EE61012 – Spring Semester 2024-25

Homework 1: Convex Optimization in Control and Signal Processing

Prof. Ashish Ranjan Hota
Department of Electrical Engineering, IIT Kharagpur

Q 1.1: Bounded Set

Let {xn }n∈N be a sequence that converges to x∗ ∈ X. Show that the set {xn }n∈N = {x1 , x2 , . . .} ⊆
X is bounded. Recall that a sequence converges if for every ϵ > 0, there exists n ∈ N such that
||x∗ − xm || ≤ ϵ for all m ≥ n.

Q 1.2: Open and Closed Set

Is the set X = {x ∈ R2 |x1 + x2 = 1} an open set? Is it a closed set?

Q 1.3: Intersection of Closed Sets

Prove that intersection of arbitrarily many closed sets is a closed set.

Q 1.4: Convex Hull

Let C = {x ∈ R|1 ≤ x ≤ 2, 3 ≤ x ≤ 4}. Find the affine hull, conic hull, and convex hull of C.

Q 1.5: Convex Set

Let x0 , v ∈ Rn . Let C = {x ∈ Rn |x = x0 + αv, α ≥ 0} be the set of points that lie on the ray
originating from x0 along the direction v. Is C an affine set? Is it a convex set? Is it a cone?

Q 1.6: Convex Set

Show that the set C = {x ∈ R2 |x1 x2 ≥ 1, x1 ≥ 0, x2 ≥ 0} is a convex set. Hint: If a, b ≥ 0 and

0 ≤ θ ≤ 1, then aθ b(1−θ) ≤ θa + (1 − θ)b.

Q 1.7: Convex Set

Let x ∈ Rn and F0 , F1 , . . . , Fn ∈ Sn where Sn is the set of symmetric matrices of dimension n.

A linear matrix inequality (LMI) F : Rn → Sn is an expression of the form:
X
F (x) := F0 + xi Fi ⪯ 0n×n .
i∈[n]

Show that the set {x ∈ Rn | F (x) ⪯ 0n×n } is a convex set.

Homework 1: Convex Optimization in Control and Signal Processing Spring 2024-25

Q 1.8: Dual Cones

Let K ⊆ Rn be a cone. Consider the following set:

K ∗ = {y ∈ Rn |x⊤ y ≥ 0, ∀x ∈ K}.

K ∗ is called the dual cone of the set K. Show that:

a. K ∗ is a convex set and is a cone.
b. Let K1 ⊆ K2 . Then the respective dual cones satisfy: K2∗ ⊆ K1∗ .

Q 1.9: Convex Sets and Cones

Is the set C = {(x, y)| ||x||22 ≤ y} ⊆ R2 convex? Is this a cone?

Q 1.10: Union of Convex Sets

Give an example of two convex sets X1 and X2 whose union is not convex.

Q 1.11: Convex Set

Let C ∈ Rn be a compact (closed and bounded) convex set. For a scalar α, let C(α) be the set
of all points in C whose first coordinate is closer to α than any other point. In other words,
C(α) = {x ∈ C||x1 − α| ≤ |y1 − α|, ∀y ∈ C}. Is C(α) a convex set for any choice of C and α?

Q 1.12: Cone

Let C = {x ∈ R2 |x2 ≥ 0, x2 ≥ |x1 |}. Show that C is a cone.

Q 1.13: Midsem Spring 2023-24

Determine if the following sets in R2 are convex sets together with suitable explanation.

ˆ S1 = {(x1 , x2 )|x2 ≥ exp(x1 −1) }.

ˆ S2 = {(x1 , x2 )|x2 ≤ x1 (x1 − 2)(x1 − 3)}.
ˆ S3 = {(x1 , x2 )|x21 + x22 ≤ 4, x1 ≥ 0}.
ˆ S4 = {x ∈ R2 |x21 + x22 ≤ 2}.
ˆ S5 = {x ∈ R2 |x21 + x22 ≤ 2, (x1 − 1)2 + (x2 + 1)2 ≤ 5}.
ˆ S6 = {x ∈ R2 |x21 + x22 ≤ 2} ∪ {x ∈ R2 |(x1 − 1)2 + (x2 + 1)2 ≤ 1}.

Q 1.14: Endsem Spring 2023-24

Determine if the following sets are convex sets.

ˆ S1 = {x ∈ Rn | ni=1 x2i = 1}.

P

x21 +4x22 +4x1 x2

ˆ S2 = {x ∈ R3 | 2x1 +x2 +x3 ≤ 10, 2x1 + x2 + x3 > 0}.

2
Homework 1: Convex Optimization in Control and Signal Processing Spring 2024-25

Q 1.15: Separating Hyperplane

Let C1 = {x ∈ R2 |x1 ≥ 0, x2 ≥ 0}, and let C2 = {x ∈ R2 |x1 ≤ 0, x2 ≤ 0}. Find a hyperplane

that separates C1 and C2 .

Q 1.16: Convex Function

Let xk ∈ Rn , k = 1, 2, . . . , T . Let x = [x⊤ ⊤ ⊤ ⊤ ∈ RnT where f : RnT → R. Let

1 , x2 , . . . , xk ]
PT
f (x) = k=1 x⊤ k Qk xk , where each Qk is positive semi-definite. Then show that f is a convex
function.

Q 1.17: Midsem Spring 2023-24

Let f : Rn → R be a differentiable convex function. Show that

(∇f (x) − ∇f (y))⊤ (x − y) ≥ 0, ∀x, y ∈ domf.

Q 1.18: Convex Function

Suppose that g(x) is convex and h(x) is concave. Suppose the domain of both functions is a
closed, convex set C such that both g(x) and h(x) are always positive. Prove that the function
g(x)
f (x) = h(x) is quasi-convex, i.e., all sub-level sets of f are convex sets.

Q 1.19: Convex Function

Determine whether the following functions fi : R2 → R are convex, concave or neither.

a. f1 (x1 , x2 ) = x1 x2 with domain x1 > 0, x2 > 0.

b. f2 (x1 , x2 ) = − x11x2 with domain x1 > 0, x2 > 0.
x21
c. f3 (x1 , x2 ) = x2with domain x1 ∈ R, x2 > 0.
 
d. f4 (x1 , x2 ) = x1 log 1 + βx
x1
2
with domain x1 , x2 > 0, and β ∈ R is a constant.

e. f5 (x) = −( ni=1 xai )1/a for a ∈ (0, 1).

P

f. f6 (x) = min(0.5, x, x2 ).
g. f7 (x) = maxi∈{1,2,...,n} xi − mini∈{1,2,...,n} xi .

Q 1.20: Convex Function

If f is a convex function, then g(x, t) = tf (x/t) is also a convex function. You may use the
epigraph definition of convex functions to show the above.

3
Homework 1: Convex Optimization in Control and Signal Processing Spring 2024-25

Q 1.21: Convex Optimization

Formulate an optimization problem to find the largest circle contained in the triangle X := {x ∈
R2 | x1 ≥ 0, x2 ≥ 0, 3x1 + 4x2 ≤ 12}. Clearly state the decision variables, constraints and cost
function. Determine if the problem is a convex optimization problem.

Q 1.22: Convex Optimization

Formulate an optimization problem to find the minimum distance between a circle X1 := {x ∈

R2 | x21 + x22 ≤ 1} and the line X2 := {x ∈ R2 | x1 − x2 = 2}. Clearly state the decision variables,
constraints and cost function. Determine if the problem is a convex optimization problem.

Q 1.23: Midsem Spring 2022-23

Answer the following questions.

x2 +2
1. Let f (x) = x+2 with dom(f ) = (−∞, −2). Is this function convex, concave or neither?
2. Show that a function is convex if and only if its epigraph is a convex set.
3. Consider an optimization problem with cost function f (x) = −x1 + x22 and constraint set
X = {x ∈ R2 | − x21 − x22 + 4 ≤ 0, x1 + x2 ≥ −2}. Explain with justificiation whether this
problem is a convex optimization problem or not.

Q 1.24: Midsem Spring 2022-23

Consider the set X := {x ∈ Rn |||x − z0 ||2 ≤ ||x − zi ||2 , i = 1, 2, . . . , k} where z0 , z1 , . . . , zk ∈ Rn .

Show that this set is a polyhendron and can be written as X := {x ∈ Rn |Ax ≤ b}. Find A and
b.

Q 1.25: Endsem Spring 2022-23

Consider the following optimization problem:

maximizex∈R1 2x − x2
subject to 0 ≤ x ≤ 3.
Is the above problem a convex optimization problem? Find a globally optimal solution of the
above.

Q 1.26: Endsem Spring 2022-23

Determine if the following statements are true or false. If true, justify. If false, give a counter
example.

1. A convex optimization problem can have at most one globally optimal solution.
2. A convex optimization problem must have at least one globally optimal solution.
3. Any optimization problem with an unbounded feasible region does not have an optimal
solution.

4
Homework 1: Convex Optimization in Control and Signal Processing Spring 2024-25

Q 1.27: Endsem Spring 2022-23

Consider a function f : R2 → R given by f (x) = (x1 + x22 )2 where x = [x1 x2 ].

1. Compute the gradient of f .

2. At x0 = [0, 1], is the direction d = [1, −1] a descent direction, i.e., with directional derivative
negative?
3. Find α > 0 that minimizes f (x0 + αd).

Q 1.28: Endsem Spring 2023-24

Consider the optimization problem max{g(y)|f1 (y) ≤ 0, f2 (y) ≤ 0} with g being a concave
function and f1 , f2 being convex functions. Let Yopt be the set of optimal solutions of the above
problem.
Now, suppose the problem max{g(y)|f1 (y) ≤ 0} has a unique optimal solution y ⋆ .
If f2 (y ⋆ ) ≤ 0, then show that Yopt = {y ⋆ }.

Q 1.29: Endsem Spring 2023-24

Determine if the following functions are convex functions.

⊤
ˆ f1 (x) = k1 ki=1 log(1 + e−ai x ) where x ∈ Rn .
P

ˆ f2 (x) = log(ex1 + ex2 ) for x ∈ R2 .