You are on page 1of 2

Computational Linear Algebra – Problem Set 7

(points: 60)

Problem 1 (points: 5)
Let λ1 , .., λn ∈ C be the eigenvalues of A ∈ Cn×n . Using the characteristic equation, prove that

trace(A) = λ1 + · · · + λn and det(A) = λ1 · · · λn .

Problem 2 (points: 10)


Let f : Mn → R be a functional defined on Mn , the space of matrices of size n × n. Suppose that

• f (I) = 1, where I ∈ Mn is the identity matrix,


• f (A) is linear in each column of A, if we keep the rest of the columns fixed, and

• f is alternating, i.e., f (Â) = −f (A) where  is obtained by switching two columns of A.

Show that f (A) = det(A) for all A ∈ Mn .

Problem 3 (points: 2+3 = 5)


Let λ and µ be two eigenvalues of A ∈ Cn×n . Suppose the corresponding eigenvectors are xλ , xµ ∈
Cn . Prove that if λ 6= µ, then xλ and xµ must be linearly independent.
Next, show that if the n eigenvalues of A ∈ Cn×n are distinct, then A can be diagonalized.

Problem 4 (points: 5+2+3 = 10)


The spectral norm of A ∈ Rn×n is defined as

kAk∗ = max kAxk : kxk = 1 ,

where k·k is the Euclidean norm on Rn .


(a) Show that
 
 max kAxk : kxk 6 1 .

max y >Ax : kxk = 1 and kyk = 1 .

kAk∗ =

largest singular value of A.

(b) If A : Rn → Rm and B : Rm → Rk , then show that

kBAk∗ 6 kBk∗ kAk∗ .

(c) If A = [Aij ], then show that |Aij | 6 kAk∗ for all i, j.

Problem 5 (points: 2+4+4 = 10)


Define the shift operator S : Cn → Cn to be

S(x1 , x2 , . . . , xn ) = (xn , x1 , . . . , xn−1 ).


1. Prove that S is unitary: kSxk = ||x|| for all x ∈ Cn .
2. Determine the eigenvalues and eigenvectors of S.
3. Verify that the eigenvectors are orthogonal.

Problem 6 (points: 5+5 = 10)


Let A ∈ Cn×n be Hermitian. Prove that the following are equivalent:

1. For all x ∈ Cn , xHAx > 0.


2. The eigenvalues of A are non-negative.

Show that if the eigenvalues of A = [Aij ] are non-negative, then Aii > 0 and |Aij | 6 max(Aii , Ajj ).

Problem 7 (points: 5)
Let σmin and σmax be the smallest and largest singular values of A ∈ Rm×n . Show that

σmin = min kAxk and σmax = max kAxk,


x∈S x∈S

where S = {x ∈ Rn : kxk = 1} is the unit sphere in Rn .

Problem 8 (points: 5)
Pn
Let A = k=1 σk uk v > k be the full SVD of A ∈ R
m×n
, where r = rank(A) and σ1 > · · · > σn . Show
that

1. range(A) = span(u1 , . . . , ur ).
2. nullspace(A>) = span(ur+1 , . . . , um ).
3. range(A>) = span(v 1 , . . . , v r ).

4. nullspace(A) = span(v r+1 , . . . , v n ).

**********

You might also like