DSA4212 Page 1

NATIONAL UNIVERSITY OF SINGAPORE

DSA4212: Optimization

Midterm Examination
March 5th, 2021

Name:

Student Number (XX):

Student Matric (MMMMMM):

Signature:

File Naming Convention: XX-MMMMMM.pdf

[e.g. 27-A01863783A.pdf]

(a) Correct answer: 2 points. Wrong answer: −3 points. No answer: 0 point.

(b) Note: the final mark cannot be negative.

(c) Write extremely clearly which question you are answering to. Try as much as possible
to answer the questions in order.

(d) This is a closed book examination. You are allowed to a one-side cheat sheet.

(e) If the grader cannot decipher your writing, the answer will be considered not answered.

(f) Do not solve any exercise on the first page.

Continued
 DSA4212 Page 2

   
1 2 1
1. Compute hu, A ui where A = and u = .
3 4 −1

2. Suppose that it takes 1 Mo to store a matrix M ∈ R1000,1000 . How much memory

does it takes to store a matrix N ∈ R3000,3000 .

3. Suppose that it takes 1 second to invert an invertible matrix M ∈ R1000,1000 . Esti-

mate roughly how long it takes to invert an invertible matrix N ∈ R3000,3000 .

4. What does it mean for a matrix to be diagonalizable?

 
0 1
5. Find the eigenvalues of A =
−2 −3

 
1 1
6. Is the matrix A = diagonalizable?
0 1

7. Is it true that any symmetric matrix S ∈ R3,3 is diagonalizable?

8. Find argmin F where F : R → R is defined as F (x) = |x + 3|.

9. Find min G where G : R → R is defined as G(x) = x2 + 6x.

Continued
 DSA4212 Page 3


1 1
10. Find the gradient of the function F (x) = hx, M xi where M = .
0 1

11. Find the Hessian of the function F : R2 → R where F (x, y) = sin(x + 2 y).

12. Is the product of two symmetric matrices always symmetric?

13. Is the sum of two invertible matrices always invertible?

14. Is the product of two invertible matrices always invertible?

15. Is the sum of two convex functions always a convex function?

16. Is the difference of two convex functions always a convex function?

m m n
17. Let 0 < m < M be two real numbers: compute limn→∞ (1 + M
) + (1 − M
) .

18. Let S be a symmetric and positive semi-definite matrix. Is it true that the elements
on the diagonal of S are always larger or equal to zero?

Continued
 DSA4212 Page 4

19. Let S be a symmetric and positive semi-definite matrix. It is true that all the entries
of S are larger or equal to zero?

20. Let S be a symmetric and positive semi-definite matrix. It is true that the deter-
minant of S is always larger or equal to zero?

21. Let F : R2 → R be a smooth function and x ∈ R2 be a vector. Is it true that its

Hessian matrix Hess(x) is necessarily a symmetric matrix.

22. Is the function F : R2 → R defined as F (x, y) = 3x − 2y convex?

23. Let X ∈ Rn,p be a matrix and Y ∈ Rn be a vector. Consider the function F : Rp → R

defined as F (β) = kY − Xβk2 + 3 kβk2 . Find β? = argmin F (β).

24. Is the function F : (0, +∞) → R defined as F (x) = x log(x) convex?

25. The matrix A ∈ R2,2 has eigenvalues 2 and 3. The matrix B ∈ R2,2 has eigenvalues
100 and 200. Is it true that the matrix A + B has eigenvalues 102 and 203?

26. Consider R ∈ R3,3 . Is it true that the matrix M = R RT is always positive definite?

Continued
 DSA4212 Page 5

27. Let A ∈ Rd,d be a matrix and b ∈ Rd be a vector. Compute the gradient of the
function F (x) = kAx + bk2 .

28. Let S ∈ Rd,d be a symmetric and positive semi-definite matrix. Compute the
gradient of the function G(x) = log[1 + hx, S xi].

29. Is it true that a convex function F : R → R always has at least one local minimum?

30. Is it true that for any two matrices A, B ∈ Rd,d we always have that A B = B A?

31. Is it true that for any two matrices A, B ∈ Rd,d we always have that Tr(A B) =
Tr(B A).

32. Let X ∈ R3,2 be a matrix and y ∈ R2 be a vector. Is it true that the expression
(X T X)−1 X T y can be simplified as X −1 y.

33. Suppose that it takes 1 second to compute the product of two matrices of dimension
1000 × 1000. Roughly how long does it take to compute the product of two matrices
of dimension 3000 × 3000?

34. Let F : R2 → R be a convex function with a unique minimum at x? = (0, 0) ∈ R2 . Is

it true that the gradient descent algorithm defined as xn+1 = xn − 0.0001 × ∇F (xn )
always converges to x? .

Continued
 DSA4212 Page 6

35. What is the minimum of the function F : R → R defined as F (x) = (3x + 2)2 ?

36. Let F : R2 → R be a smooth convex function. Is it true that the function G(x) =
F (x) + kxk2 is always strictly convex?

37. Let A ∈ R3,3 be an invertible matrix and y ∈ R3 be vector. Is it true that the
equation Ax = y always has a unique solution given by x = A−1 y ?

38. Let A be a symmetric and positive semi-definite matrix. Let B be a symmetric

a positive definite matrix. Is it true that C = A + B is always a symmetric and
positive definite matrix?

39. What is the global minimum of the function f (x) = (1 + cos[x]) + 3 (x + π)2 ?

40. The diagonalizable matrix M ∈ R2,2 has eigenvalues 2 and 7. Is it true that the
matrix 2M has eigenvalues 4 and 14 ?

41. The diagonalizable matrix M ∈ R2,2 has eigenvalues 2 and 7. Let Id be the identity
matrix. Is it true that the matrix Id + M has eigenvalues 3 and 8 ?

42. The diagonalizable matrix M ∈ R2,2 has eigenvalues 2 and 7. Is it true that the
matrix M 2 has eigenvalues 4 and 49 ?

Continued
 DSA4212 Page 7

43. Let S be a symmetric and positive definite matrix. Is it true that the matrix S −1
is also symmetric and positive definite?

44. Let f : R → R be a convex function. Is the function g(x) = 3 f (x) − 7 also convex?

45. Let f : R → R be a convex function. Is the function g(x) = 3 − 7 f (x) also convex?

46. Let f : R → R be a convex function. Is the function g(x) = f (x)2 also convex?

47. Let f : R → R be a convex function. Is it possible for the function g(x) = sin[f (x)]
to also be convex?

48. Let A be a matrix and Id the identity matrix. Is the matrix Id + AT A always
positive definite?

49. Is it true that a covariance matrix is always symmetric and positive definite?

2
 function F : R → R defined as F (x) = kxk. Compute the gradient of
50. Consider the
1
F at x = .
2

End of Paper