The BLAST Premier is a prestigious Counter Strike 2 (CS2) professional e-

sports league since 2020, primarily based in two regions: North America
and Europe. The World Final 2024 just happened in Resorts World Sentosa,
Singapore, from 30 October to 3 November. As a CS2 enthusiast, you want
to keep track of the number of fans of your favourite teams, Team Astralis
(AST), Team Vitality (VIT) and G2 Esports (G2). After every World Finals,
you note down the number of fans of the three teams and observe the
following:

- Fans of either of the three teams will keep being fans of either of the
three teams, and people not being fans of these three teams will
never become fans of these three teams.
- After every World Final, 6% of AST fans convert into VIT fans, and
4% convert into G2 fans. 3% of VIT fans convert into AST fans and
5% convert into G2 fans. 7% of G2 fans convert into AST fans and
2% convert into VIT fans.

Let the number of AST, VIT and G2 fans before World Finals 2020 be a 0, b 0
and c 0 respectively. The transition from a t, b t, c t to a t+ 1, b t+1 , c t +1 happens
after the ( 2020+t )th World Final.

a) Write down the transition matrix A . Is A a stochastic matrix?

b) Is A diagonalisable?
c) What would be the distribution of AST, VIT and G2 fans in the long
run? Hint: Something to do with equilibrium vector.

Answer

( )
0.9 0.03 0.07
a) A= 0.06 0.92 0.02 . A is a stochastic matrix.
0.04 0.05 0.91
b) No. A cannot be split into linear factors.
c) Since A is a stochastic matrix, A must have a steady state vector,
which is the probability vector and eigenvector associated with
eigenvalue 1. Solving (I – A)v = 0, you get the steady state vector v
= [1/3; 1/3; 1/3]. This means that the Markov chain will converge to
the vector (a0 + b0 + c0) * v, and in the long run, the three teams
will have equal number of fans.
Consider the following system with 2 masses and 3 springs.

By Newton’s Second Law, we have

m1 ẍ 1 + ( k 1+ k 2 ) x 1−k 2 x 2=0

m2 ẍ 2 + ( k 2+ k 1 ) x2−k 2 x 1=0

We can rewrite this into a matrix equation M ẍ + K x=0 , where M only

contains m1 and m2, K only contains k 1 and k 2, ẍ=

ẍ 1
ẍ 1
and x=
x1
x2 () ()
a) Given M = ( m1 0
0 m2 )
, find the matrix representation of K .

b) Give an assumption that will allow M to be invertible. Is it a valid

assumption?

The masses in this system exhibit oscillatory behaviour. The oscillation

equation is given by ẍ=−ω2 x , where ω is a constant.

c) Prove that ω 2 is an eigenvalue to matrix M −1 K .

From now on, we consider the case that m1=m2=m . Also, k 1 and k 2 are
positive.

d) By definition, ω is also positive. Express the value(s) of ω in terms of

k 1, k 2 and m .
e) For each ω , find the corresponding x .

Answer:

a) K= (
k 1+ k 2 −k 2
−k 2 k 1+ k 2 )
b) M is a diagonal matrix, so for M to be invertible, we must assume
that the masses m1 and m2 are positive. This is a valid assumption
(anything that occupies volume must have mass, as far as we are
concerned).
c) M ( −ω2 x ) + K x=0 ⟹−M ω2 x+ K x=0⇒ M ω 2 x=K x ⇒ M −1 K x=ω 2 x
 k1 k +2 k 2
d) ω= or ω= 1
m m
k1
()
1
e) ω= has x= . ω= 1
m 1
k +2 k 2
m ( )
has x=
1
−1
 a) (Easy one) Prove that if λ is an eigenvalue of a square matrix A , then
cλ is an eigenvalue for cA where c is a non-zero scalar.
b) (Medium to hard one) Prove that for all square matrices A and B that
satisfies A=PB P−1 , A and B have the same eigenvalues.
c) (Medium to hard one) Vieta’s formula states that for a degree-n
polynomial a n x n +a n−1 x n−1+ …+a1 x+ a0 =0, with n roots x 1 , x 2 ,… , x n, the
−an−1
sum of the roots are defined as x 1+ x2 +…+ x n= . Prove that for any
an
square matrix A , the sum of its eigenvalues is equal to the sum of
its diagonals.

a) Please don’t say that you cannot do this…

b)
Let v be an eigenvector associated to eigenvalue k of matrix A. Then, Ax
= kx.
Let x = Pw. As P is invertible, w = P^-1(x)
Rewriting B in terms of A, we have B = P^-1AP
Consider Bw = P^-1APw = P^-1Ax = P^-1(kx)=k(P^-1 * x) = kw
Hence, k is also an eigenvalue of B.
c) The key observation is that after expanding det(xI – A), the
coefficient of x^n is 1 and the coefficient of x^n-1 is negative the
sum of the diagonals.

( ) ( )
a 11 ⋯ a1 n x −a11 ⋯ −a1 n
Suppose A= ⋮ ⋱ ⋮ and xI − A= ⋮ ⋱ ⋮
an 1 ⋯ a nn −a n1 ⋯ x−ann
Define M i 1 j 1 ,i 2 j 2 ,… be the matrix minor of xI − A by deleting the i1-th row
and j1-th column, the i2-th row and j2-th column.

Also, define p ( j ) as an arbitrary polynomial with degree j .

det ( xI−A )
¿ ( x−a11 ) det (M 11 )−(−a12) det ( M 12) −… (cofactor expansion along row 1)

Notice that as this point, the degree of the first term is n, but any other
term is only n−2. That is, you can rewrite as
¿ ( x−a11 ) det (M 11 )+ p (n−2)
¿ ( x−a11 ) ¿
¿ ( x−a11 ) (x −a22)det ( M 11,22 ) +(x −a11 ) p (n−3)+ p(n−2)
¿ ( x−a11 ) ( x−a 22) det ( M 11,22 ) + p(n−2) (merge the last 2 terms to form a degree
n-2 polynomial)
¿…
¿ ( x−a11 ) ( x−a 22) … ( x−ann ) + p(n−2)
n n−1
¿ x +(−a11 −a22 …−ann )x + p(n−2)

Using Vieta’s formula,

The sum of eigenvalues
= The sum of roots of characteristic polynomial
= - coefficient of x n−1/ coefficient of x n
= sum of diagonal entries.
Andrew proposed that the price of a house, $ b, varies linearly with its area
x 1 square metres, its distance from the nearest MRT station x 2 metres, and
its number of bedrooms x 3 , that is, there exist some parameters c 1, c 2, and
c 3 such that

c 1 x 1 +c 2 x 2 +c 3 x3 =b

Andrew then chooses four random houses in Singapore. He then

expresses his findings in terms of an augmented matrix:

( )
150 1000 6 7 500 000
240 1 330 3 5 000 000
250 540 4 7 700 000
60 250 1 2300 000

a) Suppose that his hypothesis is correct, that is, the parameters c 1, c 2,

and c 3 can correctly predict the house price of any house given its
area, distance from the nearest MRT and number of bedrooms, what
is the expected house price of a house that is 180 square metres,
750 metres away from the nearest MRT and has 3 bedrooms?
b) Are there any values of c 1, c 2, and c 3 that will make Andrew’s
hypothesis correct?
c) If the answer for part (b) is yes, find the values of c 1, c 2, and c 3. Else,
calculate an approximation for c 1, c 2, and c 3 using the least squares
solution.
d) Using your answer in part (c), what is the predicted price for a house
that is 200 square metres, 700 metres away from the nearest MRT
station, and has 4 bedrooms?
e) Who is Andrew?

Answer

a) The numbers in this part are just the last row multiplied by 3.
$ 6 900 000.
b) The RREF of this matrix is the identity matrix, so no c 1, c 2, and c 3
exist that will make Andrew’s hypothesis correct.
c) c 1=16828 , c 2=−1768 , c 3=1127893.
d) $ 6 639 572.
e) Stanford CS229: Machine Learning Full Course taught by Andrew Ng
After finishing Chapter 1 or MA1522, you finally have the tools necessary
to answer one question: How is the assessment weightage of the course
CSxxxxS distributed? This is one of the few courses in NUS that do not
disclose weightage of assessments, but a Canvas loophole (probably the
professors forget to disable this) allows students to view their assessment
component results and the average results. CSxxxxS has three
assessment components: class participation, midterm examination and
final examination. After finding two friends who are comfortable in sharing
their results, you have the following information:

- You get 100% for class participation, 65% for midterm examination,
84% for final examination and your average score is 81.5%.
- Your first friend gets 90% for class participation, 75% for midterm
examination, 90% for final examination and his average score is
85.5%.
- Your second friend gets 85% for class participation, 80% for midterm
examination, 85% for final examination and her average score is
83.5%.

Calculate the weightage of each assessment component in percentage.

Answer: Class participation: 20%, midterm examination: 30%,

final examination: 50%.
A skew-symmetric matrix is a matrix that satisfies the following property:
T
K =−K .
We then have a property of square matrices: Any square matrix, not
necessarily symmetric, can be uniquely represented as the sum of a
symmetric and a skew-symmetric matrix, that is,
n×n n× n T T
∀ A ∈R , ∃! S , K ∈ R such that S =S , K =−K , A=S+ K
Furthermore, you can compute these symmetric and skew-symmetric
matrices by the following formula:
1
S= ( A+ A¿ ¿T )¿
2
K=?

a) Prove that S is a symmetric matrix.

b) Compute the value of K , and prove that K is a skew-symmetric
matrix.

Ans:

a) ST =¿ ¿
1
b) K= ( A−A ¿¿ T ) ¿.
2
T
K =¿ ¿
Random:

Qn: Let A and B be 4 by 4 matrices, if B^2 = A’A, let N be the number of

solutions for B. Which of the following the maximum range of N?
1. 2<= N <= 4
2. 1<= N <= 4
3. 1 <= N <=16
4. 2<=N <=16
remarks: 1 if A is the 0 matrix, 16 if A is a fullrank matrix. Since when
diagonalised, 4 non zero eigen value which have +- when sqrt.

Qn: Let the bijective transformation T: R3 → R3 be defined as T(v) = ABv

where A and B are matrices of unknown size. Which of the following is
true:
1. A must be a 3 by 3 matrix
2. B must be a 3 by 3 matrix
3. AB must be a 3 by 3 matrix
4. A and B must be full rank
5. A must have more rows than columns
6. B must have more rows than columns

Remarks: AB invertible, rank(AB) = 3, A and B must be full rank of 3.

hence #col A >= #row A, #row B>= #colB

Qn: If {u1,…,uk} independent, can we conclude that {T(u1),…,T(uk)} is

linearly independent?

1. True
2. False

Remarks: Depends on whether T is injective.

( )
a
Qn: Suppose T is a non-zero 3 by 3 matrix and T b = 0, which of the
a+b
following is true about T?

1. T must be rank 3
2. T must be of rank 2
3. T must be of rank 1
Remarks: Nullity T = 2 since basis of kernel is {[1;0;1], [0;1;1]}

Qn: Find all subsets X of R^n such that span(X)=X

Remarks: only finite set is the zero vector, since u need dimension of the
span to be 0.

Qn: What is span(Ø)?

Remarks: 0 vector

Qn: The span of an infinite set of vectors must also be infinite.

1. True
2. False

Remarks: Any non-empty non-zero set would give a span that is infinite

Qn: span(X) is finite iff X=Ø.

1. True
2. False

Remarks: Span of null set is 0 space

Qn: Ø can never be a basis.

1. False
2. True

Remarks: It is basis for 0 space

Qn: R^a is a subset of R^b, for positive integers a<b.

1. True
2. False

Remarks: Elements in R^a will never be an element in R^b since they

have different # of entry.

Qn: The union of subspaces is always a subspace.

 1. True
2. False

Remarks: See tutorial 5 (?)