.

(e) For every column from the second column, subtract 𝑥 1 times the previous column.

1
 0 0 ··· 0 

1 𝑥 2 − 𝑥 1 𝑥 22 − 𝑥 1 𝑥 2
 
 ··· 𝑥2 − 𝑥 1 𝑥 2 
𝑛−1 𝑛−2

𝑉 → 1 𝑥3 − 𝑥 1 𝑥 32 − 𝑥 1 𝑥 3 ··· 𝑥 3𝑛−1 − 𝑥 1 𝑥 3𝑛−2  := 𝑉˜


. .. .. .. .. 
 .. . . . . 
 
1 𝑥 𝑛 − 𝑥 1 𝑥 𝑛2 − 𝑥 1 𝑥 𝑛
 
 ··· 𝑥 𝑛 − 𝑥 1 𝑥 𝑛 
𝑛−1 𝑛−2

then
det 𝑉 = det 𝑉˜ = det 𝑉,
¯

where
𝑥2 − 𝑥1 𝑥 2 − 𝑥1 𝑥2 · · · 𝑥 2𝑛−1 − 𝑥 1 𝑥 2𝑛−2 
 2
 𝑥 3 − 𝑥 1 𝑥32 − 𝑥 1 𝑥 3 · · · 𝑥 3𝑛−1 − 𝑥 1 𝑥 3𝑛−2 
 
𝑉¯ =  . .. .. .. .
 ..

. . . 
 
𝑥 𝑛 − 𝑥 1 𝑥 2 − 𝑥 1 𝑥 𝑛 · · · 𝑥 𝑛 − 𝑥 1 𝑥 𝑛 
𝑛−1 𝑛−2
 𝑛

Note that
𝑛
Ö
det 𝑉¯ = (𝑥 𝑖 − 𝑥 1 ) · det 𝑉𝑛−1 ,
𝑖=2

where
1 𝑥 2 𝑥 2 · · · 𝑥2𝑛−2 
 2
1 𝑥 3 𝑥 32 · · · 𝑥3𝑛−2 
 
𝑉𝑛−1 =  . . .. . . ..  .
 .. .. . . . 
 
1 𝑥 𝑛 𝑥 2 · · · 𝑛−2
𝑥 𝑛 
 𝑛

Similarly, we have
𝑛
Ö
det 𝑉𝑛−1 = (𝑥 𝑖 − 𝑥1 ) · det 𝑉𝑛−2 ,
𝑖=3

8
where
1 𝑥 3 𝑥 2 · · · 𝑥 3𝑛−3 
 3
1 𝑥 4 𝑥 42 · · · 𝑥 4𝑛−3 
 
𝑉𝑛−2 =  . . .. . . ..  .
 .. .. . . . 
 
1 𝑥 𝑛 𝑥 2 · · · 𝑛−3
𝑥 𝑛 
 𝑛

Moreover,
det 𝑉2 = (𝑥 𝑛 − 𝑥 𝑛−1 ) · det 𝑉1

and
det 𝑉1 = 1.

Therefore, we have Ö
det 𝑉 = (𝑥 𝑗 − 𝑥 𝑖 ).
𝑎≤𝑖< 𝑗 ≤𝑛