Scribd
Upload
51 views

Homework 3

This document provides instructions for Homework 3, which includes 5 problems related to linear programming, quadratic programming, and convex optimization. Students are asked to solve the problems, write their answers legibly on paper or using LaTeX, convert their solutions to PDF, and submit it to the Assignments folder in Canvas by October 26th, 2023. The problems cover topics such as convexity/concavity of linear programs, quadratic programming duality, duality gaps, and applying duality theory and Lagrangian methods to simple convex optimization problems.

Uploaded by

zenden626
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
51 views

Homework 3

This document provides instructions for Homework 3, which includes 5 problems related to linear programming, quadratic programming, and convex optimization. Students are asked to solve the problems, write their answers legibly on paper or using LaTeX, convert their solutions to PDF, and submit it to the Assignments folder in Canvas by October 26th, 2023. The problems cover topics such as convexity/concavity of linear programs, quadratic programming duality, duality gaps, and applying duality theory and Lagrangian methods to simple convex optimization problems.

Uploaded by

zenden626
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

DSA3102: Homework 3

Assigned: 05/10/23, Due: 26/10/2023

Instructions:
• Please do all the five problems below. A (non-empty) subset of them will be graded.
• Write (legibly) on pieces of paper or Latex your answers. Convert your submission to PDF format.
• Submit to the Assignment 3 folder under Assignments in Canvas before 23:59 on 26/10/2023.

1. (Convexity and concavity of optimal value of an LP)


Consider the linear programming problem

p∗ = minn cT x subject to Ax ≤ b,
x∈R

where c ∈ Rn , A ∈ Rm×n and b ∈ RM . Prove the following statements, or provide a counterexample.


(a) The optimal value of the problem p∗ , as a function of c, is concave in c.
(b) The optimal value of the problem p∗ , as a function of b, is convex in b (you may assume that the
problem is feasible).
(c) The optimal value of the problem p∗ , as a function of A, is concave in A.
2. (Quadratic Programming Duality)
Consider the following quadratic program with H being a positive definite symmetric matrix and x
being the decision variable.
1 T
min x Hx s.t. Ax ≥ b.
x 2
Show that the dual problem is also a quadratic optimization problem that looks like
1
max λT f (b, A, H) − z T g(b, A, H)z s.t. λ ≥ 0, h(b, A, H)T λ = z.
λ,z 2

Find the functions f (b, A, H), g(b, A, H) and h(b, A, H).


3. (Duality Gap Problem)
Consider the two-dimensional problem

min ex2 subject to kxk2 − x1 ≤ 0

where the domain of the optimization variable x = (x1 , x2 ) is R2 .


(a) Is this a convex program?
(b) Calculate the primal optimal value p∗

1
(c) Calculate the dual optimal value d∗ . What is the duality gap?
(d) Can you find a Slater vector? Explain intuitively why there is or isn’t a duality gap.
4. (Duality Gap?)
Consider the two-dimensional problem

min f (x) subject to x1 ≤ 0


x∈R2

where √
f (x) = e− x1 x2
for all x ∈ dom f = R2+
(a) Is this a convex optimization problem?
(b) Evaluate the primal optimal value p∗ .
(c) Form the Lagrangian L(x, λ), dual function g(λ), and find the dual optimal value d∗ = supλ≥0 g(λ).
(d) Is there a duality gap and why?
5. (Simple Convex Optimization)
Consider the problem

min x2 + 1 subject to (x − 2)(x − 4) ≤ 0.


x∈R

The decision variable is x ∈ R.

(a) State the feasible set, the optimal value p∗ , and the optimal solution.
(b) Plot the objective x2 + 1 versus x. On the same plot, show the feasible set, optimal point and
value, and plot the Lagrangian L(x, λ) versus x for a few positive values of λ. Verify the lower
bound property p∗ ≥ inf x∈R L(x, λ) for λ ≥ 0. Derive and sketch the Lagrange dual function g(λ).
(c) State the dual problem, and verify that it is a concave maximization problem. Find the dual
optimal value and dual optimal solution d∗ . Does strong duality hold?

You might also like