Scribd
Upload
2 views

Spring2425 Practice Set IV Solutions

The document contains solutions to Practice Set IV for the IE 303 course at Bilkent University, focusing on optimization methods. It includes problems related to Minimum and Maximum Weight Spanning Trees, the Asymmetric Traveling Salesperson Problem (ATSP), and matching exercises, providing explicit formulations and constraints. Additionally, it discusses the implications of adding constraints like subtour elimination and presents lower bounds for various relaxation formulations in the context of the Traveling Salesperson Problem.

Uploaded by

emreuyan52
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
2 views

Spring2425 Practice Set IV Solutions

The document contains solutions to Practice Set IV for the IE 303 course at Bilkent University, focusing on optimization methods. It includes problems related to Minimum and Maximum Weight Spanning Trees, the Asymmetric Traveling Salesperson Problem (ATSP), and matching exercises, providing explicit formulations and constraints. Additionally, it discusses the implications of adding constraints like subtour elimination and presents lower bounds for various relaxation formulations in the context of the Traveling Salesperson Problem.

Uploaded by

emreuyan52
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

IE 303 Modeling and Methods in

Optimization

Practice Set IV Solutions

Bora Çetin

Bilkent University
Department of Industrial Engineering
Spring 2024-2025
IE 303 Practice Set IV

Question 1 (The Greedy Algorithm for MST)


a) The Minimum Weight Spanning Tree is

4
B 5
5
A C
3
F
4
D E
5

b) The Maximum Weight Spanning Tree is

B 7

A C 9
7
8 F
6
D E
5

Question 2 (ATSP Discussion)


a) Construct a graph with 6 nodes as you desire. Write down the explicit formula-
tions of the matching formulation.
b) In the formulation of the matching Traveling Salesperson Problem (ATSP) overy-
our graph, the Matching constraints are NOT enough to formulate the problem.
To model the problem fully, we require some additional constraints, such as
SUBTOUR elimination constraints. Why do we require subtour elimination con-
straints in this case? Clearly explain verbally. Give one example graph where
the matching formulation without the subtour constraints fails to model ATSP.
Write down the subtour elimination constraints for your graph explicitly.
c) To your model in part a), add MTZ-type constraints to your model explicitly.
d) Show that adding MTZ constraints eliminate potential subtours in the graph.

1 April 20, 2025


IE 303 Practice Set IV

Question 3 (One-Tree Formulation)


9 5 4
1 3 6 8

6 1
2
4 7

Indeed, this gives a lower bound to our minimization problem. The explicit formulation
is left as an exercise. (because it is just restatement of the lecture material) However,
below you can find one example for the cycle elimination constraints.
x67 + x78 + x68 ≤ 2

Question 4 (Matching Exercise)


Observe that we have the following graph.

Consider the formulation of Maximum Cardinality Matching Problem. For every node
we can pick at most 1 of its neighboring arcs. Take the LP relaxation. We have
X
xe ≤ 1 For every node
Every Neighbor

Sum up all these constraints to get


X
2xe ≤ 9
All arcs

This gives X
xe ≤ 4.5
All arcs

2 April 20, 2025


IE 303 Practice Set IV

This is nothing but the objective function. So you can have at most 4.5 in LP relaxation.
We are allowed to have the following solution with the objective function value of 4.5.

Corresponding odd cycle inequality is

x12 + x13 + x23 ≤ 1

This is actually easy to observe. In a triangle, we cannot pick any 2 edges, because
it would violate our matching constraints. If the edge (1,2) were not present, then we
wouldn’t have any odd cycles. Since there would be no odd cycles in this graph, the
LP relaxation would be exact, and you could match at most 4 couples out of 9 people,
meaning that the optimal solution would be 4.

Question 5 (Delivery TSP)


Now we need to apply 1 − tree and 2 − M atching relaxation formulations. For 2 −
matching relaxation notice that some distances are drastically lower than others. For
instance, consider nodes 1, 2 and 3. A triangle formed by 1, 2 and 3 satisfies 2 −
matching conditions for all these nodes. Similarly take {4, 5, 6} and remaining nodes
form {7, 8, 9, 10}. 2 − matching relaxation will be the following.

3 April 20, 2025


IE 303 Practice Set IV

This relaxation has the distance of 2.83 + 2.24 + 1.00 + 7.21 + 5.10 + 5.10 + 3.16 +
6.32 + 8.06 + 8.06 = 49.08. Moving forward with the 1 − tree relaxation, use the
1 − tree algorithm to find a suitable tree. Remove node 1, then find The Minimum
Weight Spanning Tree on this graph. Pick arcs 2 − 3, 7 − 8, 8 − 9, 4 − 6, 5 − 6,9 − 10,
2 − 5, 5 − 7. This is a spanning tree. Now add node 1 and two shortest edges. Add
1 − 2 and 1 − 3. We get the following 1 − tree with a cost of 56.73.

These 2−M atching and 1−T ree relaxations are lower bounds for our problem because
when extra constraints are added to their formulations, we expect the objective function
value to get worse, or since this is a minimization problem, increase. Also, notice that
every TSP is a 2 − M atching but the converse is not true. You can think of TSP as
a specific member of the 2 − M atching family in this graph. When you find the best
2 − M atching in this graph, it may not be the TSP solution. Considering all lower
bounds we now have
49.08 ≤ 56.73 ≤ z T SP
Thus, the best interval we have found is [56.73, +∞).

4 April 20, 2025

You might also like