0% found this document useful (0 votes)
169 views9 pages

Problem Set 1: January 7, 2014

Here are the key steps to solve this quadratic optimization problem: 1) Write the objective function and constraint as: min cTx s.t. xTx ≤ 1 2) Form the Lagrangian: L(x,λ) = cTx + λ(1 - xTx) 3) Take the partial derivatives and set equal to 0: ∇xL = c - 2λx = 0 4) Solve for x: x = c/(2λ) 5) Substitute x back into the constraint: cTc/(2λ) ≤ 1 6) Solve for λ: λ = cTc/2 7) Substitute λ back into the

Uploaded by

VishnuVardhan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
169 views9 pages

Problem Set 1: January 7, 2014

Here are the key steps to solve this quadratic optimization problem: 1) Write the objective function and constraint as: min cTx s.t. xTx ≤ 1 2) Form the Lagrangian: L(x,λ) = cTx + λ(1 - xTx) 3) Take the partial derivatives and set equal to 0: ∇xL = c - 2λx = 0 4) Solve for x: x = c/(2λ) 5) Substitute x back into the constraint: cTc/(2λ) ≤ 1 6) Solve for λ: λ = cTc/2 7) Substitute λ back into the

Uploaded by

VishnuVardhan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

Problem Set 1

January 7, 2014
1.

(a) Prove that for rectangular matrices B, D, and invertible matrices A, C


A(A + BCD)1 = I B(C1 + DA1 B)1 DA1 .
(b) Using the above result, prove that
(A1 + BT C1 B)1 BT C1 = ABT (BABT + C)1 .
(c) Find an expression for (A + xyT )1 in terms of A1 , x and y for invertible
matrix A and (column) vectors x and y.

2. Let i (A) denote an eigenvalue of A


(a) i (AB) = i (BA) where A, B are not necessarily square.
(b) If A = AT , find the following in terms of i (A)
i.
ii.
iii.
iv.

Tr(Ap ).
i (I + cA).
i (A cI).
i (A1 ).

3. The notation A 0 denotes the fact that A is positive definite. Prove the following
results for A 0:
(a) A1 0

(b) [A]ii 0 for all i, where [A]ii denotes the i-th diagonal entry of A.
(c) For any B, rank(BABT ) = rank(B).

(d) If B is full row-rank, then BABT 0.

(e) If A  0, then for any matrix X, XAXT = 0 AX = 0.

4. Prove the following for the operator norm (with a = b, and all vector norms are
k.ka ).
(a) kIka,a = 1

(b) kAka,a = supkxk=1 kAxka


(c) kAka,a = supkxk6=0

kAxk
.
kxk

(d) Let AT A  0, then kAk22,2 = maxi i (AT A)


(e) kAk1,1 = maxj

i |[A]i,j |

(f) kAxk kAka,a kxk

(g) kABka,a kAka,a kBka,a


5. Consider X = {x Rn |Ax = b} for A Rmn and b Rm with b R(A).
Prove that the following two statements are equivalent (i.e., (a) (b) and (b)
(a))
(a) cT x = d for all x X

(b) There exists v Rm such that c = AT v and d = bT v.


6. Prove that
(a) ku + vk22 = kuk22 + kvk22 if and only if uT v = 0.

(b) 2 < a, b > +2 < x, y >=< a + x, b + y > + < a x, b y >

(c) kxk1 n kxk2

(d) kxk1 kxk2 kxk

7. Consider A Rmn where m n and rank of A is n. Suppose there exists B such


that BA = I and that AB = BT AT .
(a) Let X = AB. Show that
i. X2 = X
ii. Tr(X) = rank(X)
(b) Find an expression for B in terms of A.
(c) For any two vectors b Rn and c Rm , prove that kAb ck2
kABc ck2 .

EE609

Assignment 2

January 8, 2015

Weight: 1% if submitted by Jan 15, 2014.


Assignment will not be graded. No need to submit practice problems.

1 Assignment Problems
1. Show that the set {x Rn |kx x1 k2 kx x2 k2 } is actually a half space. and express it
in canonical form.
2. Show that the set {x Rn |kxk1 2} is a polyhedra and express it as an intersection of half
spaces (and hyperplanes, if required).
3. Show that the set {x R2++ |x1 x2 1} is convex and express it as an intersection of infinite
number of half-spaces.
4. Is the following set convex: {x Rn |x 0, xT y 1 for all y with kyk2 = 1}? Show that
the set {x Rn |xT y 1 for all y with y C} is always convex, even if C is not convex.
5. Consider the two-dimensional positive semidefinite cone S2+ defined as
{X = [ xy yz ] |x, y, z R, X  0}

(1)

Show that it can equivalently be expressed as {x, y, z R|x 0, z 0, xz y 2 }.

2 Practice Problems
6. Show that a set is affine if and only if its intersection with any line is affine.
7. What is the minimum distance between two parallel halfspaces {x Rn |aT x b1 } and
{x Rn |aT x b2 } (Hint: it depends on where the origin is)?
, b, and b, is the following true?
8. Under what conditions on a, a
{x Rn |aT x b} {x Rn |
aT x b}

(2)

9. For m < n, let b Rm and A Rmn be full row rank. Show that any affine set {x
Rn |Ax = b} can be expressed in the form {Cu + v|u Rm }. For example, the set
{x R2 |x1 + x2 = 1} can be expressed as {[u 1 u]T |u R}
10. What is the affine hull of the set {x R3 |x21 + x22 = 1, x3 = 1}.
11. Is the following set affine: {x Rn |kx x1 k1 kx x2 k1 }?
12. Given two vectors y, z Rn , consider the set S{a,b} = {x Rn |x = ay + bz}. Show that
the following set is a polyhedra, and find its boundaries.
[
[
S{a,b}
(3)
a[1,1] b[1,1]

13. Show that the set {x Rn |kxk 1} is a polyhedra and express it as an intersection of half
spaces (and hyperplanes, if required).

EE609

Assignment 2, Page 2 of 2

January 8, 2015

14. Show that the set of all doubly stochastic matrices is convex polyhedral in Rnn . A doubly
stochastic matrix is a square matrix with nonnegative entries with the property that the sum of
entries in every row and column is exactly 1.
15. What kind of set is C = {x R2 |p(0) = 1, |p(t)| 1, for 1 t 2} where p(t) =
x1 + x2 t?
16. Given , consider the set S = {x Rn |kx ak2 kx bk2 } for a 6= b. Show that S is
a halfspace for = 1, convex for < 1. Give an example to prove that S can be non-convex
for > 1.
17. Find the separating hyperplane between the two sets C = {x R2 |x2 0} and D = {x
R2+ |x1 x2 1}.
18. Express the following norm balls as intersection half-spaces (a) {x Rn |kxk2 1}, (b)
{x Rn |kxk1 1}, and (c) {x Rn |kxk 1}.
19. Which of the following sets are convex (provide proof or counterexample)
(a) {x R2 |x21 + 2ix1 x2 + i2 x22 1 i = 1, 2, . . . , 10}
(b) {x R2 |x21 + ix1 x2 + i2 x22 1 i = 1, 2, . . . , 10}
(c) {x Rn | mini xi = 1}
(d)
20. Show that the perspective function transformation, P(C) := {x/t|[xT t]T C} preserves
convexity.
21. Show that the set {x Rn |(Ax + b)T (Ax + b) (cT x + d)2 , cT x + d > 0} is a convex
cone.

EE609

Assignment 3

January 16, 2015

Weight: 1% if submitted by Jan 22, 2014.


Assignment will not be graded. No need to submit practice problems.

1 Assignment Problems
1. Starting from Jensens inequality, show that x y 1 x + (1 )y.
2. Consider a differentiable function f : R R with R+ dom f , and its running average,
defined as
Z
1 x
f (t)dt
(1)
F (x) =
x 0
with dom F = R++ . Show that F (x) is convex if f (x) is convex.
P
3. Show that the harmonic mean f (x) = ( ni=1 1/xi )1 is concave.

4. Define x[j] as the j-th largest component of x; e.g., x[1] = maxi {xi }, and x[n] =Pmini {xi }.
Given any non-negative numbers 1 2 . . . r 0, prove that f (x) = ri=1 i x[i]
is convex.

n
5. Prove
Pn the reverse Jensens inequality for a convex f with dom f = R , i > 0 and 1
i=2 i = 1

f (1 x1 2 x2 . . . n xn ) 1 f (x1 ) 2 f (x2 ) . . . n f (xn )

(2)

2 Practice Problems
6. Consider an increasing and convex function f : R R with dom f = {a, b}. Let the function
g : R R denote its inverse, i.e., g(f (x)) = x for a x b and dom g = {f (a), f (b)}.
Show that g is a concave function.
7. Show that the running average of the non-differentiable function f (x) is also convex. Since f
is not differentiable, you must use the zeroth order condition to prove the convexity of F .
8. Give an example of a function f (x) whose epigraph is (a) half-space, (b) norm cone, and (c)
polyhedron.
9. Give an example of a concave positive function (f (x) > 0) with domain Rn .
10. Show that the function f (x) is convex if and only if the function f (a + tb) is convex for all
a + tb dom f and t.
P
11. Show that the harmonic function: f (x) = ( ni=1 xai )1/a for a < 1, and a 6= 0 is concave.
P
12. Show that the entropy function H(x) = ni=1 xi log(xi ) with dom H = {x Rn++ |1T x =
1} is concave.
13. Determine if the following functions with dom f = R2++ are convex or concave
(a) f (x1 , x2 ) = x1 x2
(b) f (x1 , x2 ) = x1 /x2
(c) f (x1 , x2 ) = x21 /x2

EE609

Assignment 3, Page 2 of 2

January 16, 2015

(d) f (x1 , x2 ) = x1 x21 for 0 1.

(e) f (x1 , x2 = x1 x2
P
14. Show that the harmonic mean f (x) = ( ni=1 1/xi )1 is concave.

15. Show that the function f (x, t) = xT x/t on dom f = Rn R++ is convex in (x, t).
16. Show that the function f (x, s, t) = log(st xT x) is convex on dom f = {(x, s, t)
Rn+2 |st > xT x, s, t > 0}.
17. Show that the function f (x, t) = kxk33 /t2 is convex on {(x, t)|t > 0}.
18. Show that the function f (x) =

kAx+bk
cT x+d

is convex on {x|cT x + d > 0}.

19. Prove the following for non-differentiable functions f and g.


(a) If f and g are convex, both nondecreasing (or nonincreasing), and positive functions on
an interval, then f g is convex.
(b) If f , g are concave, positive, with one nondecreasing and the other nonincreasing, then
f g is concave.
(c) If f is convex, nondecreasing, and positive, and g is concave, nonincreasing, and positive,
then f /g is convex.
(d) If f is nonnegative and convex, and g is positive and concave, then f 2 /g is convex.
20. A function f (x) is log-concave if log(f (x)) is a concave function on {x|f (x) > 0}. Show
that the Gaussian cdf F (x) is log-concave for all x.
21. Show that the maximum of a convex function f over the box B := {xl x xu } is achieved
at one of its 2n vertices.
22. Using the following intermediate steps, show that the function f (x, t) = log(t2 xT x) over
dom f = {(x, t) Rn+1 |t > kxk2 } is convex.
(a) Show that t xT x/t is convex over dom f .

(b) Show that log(t xT x/t) is convex over dom f .


23. Show that the following function is convex on {x|kxk2 1}
f (x) =

kAx bk22
1 kxk22

(3)

24. Show that the weighted geometric mean


f (x) =

n
Y

xk k

(4)

i=1

with dom f = Rn++ is concave for k 0 and


25. Show that the Huber function
f (x) =
is convex on Rn .

Pn

i=1 k

kxk22 /2
kxk22 1/2

= 1.

kxk2 1
kxk2 > 1

(5)

EE609

Assignment 4

January 26, 2015

No submission required.

1 Assignment Problems
1. What is the solution of the following linear program
min cT x

(1)

s. t. 0 xi 1

i = 1, . . . , n.

(2)

2. Consider the following linear program


min cT x

(3)

s. t. Ax b

(4)

where A is square and full rank.


(a) When is the problem infeasible?
(b) When is the problem unbounded below?
(c) When does the problem have a finite solution, and what is it?
3. Show that any linear programming problem can be expressed as
min cT x

(5)

s. t. Ax = b

(6)

xi 0

i = 1, . . . , n

(7)

4. Solve the following quadratic optimization problem


min cT x

(8)

s. t. kxk2 1

(9)

5. Consider the sinusoidal measurement model in Gaussian noise described as,


y(n) = d + c cos(2f0 n) + s sin(2f0 n) + w(n), 0 n N 1,

(10)



where w(n) is additive white Gaussian noise and E |w(n)|2 = n2 . Answer the questions
that follow
(a) Formulate the LS estimation problem for parameters d , c , s .
(b) Derive the LS estimator of d , c , s with a suitable approximation for large N .

EE609

Assignment 4, Page 2 of 3

January 26, 2015

2 Practice Problems
6. Consider the following linear program
min cT x

(11)

s. t. Ax = b

(12)

(a) When is the problem infeasible?


(b) When is the problem unbounded below?
(c) When does the problem have a finite solution, and what is it?
7. Consider the following linear program
min cT x

(13)

s. t. a x b

(14)

where a 6= 0.
(a) When is the problem infeasible?
(b) When is the problem unbounded below?
(c) When does the problem have a finite solution, and what is it?
8. Given that ci 0 for i = 1, 2, . . . , n, solve the following problem
min cT x

(15)

s. t. i xi bi

i = 1, . . . , n

(16)

where i bi for all i = 1, 2, . . . , n.


9. What is the solution of the following linear program
min cT x
n
X
xi = 2
s. t.

(17)
(18)

i=1

0 xi 1

i = 1, . . . , n

(19)

10. What is the solution of the following linear program


min cT x
n
X
xi = 1
s. t.

(20)
(21)

i=1

xi 0

i = 1, . . . , n

(22)

11. Consider the two sets:


S1 = {v1 , v2 , . . . , vp }

(23)

S2 = {u1 , u2 , . . . , uq }

(24)

EE609

Assignment 4, Page 3 of 3

January 26, 2015

Formulate the linear optimization problem to determine the separating hyperplane, i.e., find
a Rn and b R such that
aT v i b

i = 1, . . . , p

(25)

aT u i b

i = 1, . . . , q

(26)

Ensure that your problem excludes the trivial solution a = b = 0.


12. Consider the following quadratic optimization problem
min cT x

(27)

s. t. x Ax 1

(28)

Let i for i = 1, 2, . . . n be the real eigenvalues of a symmetric matrix A. What is the solution
of this problem for the following cases.
(a) If i > 0 for i = 1, 2, . . . n.
(b) If 1 < 0, while i > 0 for i = 2, . . . n.
(c) If 1 = 0, while i > 0 for i = 2, . . . n.
13. Solve the following optimization problem for A 0,
min cT x

(29)
T

s. t. (x xc ) A(x xc ) 1

(30)

14. Solve the following optimization problem for A 0,


min xT Ax
s. t.

kxk22

(31)
(32)

Hint: Given that the eigenvalue decomposition A = UUT , use the change of variable
y = UT x.

You might also like