Indian Institute of Science

Department of Electrical Communication Engineering

E2 212: Matrix Theory

Aug - Dec 2022

Homework 1
Deadline 15 September 17:15 hrs after the class

This homework consists of six exercises and a programming assignment related Vector Spaces.
You can turn in your hand written solutions after the class. For the programming part, turn
in your SINGLE .ipynb file or .m file via MS Teams. MS Teams submission window will
automatically close by Sept. 15th 6 pm. So plan ahead.
Note that we plan to have eight homework sets out of which we will consider the best 5
towards your final grade. Hence no deadline extension requests will be entertained.

Part A: Exercises
1. Let V be a vector space over a field F. Let Xi for i = 1, 2, 3...., K be a non empty family
of subspaces of V. Then show that
K
\
X = Xi
i=1

is a subspace of V. Further, for K = 2

T S 2, show with an example (or with a sketch in R )
that X1 X2 is a subspace and X1 X2 is not a subspace.
2. Let X and Y be subspaces of V. Assume V = X + Y. Then X ∩ Y = {0}, i.e., V = X ⊕ Y
(here, ⊕ denotes a direct sum), is equivalent to the statements:

a. If x1 + y1 = x2 + y2 , then x1 = x2 and y1 = y2 for all x1 , x2 ∈ X and y1 , y2 ∈ Y.

b. If x + y = 0, then x = y for all x ∈ X and y ∈ Y.

3. Let us denote the set of all n × n matrices over a field F as Mn (F). Let us denote the
set of all n×n symmetric matrices as Sn , i.e. A ∈ Sn means that A = AT .Let us denote
the set of all n × n skew symmetric matrices as Kn , i.e. B ∈ Kn means B = −BT .

a. Show that Sn is a subspace of Mn (F).

b. Show that Kn is a subspace of Mn (F).
c. Show that Mn (F) = Sn ⊕ Kn .

4. For A ∈ Rm×n with rank(A) = n and a subspace S ⊆ Rn×1 . Let

A(S) = {Ax | x ∈ S}

and B = {s1 , s2 , · · · , sk } be a basis for S. Find a basis for A(S) in terms of B. Explain
why it is a basis.
 5. Suppose rank(A) = 2 and rank(B) = 1. Then, is the following true or false:

1 ≤ rank(A + B) ≤ 3.

Explain your claim and give example matrices A and B to justify it.

6. For a linear transformation T : R2 7→ R2

defined
  byT(x, y) = (x−y, 2x−4y), determine
1 1
the coordinate matrix [T]B , where B = , .
1 2

Part B: Programming
Make Python routines in Google CoLab that computes a basis for each of the four fundamental
subspaces associated with a matrix A ∈ Rm×n . To do so, you will have to use the discussions
from lectures on computing the spanning sets for the fundamental subspaces. Using inbuilt
functions that output a basis is not allowed.
Test your routines to compute the fundamental subspaces and its dimension of
 
1 2 0 2 1
A = 3 6 1 9 6 .
2 4 1 7 5
Although Python is strongly encouraged, you may use Matlab. In that case, submit a
SINGLE *.m file that has all the four functions at the end of the script via MS Teams.

2
Indian Institute of Science
Department of Electrical Communication Engineering

E2 212: Matrix Theory

Aug - Dec 2022

Homework 2
Deadline 27 October 17:15 hrs after the class

This homework consists of exercises on Norms. Deadline extension requests will not be
entertained.

ith standard basis in Fn and suppose that ∥ · ∥ is a seminorm on Fn .

1. Let ei denote the P
Show that ∥x∥ ≤ ni=1 |xi |∥ei ∥.

2. Parallelogram identity: Let ∥ · ∥ be a norm on V that is derived from an inner

product.

a. Show that ∥ · ∥ satisfies

1
(∥x + y∥2 + ∥x − y∥2 ) = ∥x∥2 + ∥y∥2 .
2

b. Does ∥x∥∞ satisfy the above identity? Explain why.

3. If ∥ · ∥ is a norm on V that is derived from an inner product ⟨·, ·⟩, show that

∥x + y∥ ∥x − y∥ ≤ ∥x∥2 + ∥y∥2 .

4. ℓp norm: For real numbers p ≥ 1 and q ≥ 1 such that 1/p+1/q = 1, show the following.

a. !1/p !1/q
n
X n
X n
X
p q
|xi yi | ≤ |xi | |yi | .
i=1 i=1 i=1

b.
∥x + y∥p ≤ ∥x∥p + ∥y∥p ,
where x = [xi ], y = [yi ], and ∥x∥p = (|x1 |p + . . . + |xn |p )1/p is the ℓp norm.

5. Let ∥ · ∥α and ∥ · ∥β be two given norms on Cn , and suppose that there is some C > 0
such that ∥x∥α ≤ C∥x∥β for all x ∈ Cn . Explain why ∥x∥D D
β ≤ C∥x∥α for all x ∈ C .
n

√
6. Show that ∥x∥2 ≤ ∥x∥1 ≤ n∥x∥2 .

7. Explain why |||UAV||| = |||A||| for any A ∈ Mn and any unitary U, V ∈ Mn .

8. Explain why the spectral radius ρ : Mn → R is not a matrix norm.

 9. Let |||·||| be a matrix norm on Mn and let A ∈ Mn . Then show the asymptotic behavior
1/k
of the spectral radius ρ(A) = limk→∞ Ak .

10. For A ∈ Mn such that |||A||| < 1, show that

1 1
≤ (I − A)−1 ≤ .
1 + |||A||| 1 − |||A|||

11. Explain why the function |||A||| = max{|||A|||1 , |||A|||∞ } is a matrix norm on Mn that is
unital.

12. If |||·||| is a matrix norm on Mn , show that its adjoint is also a matrix norm.

13. Let ∥ · ∥ be a norm on Mn and let A ∈ Mn be given. Explain why for any Y ∈ Mn
with ∥Y∥ = 1, |trYH A| ≤ ∥A∥D .

14. Let |||·||| is a matrix norm on Mn that is induced by the norm ∥ · ∥ be a norm on Cn .
Then AH = max{|tr BH A| : |||B||| = 1 and rank B = 1}.
√
15. Let |||·||| is a matrix norm on Mn . Show that √1n |||X|||2 ≤ |||X|||1 ≤ n|||X|||2 .

2
Scanned by CamScanner
Scanned by CamScanner
Scanned by CamScanner
Scanned by CamScanner
Scanned by CamScanner
Scanned by CamScanner
Scanned by CamScanner
Scanned by CamScanner
Scanned by CamScanner
Indian Institute of Science
Department of Electrical Communication Engineering

E2 212: Matrix Theory

Aug - Dec 2022

Homework 3
Deadline 10 November 17:15 hrs after the class

This homework consists of exercises on Eigenvalues, eigenvectors, similarity, and unitary

equivalence. Deadline extension requests will not be entertained.

1. If Ax = λx. Explain why σ(Ā) = σ(A).

2. Consider the block diagonal matrix

 
A11 0
A= , Aii ∈ Mni
0 A22

Show that σ(A) = σ(A11 ) ∪ σ(A22 ). Also, show that if λ is an eigenvalue of A11 , then
it must be an eigenvalue of A; and if λ is an eigenvalue of A, then it is an eigenvalue of
either A11 or A22 .

3. Let A ∈ Mn be an idempotent matrix (i.e., A2 = A). Show that each eigenvalue of A

is either 0 or 1.

4. Let A ∈ Mn (R) be a real matrix. (a) Explain why all the coefficient of pA (t) are real.
(b) Show that the eigenvalues of A occur as complex conjugate pairs. (c) If n is odd,
show that A has at least one real eigenvalue.

5. Let x, y ∈ Cn , x ̸= 0, and A ∈ Mn . Suppose that Ax = λx and the eigenvalues of A

be λ, λ2 , . . . , λn . What are the eigenvalues of A + xyH .

6. If A ∈ Mn , SAS−1 = Λ = diag(λ1 , λ2 , . . . , λn ), and p(t) is a polynomial. Show that

p(A) = S−1 p(Λ)S and that p(Λ) = diag(p(λ1 ), p(λ2 ), . . . , p(λn )).

7. Let A ∈ Mn and a polynomial p(t) be given. If A is diagonalizable, show that p(A) is

diagonalizable. What about the converse?

8. Explain why the algebraic mulitplicity of an eigenvalue is greater than or equal to its
geometric multiplicity?

9. Suppose that n ≥ 2 and let T = [tij ] ∈ Mn be upper triangular. Let x be an eigenvector

associated with the eigenvalue tnn . Explain why en is a left eigenvector associated with
tnn .

10. Show that U ∈ Mn and V ∈ Mm are unitary if and only if U ⊕ V ∈ Mm+n is unitary.
(Here, ⊕ indicated block diagonal operator.)
11. If U ∈ Mn is unitary, show that |detU| = 1. Further, show that the eigenvalues of a
Householder matrix are always −1, 1, . . . , 1.

12. Cholesky factorization. Show that any B ∈ Mn of the form B = AAH , A ∈ Mn may
be written as B = LLH , in which L ∈ Mn is lower triangular and has non-negative
diagonal entries. Explain why this factorization is unique if A is nonsingular.

13. Let A ∈ Mn be factored as A = QR as in the QR factorization. Partition A =

[a1 , a2 , · · · , an ] and Q = [q1 , q2 , · · · , qn ] according to their columns. Explain why
{q1 , · · · , qk } is an orthonormal basis for span{a1 , a2 , · · · , ak } for each k = 1, . . . , n.

14. Let A, B ∈ Mn be given, and suppose that A and B are simultaneously unitarily similar
to upper triangular matrices, i.e., UH AU and UH BU are both upper triangular for
some unitary U ∈ Mn . Show that every eigenvalue of AB − BA must be zero.

15. Suppose that a diagonalizable matrix A ∈ Mn has d ≤ n distinct eigenvalues λ1 , λ2 , . . . , λd .

Let q(t) = (t − λ1 )(t − λ2 ) · · · (t − λd ). Show that q(A) = 0, so that A satisfies a poly-
nomial of degree d. Why is there no polynomial g(t) of degree strictly less that d such
that g(A) = 0?.

16. If A ∈ Mn and ϵ > 0, there −1

P is a nonsingular Sϵ ∈ Mn such that Sϵ ASϵ = Tϵ = [tij (ϵ)]
is upper triangular and j>i |tij (ϵ)| ≤ ϵ.

2
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Indian Institute of Science
Department of Electrical Communication Engineering

E2 212: Matrix Theory

Aug - Dec 2022

Homework 4 and 5
Deadline 2 December 17:30 hrs

This homework consists of exercises on Normal matrices, singular value decomposition, and
Jordan canonical form,. For logistics purpose, homework 4 and 5 are combined as one.
Handover the homework to the TA (Robin Francis) by the deadline. Deadline extension
requests will not be entertained.

Homework 4
1. Explain why a normal matrix is non-defective, i.e., the geometric multiplicity of every
eigenvalue is the same as its algebraic multiplicity.

2. If A ∈ Mn is normal, show that x ∈ Cn is a right eigenvector of A associated with the

eigenvalue λ of A if and only if x is a left eigenvector of A associated with λ.

3. Suppose that A ∈ Mn is normal, ĀA = AĀ, and A = B + iC, in which B and C are
real. Explain why B and C are normal and commute.

4. Let A ∈ Mn be normal. Show that the following statements are equivalent: (a) ĀA =
AĀ. (b) AT A = AAT .

5. Suppose we write A ∈ Mn as A = H(A) + iK(A) in which H(A) and K(A) are

Hermitian. Show that A is normal if and only if H(A) and K(A) commute.

6. Suppose we write A ∈ Mn as A = H(A) + iK(A) in which H(A) and K(A) are

Hermitian. If every eigenvector of H(A) is an eigenvector of K(A), show that A is
normal. What about the converse?

7. If A ∈ Mn is both normal and nilpotent, show that A = 0.

8. Show that A ∈ Mn is Hermitian if and only if trA2 = trAH A.

9. Let A, B ∈ Mn and B is invertible. Suppose A = VTA WH and B = VTB WH with

both TA and TB being upper triangular. Express the eigenvalues of B−1 A in terms of
TA and TB .

10. Let A ∈ Mn,m with rank(A) = r. Show that its singular values σ1 , · · · , σr are the
positive square roots of the decreasingly ordered nonzero eigenvalues of AAH , which
are the same as the decreasingly ordered nonzero eigenvalues of AH A.

11. Derive the closed-form expression for the two squared singular values of A ∈ M2 .
 12. For A ∈ Mn , show that the product of its singular values is |det(A)|.

13. Let A ∈ Mn,m with n ≥ m. Show that A has full column rank if and only if all of its
singular values are positive.

14. Let A ∈ Mn,k and B ∈ Mk,m be given. Use the singular value decomposition to show
that rank(AB) ≤ min{rank(B), rank(B)}.

15. Show that A ∈ Mn is unitary if and only if Σ = I, where Σ contains the singular values
of A.

16. Show that two complex matrices of the same size are unitarily equivalent if and only if
they have the same singular values.

Homework 5
1. Explain why every Jordan block Jk (λ) has a one-dimensional eigenspace associated with
the eigenvalue λ. Conclude that λ has geometric multiplicity 1 and algebraic multiplicity
k as an eigenvalue of Jk (λ).

2. Let A ∈ Mn , let λ be an eigenvalue of A, and let k ∈ {1, . . . , n}. Explain why

rk−1 (A, λ) − 2rk (A, λ) + rk+1 (A, λ) is the number of Jordan blocks of A that have
size k and eigenvalue λ. Here, rk (A, λ) = rank(A − λI)k and r0 (A, λ) := n.

3. Show that every principal submatrix of a Hermitian matrix is Hermitian.

4. Let B ∈ Mn be Hermitian. If B is singular and rank(B) = r, explain why λn−r (B) ≤ 0

and λr+1 (B) ≥ 0.