CO 250: Introduction To Optimization: Module 1: Formulations (Optimization On Graphs)

CO 250: Introduction to Optimization
Module 1: Formulations (Optimization on Graphs)
University
c of Waterloo 1 / 20
Shortest Paths
s a Columbia Ave
Albert St
Fisher Rd d
King St
c
Westmount Rd
Erb St
b f g
Victoria St
t
• Familiar problem: Starting at location s, we wish to travel to t.

What is the best (i.e., shortest) route?
• In the figure above, such a route is indicated in bold.
University
Shortest Paths
s a Columbia Ave
Albert St
Fisher Rd d
King St
c
Westmount Rd
Erb St
b f g
Victoria St
t
• Goal: Write the problem of finding the shortest route between s and
t as an integer program!
... How?
University
Graph Theory 101
Rephrasing this problem in the language of

graph theory helps!
y
A graph G consists of ...
x
• vertices u, w, . . . ∈ V z
(drawn as filled circles)
w t
• edges uw, wz, . . . ∈ E u
(drawn as lines connecting circles) v
Two vertices u and v are adjacent if

uv ∈ E. Vertices u and v are the
endpoints of edge uv ∈ E, and edge e ∈ E
is incident to u ∈ V if u is an endpoint of e.
University
Graphs – Why do we care?
Graphs are useful to

compactly model
many real-world
entities. For example:
• Modeling circuits
in chip design
• Social networks
Lampman, 2008 [Online Image]. Late Medieval Trade Routes. Wikimedia Commons.
http://commons.wikimedia.org/wiki/File:Late Medieval Trade Routes.jpg
• Trade networks
• .... and many

more!
University
The Map of Water Town
s a Columbia Ave
Albert St
Fisher Rd d
King St
c
Westmount Rd
Erb St
b f g
Victoria St
t
• We can think of the street map as a graph, G.
• Vertices: Road intersections
• Edges: Road segments connecting adjacent intersections
University
The Map of Water Town
s 650 a 830 600
90
80
450
630
900
700
d
590
610
5
250
41
325 c 830 600
325 b g
f
350
0
50
720
700
330
650 800 610

t
• Each edge e ∈ E is labelled by its length ce ≥ 0.
• We are looking for a path connecting s and t of smallest total length!
University
Paths
s 650 a 830 600
90
80
450
630
900
700
d
590
610
5
250
41
325 c 830 600
325 b g
f
350
0
50
720
700
330
650 800 610

t
An s, t-path in G = (V, E) is a sequence

v1 v2 , v2 v3 , v3 v4 , . . . , vk−2 vk−1 , vk−1 vk
where
• vi ∈ V and vi vi+1 ∈ E for all i, and
• v1 = s, vk = t, and vi 6= vj for all i 6= j.

(Without this, it is called an s, t-walk)
University
Paths
s 650 a
450
d
c(P ) = csa + cad + cdb + cbf + cf g + cgt
= 650 + 490 + 250 + 830 + 600 + 700
250
830 600
= 3520
b f g
700
P= sa, ad, db, bf, fg, gt
The length of a path P = v1 v2 , . . . , vk−1 vk is the sum of the lengths of

the edges on P : X
c(P ) := (ce : e ∈ P ).
University
s 650 a 830 600
90
80
450
630
900
700
d
590
610
5
250
41
325 c 830 600
325 b g
350 f
0
50
720
700
330
650 800 610

t
Shortest Path Problem
• Given: Graph G = (V, E), lengths ce ≥ 0 for all e ∈ E, s, t ∈ V
• Find: Minimum-length s, t-path P
Goal: Write an IP whose solutions are the shortest s, t-paths!

−→ Later!
University
Example: Matchings
University
WaterTech - Job Assignment
WaterTech has a collection of

Jobs
important jobs: Employees
10 20 30 40
J = {10 , 20 , 30 , 40 } 1 - 5 - 7
2 8 - 2 -
that it needs to handle urgently. 3 - 1 - -
4 8 - 3 -
It also has 4 employees:
Note: Some workers are not able to
E = {1, 2, 3, 4} handle certain jobs!
that need to handle these jobs.
Goal: Assign each worker to exactly
Employees have different skill-sets one task so that the total execution
and may take different amounts of time is smallest!
time to execute a job. −→ We will rephrase this in the
language of graphs
University
Matchings
1 1’
5 8
7 8
2 2’
Create a graph with one vertex for 4
each employee and job. 3 1 3’
Add an edge ij for i ∈ E and j ∈ J 3

4 4’
if employee i can handle job j.
E J
Let the cost cij of edge ij be the
amount of time needed by i to Jobs
complete j. Employees
10 20 30 40
1 - 5 - 7
2 8 - - 4
3 - 1 - -
4 8 - 3 -
University
Matchings
1 1’
5 8
7 8
2 2’
4
Definition 1
3 3’
A collection M ⊆ E is a matching if no two
edges ij, i0 j 0 ∈ M (ij 6= i0 j 0 ) share an 3
4 4’
endpoint; i.e., {i, j} ∩ {i0 , j 0 } = ∅.
E J
Examples
1. M = {140 , 210 , 320 , 430 } is a matching.
2. M = {140 , 320 , 410 , 430 } is not a

matching.
University
Matchings
The cost of a matching M is the 1 1’

5 8
sum of costs of its edges: 7 8
2 2’
4
X
c(M ) = (ce : e ∈ M )
3 1 3’
3
4 4’
e.g., M = {140 , 210 , 320 , 430 }
−→ c(M ) = 19 E J
Definition
A matching M is perfect if every E.g., matching in figure is perfect,
vertex v in the graph is incident to and this one is not!
an edge in M .
University
Restating the Assignment Problem
1 1’
Definition 7
5 8
8
A matching M is perfect if every vertex v in 2 2’
the graph is incident to an edge in M . 4
3 1 3’
Note: Perfect matchings correspond to 3
feasible assignments of workers to jobs! 4 4’
E.g., the matching shown corresponds to E J

the following assignment:
1 → 40 , 2 → 10 , 3 → 20 , and 4 → 30 Restatement of original

question:
whose execution time equals c(M ) = 19! Find a perfect matching M
in our graph of smallest cost.
University
A Little More Notation...
Notation: Use δ(v) to denote the set of 1 1’

5 8
edges incident to v; i.e., 7 8
2 2’
δ(v) = {e ∈ E : e = vu for some u ∈ V }. 4
3 1 3’
3
4 4’
Definition
E J
Given G = (V, E), M ⊆ E is a perfect
matching iff M ∩ δ(v) contains a single
edge for all v ∈ V .
Examples
• δ(2) = {210 , 240 }

• δ(30 ) = {430 }
University
An IP for Perfect Matchings
1 1’
Definition 7
5 8
8
Given G = (V, E), M ⊆ E is a perfect 2 2’
matching iff M ∩ δ(v) contains a single 4
edge for all v ∈ V . 3 1 3’
3
The IP will have a binary variable xe for 4 4’
every edge e ∈ E. Idea: E J
xe = 1 ↔ e ∈ M
Objective:
Constraints: For all v ∈ V , need X
X (ce xe : e ∈ E)
(xe : e ∈ δ(v)) = 1
University
An IP for Perfect Matchings
X
min (ce xe : e ∈ E) 2
X 1 5
s.t. (xe : e ∈ δ(v)) = 1 (v ∈ V )
4
x ≥ 0, x integer 1
3
3
min (5, 1, 3, 4)x 4
12 13 14 23
 
1 1 1 1 0
 
2
 1 0 0 1 x = 1

s.t.
3 0 1 0 1
 

4 0 0 1 0
x ≥ 0 integer
University
Optimization on Graphs
Recap
• Graphs consist of vertices V and edges E ... and are very useful in
modeling many practical problems.
• In particular, graphs can be used to model road networks, where

roads are edges and street intersections are vertices.
• In the shortest path problem, each edge e ∈ E has an associated

weight ce , and we are looking for a path connecting two specific
vertices of smallest total weight.
• A matching is a collection of edges, no two of which share an

endpoint. A perfect matching is a matching that covers all vertices
in V .
University

CO 250: Introduction To Optimization: Module 1: Formulations (Optimization On Graphs)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CO 250: Introduction To Optimization: Module 1: Formulations (Optimization On Graphs)

Uploaded by

Copyright:

Available Formats

CO 250: Introduction to Optimization

Module 1: Formulations (Optimization on Graphs)

• Familiar problem: Starting at location s, we wish to travel to t.

• In the figure above, such a route is indicated in bold.

Rephrasing this problem in the language of

(drawn as lines connecting circles) v

Two vertices u and v are adjacent if

Graphs are useful to

• .... and many

• We can think of the street map as a graph, G.

• Vertices: Road intersections

• Edges: Road segments connecting adjacent intersections

s 650 a 830 600

650 800 610

• Each edge e ∈ E is labelled by its length ce ≥ 0.

• We are looking for a path connecting s and t of smallest total length!

650 800 610

An s, t-path in G = (V, E) is a sequence

• v1 = s, vk = t, and vi 6= vj for all i 6= j.

The length of a path P = v1 v2 , . . . , vk−1 vk is the sum of the lengths of

650 800 610

Shortest Path Problem

• Given: Graph G = (V, E), lengths ce ≥ 0 for all e ∈ E, s, t ∈ V

• Find: Minimum-length s, t-path P

Goal: Write an IP whose solutions are the shortest s, t-paths!

WaterTech has a collection of

Add an edge ij for i ∈ E and j ∈ J 3

1. M = {140 , 210 , 320 , 430 } is a matching.

2. M = {140 , 320 , 410 , 430 } is not a

The cost of a matching M is the 1 1’

E.g., the matching shown corresponds to E J

1 → 40 , 2 → 10 , 3 → 20 , and 4 → 30 Restatement of original

Notation: Use δ(v) to denote the set of 1 1’

• δ(2) = {210 , 240 }

min (5, 1, 3, 4)x 4

• In particular, graphs can be used to model road networks, where

• In the shortest path problem, each edge e ∈ E has an associated

• A matching is a collection of edges, no two of which share an

You might also like