You are on page 1of 4

EE609 Assignment 7 March 6, 2022

1. Let c be a Gaussian random variable with mean c̄ 6= 0 and covariance matrix R. Solve the following
problem either by using bisection or by converting it into a convex program.

max P[cT x ≥ α] (1)


s. t. Fx ≤ g
Ax = b
c̄T x > α

2. Show the the following problem can be solved using bisection

min f0 (x)/h(x) (6)


s. t. fi (x) ≤ 0 i = 1, . . . , m
Ax = b

where {fi (x)} are convex, h(x) is concave, and f0 (x) is convex with domain {x ∈ domf0 |h(x) > 0}
and f0 (x) > 0. Give an example of such a problem where f0 (x) and h(x) are not affine.

3. Let D be a diagonal matrix with Dii > 0. Show that the following problem is an unconstrained
geometric program in d = diag(D):

min kDAD−1 k2F (10)

where the matrix A is given.

4. Given posynomials {fi }m m n n


i=0 and monomials {qi }i=1 , {ri }i=1 , and {si }i=1 , show that the following
problem can be expressed as a convex GP

min f0 (x) (12)


s. t. fi (x) − qi (x) ≤ 0 (13)
ri (x) − si (x) = 0 (14)

5. Pose the following problem as an SOCP and solve it in cvx


m
X n
X
min vi2 + wi2 /xi (18)
i=1 i=1
s. t. v + Bw = Ax + b (19)
x>0 (20)

See www.cvxr.com for information about installing cvx. Take m = 16, n = 8, and use the following
commands to generate A, B, and b:

A = randn(m,n);
b = randn(m,1);
B = randn(m,n);

Comment on the data that is printed by cvx.


EE609 Assignment 7, Page 2 of 4 March 6, 2022

6. Solve the following minimum length problem in x ∈ Rn

min length(x) (26)


s. t. kAx − bk ≤  (27)

where length(x) = min{k|xi = 0 for i > k}. In other words, we are trying to find the minimum
number of columns of A, taken in order, that can approximate the vector b within .

7. Show that the following problem can be solved by using bisection on a series of SOC problems:

min max{∠(y1 , z1 ), . . . , ∠(yn , zn )}


x,ym

s. t. yi = Ai x + bi i = 1, . . . , n (28)

where yi ∈ Rk , zi ∈ Rk , and

uT v
 
∠(u, v) = cos−1 (29)
kukkvk

where we take cos−1 (a) ∈ [0, π]. We are only interested in the case when the optimal objective value
does not exceed π/2.

8. Solve the following problem via bisection



p(ti )
min max − yi (33)
i=1,...,k q(ti )

where

p(t) = a0 + a1 t + . . . + am tm q(t) = 1 + b1 t + . . . + bn tn (34)

and the domain of the objective function is defined as

D = {a ∈ Rm+1 , b ∈ Rn |q(t) > 0, α ≤ t ≤ β}. (35)

The points {ti } and {yi } are given.

9. Consider a wireless network consisting of n data links, labeled 1, . . . , n. Link i transmits with power
pi > 0, and supports a data rate ri = log(1 + γi ), where γi is the signal-to-interference-plus-noise
ratio (SINR). These SINR ratios depend on the transmit powers, as described below. The system is
characterized by the link gain matrix G ∈ Rn×n ++ , where gij := [G]ij is the gain from the transmitter on
link j to the receiver for link i. The received signal power for link i is gii pi ; the noise plus interference
power for link i is given by
X
σi2 + gij pj , (39)
j6=i

where σi2 > 0 is the receiver noise power for link i. The SINR is the ratio
g p
γi := Pii i (40)
σi2 + j6=i gij pj

Page 2
EE609 Assignment 7, Page 3 of 4 March 6, 2022

The problem is to choose transmit powers {pi } subject to 0 < pi ≤ Pmax , in order to maximize the log
utility function
n
X
U (p) = log(Ri ) (41)
i=1

(This utility function can be argued to yield a fair distribution of rates.) The data are G, σi2 , and Pmax .
Formulate this problem as a convex or quasiconvex optimization problem. If you make any transforma-
tions or use any steps that are not obvious, explain.
Hint: The function log log(1 + ex ) is concave. You might find the new variables defined by zi = log pi
useful.

10. In problem 9, consider instead the problem of maximizing the minimum SINR, i.e.,

max min γi (47)


i
g p
s. t. γi := Pii i (48)
σi2 +
j6=i gij pj
max
0 ≤ p i ≤ Pi ∀ i = 1, . . . , n (49)
gp
X
pk ≤ P` ∀` (50)
k∈K`
X n
gij pj ≤ Pirc ∀ i = 1, . . . , n (51)
j=1

Here the transmitters are partitioned into groups sharing the same power supply, so that for each subset
gp
K` , the total power cannot exceed P` . Formulate the problem as a quasiconvex optimization problem
and solve it using bisection in CVX. Take n = 5 transmitters, partitioned into groups K1 = {1, 2} and
gp gp
K2 = {3, 4, 5}, Pimax = 3, P1 = 4, P2 = 6, noise σi2 = 0.5, and Pirc = 5. The channel matrix is
given by
 
1.0 0.1 0.2 0.1 0.0
0.1 1.0 0.1 0.1 0.0
 
G= 0.2 0.1 2.0 0.2 0.2 (52)

0.1 0.1 0.2 1.0 0.1
0.0 0.0 0.2 0.1 1.0

When implementing bisection, you will have to think of some reasonable starting points. Make sure
that the problem is solved to an accuracy of 0.01 in SINR.

11. Let A be an n × n matrix with positive entries, and let c and d be positive vectors that satisfy 1T c =
1T d = 1. Consider the geometric program

min xT Ay (53)
n
Y
s. t. xci i = 1 (54)
i=1
n
d
Y
yj j = 1 (55)
i=1

Page 3
EE609 Assignment 7, Page 4 of 4 March 6, 2022

with variables x, y ∈ Rn and implicit constraints x, y > 0. Write the constraints of this geometric
program in convex form and derive the first order optimality conditions. Show that if x? and y? are
optimal, then the matrix
1
B= diag(x? )Adiag(y? ) (56)
x? T Ay?

satisfies B1 = c and BT 1 = d.

Page 4

You might also like