GRAPH ALGORITHMS

Lecture 20
CS 2110 — Fall 2019
 Prelim 2 is on it's way!
2

 Check Canvas to see what time you have been

assigned to take the exam (5:30 or 7:30)
 If you can't make your assigned time, fill out the
Prelim 2 conflict survey by next Thursday
 Locations TBA
 JavaHyperText Topics
3

“Graphs”, topics:
 4: DAGs, topological sort

 5: Planarity

 6: Graph coloring
4 Sorting
 CS core course prerequisites (simplified)

2110 3410 4410

1110 3110

2800 4820

Problem: find an order in which you can

take courses without violating prerequisites

e.g. 1110, 2110, 2800, 3110, 3410, 4410, 4820

 Topological order
6

A topological order of directed graph G is an

ordering of its vertices as v1, v2, …, vn, such that for
every edge (vi, vj), it holds that i < j.

Intuition: line up the vertices with all edges pointing left

to right.

3410 4410
2110
1110 3110 4820
2800

Other applications: dependencies between lines of code (data

flow graphs, control flow graphs), task scheduling, etc.
 Cycles
7

 A directed graph can be topologically ordered if

and only if it has no cycles
 A cycle is a path v0, v1, ..., vp such that v0 = vp
 A graph is acyclic if it has no cycles
 A directed acyclic graph is a DAG
A A

B B
C C

E D D
E
DAG Not a DAG
 Is this graph a DAG?
8

B D

C
Yes!
A F It was a DAG.
E

 Deleting a vertex with indegree zero would not

remove any cycles
 Keep deleting such vertices and see whether graph
“disappears”
And the order in which we removed vertices was a topological order!
 Algorithm: topological sort
9

k= 0;
A 0
// inv: k nodes have been given numbers in 1..k in such a way that
B 1
if n1 <= n2, there is no edge from n2 to n1.
C
while (there is a node of in-degree 0) { D
k= 0 1
Let n be a node of in-degree 0; E
Give it number k; F
Delete n and all edges leaving it from the graph.
k= k+1;
1 0 2
} B D
2 1
JavaHyperText shows how to C
implement efficiently:
O(V+E) running time. A F
0 E 3
3 2
10 Graph Coloring
 Map coloring
11

How many colors are needed to ensure adjacent states

have different colors?
 Graph coloring
12

Coloring: assignment of color to each vertex.

Adjacent vertices must have different colors.

B D

A F
E

How many colors needed?

 Uses of graph coloring
13

And more! http://ijcit.org/ijcit_papers/vol3no2/IJCIT-130101.pdf

 How to color a graph
14

void color() { B D
for each vertex v in graph:
C
c= find_color(neighbors of v);
A F
color v with c; E
}
int find_color(vs) { Assume colors are integers 0, 1, …
int[] used;
assign used[c] the number of vertices in vs that are colored c

return smallest c such that used[c] == 0;

}
 How to color a graph
15

void color() { B D
for each vertex v in graph:
C
c= find_color(neighbors of v);
A F
color v with c; E
}
int find_color(vs) { Assume colors are integers 0, 1, …
int[] used= new int[vs.length() + 1];
for each vertex v in vs:
if color(v) <= vs.length(): If there are d
used[color(v)]++; vertices, need at
} most d+1
return smallest c such that used[c] == 0; available colors
}
 Analysis
16

void color() { Total time: O(E)

for each vertex v in graph:
c= find_color(neighbors of v);
color v with c;
} Time: O(# neighbors of v)
int find_color(vs) {
int[] used= new int[vs.length() + 1];
for each vertex v in vs:
Time:
if color(v) <= vs.length(): O(vs.length())
used[color(v)]++;
}
return smallest c such that used[c] == 0;
}
 Analysis
17

void color() {
for each vertex v in graph: Use the minimum
c= find_color(neighbors of v);
number of colors?
color v with c;
Maybe! Depends
}
on order vertices
int find_color(vs) {
processed.
int[] used= new int[vs.length() + 1];
for each vertex v in vs:
if color(v) <= vs.length():
used[color(v)]++;
}
return smallest c such that used[c] == 0;
}
 Analysis
18

Best coloring Worst coloring

Vertices labeled in order of processing

Only 2 colors needed for this special kind of graph…
Source: https://en.wikipedia.org/wiki/Grundy_number#/media/File:Greedy_colourings.svg
 Bipartite graphs
19

Bipartite: vertices can be

A
partitioned into two sets such that
1
no edge connects two vertices in
B
the same set
2
C
Matching problems: 3
 Med students & hospital D
residencies Fact: G is bipartite
 TAs to discussion sections iff G is 2-colorable
 Football players to teams
 Four Color Theorem
20

Every “map-like” graph is 4-colorable

[Appel & Haken, 1976]
Four Color Theorem
Proof required checking that 1,936 special graphs had
a certain property
 Appel & Haken used a computer program to check the
1,936 graphs
 Does that count as a proof?
 Gries looked at their computer program and found an
error; it could be fixed

In 2008 entire proof formalized in Coq proof

assistant [Gonthier & Werner]: see CS 4160
 Four Color Theorem
22

Every “map-like” graph is 4-colorable

[Appel & Haken, 1976]

…“map-like”?
= planar
23 Planar Graphs
 Planarity
24

A graph is planar if it can be drawn in the plane

without any edges crossing

B D

A F
E

Discuss: Is this graph planar?

 Planarity
25

A graph is planar if it can be drawn in the plane

without any edges crossing

B D

A F
E

Discuss: Is this graph planar?

 Planarity
26

A graph is planar if it can be drawn in the plane

without any edges crossing

B D

A F
E

Discuss: Is this graph planar?

YES!
 Detecting Planarity
27

Kuratowski's Theorem:

K5
K3,3

A graph is planar if and only if it does not contain a

copy of K5 or K3,3 (possibly with other nodes along
the edges shown).
 John Hopcroft & Robert Tarjan
28

 Turing Award in 1986 “for fundamental

achievements in the design and analysis of
algorithms and data structures”

 One of their fundamental achievements was a O(V)

algorithm for determining whether a graph is
planar.
 David Gries & Jinyun Xue
29

Tech Report, 1988

Abstract: We give a rigorous, yet, we hope, readable,
presentation of the Hopcroft-Tarjan linear algorithm for
testing the planarity of a graph, using more modern
principles and techniques for developing and presenting
algorithms that have been developed in the past 10-12
years (their algorithm appeared in the early 1970's).
Our algorithm not only tests planarity but also
constructs a planar embedding, and in a fairly
straightforward manner. The paper concludes with a
short discussion of the advantages of our approach.