Graph Theory Notes

Discrete Mathematics & Mathematical Reasoning Chapter 10: Graphs
Kousha Etessami
U. of Edinburgh, UK
Kousha Etessami (U. of Edinburgh, UK)
Discrete Mathematics (Chapter 6)
1 / 13
Overview
Graphs and Graph Models Graph Terminology and Special Types of Graphs Representations of Graphs, and Graph Isomorphism Connectivity Euler and Hamiltonian Paths Brief look at other topics like graph coloring
2 / 13
What is a Graph?
Informally, a graph consists of a non-empty set of vertices (or nodes), and a set E of edges that connect (pairs of) nodes. But different types of graphs (undirected, directed, simple, multigraph, . . .) have different formal denitions, depending on what kinds of edges are allowed. This creates a lot of (often inconsistent) terminology. Before formalizing, lets see some examples....
During this course, we focus almost exclusively on standard (undirected) graphs and directed graphs, which are our rst two examples.
3 / 13
A (simple undirected) graph:
ED
NY SF
LA
Only undirected edges; at most one edge between any pair of distinct nodes; and no loops (edges between a node and itself).
Kousha Etessami (U. of Edinburgh, UK) Discrete Mathematics (Chapter 6) 4 / 13
A directed graph (digraph) (with loops):
ED
NY SF
LA
Only directed edges; at most one directed edge from any node to any node; and loops are allowed.
A simple directed graph:
ED
NY SF
LA
Only directed edges; at most one directed edge from any node to any other node; and no loops allowed.
An undirected multigraph:
ED
NY SF
LA
Only undirected edges; may contain multiple edges between a pair of nodes; but no loops.
An undirected pseudograph:
ED
NY SF
LA
Only undirected edges; may contain multiple edges between a pair of nodes; and may contain loops (even multiple loops on the same node).
A directed multigraph:
ED
NY SF
LA
Only directed edges; may contain multiple edges from one node to another; but no loops allowed. Warning: this differs slightly from the Rosen book terminology. The books notion of directed multigraph would allow loops.
An directed pseudograph:
ED
NY SF
LA
Only directed edges; may contain multiple edges from one node to another; and may contain loops (even multiple loops on one node).
Graph Terminology Zoo (ridiculous)

1. 2. 3. 4. 5. 6. 7. 8. Type (simple undirected) graph (undirected) multigraph (undirected) pseudograph directed graph simple directed graph directed multigraph directed pseudograph mixed graph Edges Undirected Undirected Undirected Directed Directed Directed Directed Both Multi-Edges? No Yes Yes No No Yes Yes Yes Loops? No No Yes Yes No No1 Yes Yes
We will focus on the two most standard types: (1.) graphs (simple undirected), and (4.) directed graphs (also known as digraphs).
1
differs from book.

Discrete Mathematics (Chapter 6) 11 / 13
Formal Dention of Directed Graphs
A directed graph (digraph), G = (V , E ), consists of a non-empty set, V , of vertices (or nodes), and a set E V V of directed edges (or arcs). Each directed edge (u , v ) E has a start (tail) vertex u , and a end (head) vertex v . Note: a directed graph G = (V , E ) is simply a set V together with a binary relation E on V .
12 / 13
Denition of (Undirected) Graphs
For a set V , let [V ]k denote the set of k -element subsets of V . (Equivalently, [V ]k is the set of all k -combinations of V .) A (simple,undirected) graph, G = (V , E ), consists of a non-empty set V of vertices (or nodes), and a set E [V ]2 of (undirected) edges. Every edge {u , v } E has two distinct vertices u = v as endpoints, and such vertices u and v are then said to be adjacent in the graph G. Note: the above denitions allow for innite graphs, where |V | = . In this course we will focus on nite graphs.
13 / 13
Graph Models: Computer Networks

network where we care about the number of links: we use a multigraph.
diagnostic self-links at data centers: we use a pseudograph.
network with multiple one-way links we use a directed !multi"graph.
Applications of Graphs
#$%$&'()*NG *+ A G&A,)-
!labeled directed etc. ..." graph theory can be used in modelling of:
+ocial networks Communications networks *nformation networks +oftware design (ransportation networks .iological networks ......
Graph Models: +ocial Networks

model social structures: relationships between people or
groups.
/ertices represent indi/iduals or organi0ations edges
represent relationships between them.
1seful graph models of social networks include:
friendship graphs - undirected graphs where two people are connected if they are friends !e.g. on 2acebook" collaboration graphs - undirected graphs where two people are connected if they collaborate in a specific way influence graphs - directed graphs where there is an edge from one person to another if the first person can influence the second
Graph Models: +ocial Networks

Example: A friendship graph: two people are connected if they are 2acebook friends.
Example: An influence graph
*nformation Networks
*n a web graph web pages are represented
by /ertices and links are represented by directed edges. *n a citation network:

&esearch papers are represented by /ertices. When paper A cites paper . there is an edge
from the /erte3 representing paper A to the /erte3 representing paper ..
(ransportation Graphs
Graph models are e3tensi/ely used to study
transportation networks. Airline networks can be modeled using directed multigraphs where:
airports are represented by /ertices each flight is represented by a directed edge
from the /erte3 representing the departure airport to the /erte3 representing the destination airport
&oad networks modeled using graphs
.iological Applications
Graph models are used e3tensi/ely in many
areas of the biological science. Niche overlap graphs model competition between species in an ecosystem:
Example: niche o/erlap graph for a forest ecosystem.
Degree and neighborhood of a vertex

Definition 3. (he degree of a vertex v in a undirected graph is the number of edges incident with it. (he degree of the /erte3 v is denoted by deg!v". Definition 3. (he neighborhood !neighbor set" of a vertex v in a undirected graph, denoted N(v) is the set of /ertices ad9acent to /.
4egrees and Neighborhoods of %ertices

Example: What are the degrees and neighborhoods of the /ertices in the graphs G and H:
Solution: deg!a" ; 7 deg!b" ; < deg!d " ; 5 N!a" ; =b, f > N!b" ; =a, c, e, f > N!d" ; =c>.
)andshaking (heorem
THEOREM 1 (Handsha in! "emma#: *f G;!% $" is a undirected graph with m edges then:
2m =v V deg ( v )
Proof: Each edge contributes twice to the degree count of all vertices. Hence, both the left-hand and right-hand sides of this equation equal twice the number of edges. QED
4egree of %ertices !continued"

Theorem $% An undirected graph has an e/en number of /ertices of odd degree. Proof% ?et V5 be the /ertices of e/en degree and V7 be the /ertices of odd degree in graph G ; !V E" with m edges. (hen
e/en must be e/en because 7m is e/en and the sum of degrees of /ertices of e/en degree is e/en. (hus since this is the sum of degrees of all /ertices of odd degree there must be an e/en number of them.
must be e/en since deg!v" is e/en for each v @ V5
)andshaking (heorem:$3amples
Example: )ow many edges are there in a graph with 56 /ertices each ha/ing degree si3: Solution: the sum of the degrees of the /ertices is A 56 ; A6. (he handshaking theorem says 7m ; A6. +o the number of edges is m ; B6. Example: *f a graph has C /ertices can each /erte3 ha/e degree B: Solution: (his is not possible by the handshaking thorem because the sum of the degrees of the /ertices B C ; 5C is odd.
4irected Graphs
Definition% (he in degree of a vertex v denoted deg!!v" is the number of edges directed into v. (he out degree of v denoted deg"!v", is the number of edges directed out of v. Note that a loop at a /erte3 contributes 5 to both in-degree and out-degree. Example% *n the graph G we ha/e
deg!!a" ; 7 deg!!b" ; 7 deg!!c" ; B deg!!d" ; 7 deg!!e" ; B deg!!f" ; 6.
4irected Graphs !continued"

Theorem 3: ?et G # !V, E" be a directed graph. (hen:
Proof: (he first sum counts the number of outgoing edges o/er all /ertices and the second sum counts the number of incoming edges o/er all /ertices. .oth sums must be D$D.
+pecial (ypes of Graphs: Complete Graphs

A complete graph on n vertices denoted by K n is the simple graph that contains e3actly one edge between each pair of distinct /ertices.
+pecial (ypes of Graphs: Cycles

A c$cle C n for n E B consists of n /ertices v5 v7 ,% , vn and edges =v5, v7>, =v7, vB> ,% , =vn 5, vn>, =vn, v5>&
+pecial (ypes of +imple Graphs: n-Cubes

An n dimensional h$percube or n cube is a graph with 2n /ertices representing all bit strings of length n where there is an edge between two /ertices if and only if they differ in e3actly one bit position.
.ipartite Graphs
Definition% An eFui/alent definition of a bipartite graph is one where it is possible to color the /ertices either red or blue so that no two ad9acent /ertices are the same color.
H is not bipartite: if we color a red then its neighbors f and b must be blue. .ut f and b are ad9acent.
G is bipartite
.ipartite Graphs !continued"

Example: +how that C 6 is bipartite. Solution: ,artition the /erte3 set into =v5 vB vC> and V7 ; =v7 v< vA>: V5 ;
Example: +how that '() is not bipartite. Solution: *f we partition /ertices of '(B into two nonempty sets one set must contain two /ertices. .ut e/ery /erte3 is connected to e/ery other. +o the two /ertices in the same partition are connected. )ence '(B is not bipartite.
Complete .ipartite Graphs

Definition% A complete bipartite graph is a graph that has its /erte3 set partitioned K m,n into two subsets V(5 of si0e m and V(7 of si0e n such that there is an edge from e/ery /erte3 in V(5 to e/ery /erte3 in V(7&
Examples%
+ubgraphs A subgraph of a graph G ; !V E" is a graph !* +" where W V and F E. A subgraph H of G is a proper subgraph of G if H , G& Example: here is K 5 and one of its !proper"
Definition% subgraphs:
*nduced +ubgraphs
Definition% ?et G ; !V E" be a graph. (he subgraph induced by a subset * of the /erte3 set V is the graph ); !* +" whose edge set + contains an edge in E if and only if both endpoints are in *& Example: )ere is -(C and its induced subgraph induced by * ; =a,b,c,e>.
.ipartite Graphs and Matchings

.ipartite graphs used e3tensi/ely in appGs in/ol/ing matching elements of two sets: .ob assignments - /ertices represent the 9obs and the employees edges link employees with 9obs they are Fualified for. Ma3imi0e H of employees matched to 9obs.
/arriage0dating - /ertices represent men I women and edges link a man I woman if they are acceptable to each other as partners.
Bipartite graphs
A bipartite graph is a (undirected) graph G = (V , E ) whose vertices can be partitioned into two disjoint sets (V1 , V2 ), with V1 V2 = and V1 V2 = V , such that for every edge e E , e = {u , v } such that u V1 and v V2 . In other words, every edge connects a vertex in V1 with a vertex in V2 . Equivalently, a graph is bipartite if and only if it is possible to color each vertex red or blue such that no two adjacent vertices are the same color.
1/9
Example of a Bipartite Graph
V1
V2
2/9
Matchings in Bipartite Graphs
A matching, M , in a graph, G = (V , E ), is a subset of edges, M E , such that there does not exist two distinct edges in M that are incident on the same vertex. In other words, if {u , v }, {w , z } M , then either {u , v } = {w , z } or {u , v } {w , z } = . A maximum matching in graph G is a matching in G with the maximum possible number of edges.
3/9
Perfect/complete matchings
For a graph G = (V , E ), we say that a subset of edges, W E , covers a subset of vertices, A V , if for all vertices u A, there exists an edge e W , such that e is incident on u , i.e., such that e = {u , v }, for some vertex v . In a bipartite graph G = (V , E ) with bipartition (V1 , V2 ), a complete matching with respect to V1 , is a matching M E that covers V1 , and a perfect matching is a matching, M E , that covers V . Question: When does a bipartite graph have a perfect matching?
4/9
Halls Marriage Theorem

For a bipartite graph G = (V , E ), with bipartition (V1 , V2 ), there exists a matching M E that covers V1 if and only if for all S V1 , |S | |N (S )|. Proof: For G = (V , E ), with A V , let NG (A) denote the neighbors of A in G. First, only if direction: Suppose there is a matching M in G that covers V1 . We show that S V1 , |S | |NG (S )|. Suppose, for contradiction, that there is a subset S V1 such that |S | > |NG (S )|. Then no matching M could possibly covers S , because there arent enough neighbors NG (S ). Done. The if direction of the proof is harder...
5/9
proof of Halls Theorem, continued...

If direction: Suppose S V1 , |S | |NG (S )|. Then we prove a matching M exists which covers V1 , by induction on the size |V1 |. Base case: |V1 | = 1. Since |V1 | |NG (V1 )|, there must be an edge covering the vertex u in V1 = {u }. Inductive step: Suppose (by inductive hypothesis) that the claim holds for bipartite graphs G with |V1 | = j k . Suppose |V1 | = k + 1.
6/9
proof of Halls Theorem (continued)

Case 1: Suppose that for every nonempty strict subset S V1 , we have |S | |NG (S )| 1. Take any {u , v } E , with u V1 . Remove u and v (and the edges incident on them) from G. Call the resulting bipartite graph G , with bipartition (V1 {u }, V2 {v }). By the induction hypothesis, there must exist a matching M in G that covers V1 {u }, because for every subset S V1 {u }, NG (S ) NG (S ) {v }, and thus |NG (S )| |NG (S )| 1 |S |. But then M = M {{u , v }} is a matching in G which covers V1 .
7/9
Case 2: Suppose, on the contrary, that there exists a nonempty strict subset S V1 with |S | = |NG (S )|. Any matching that covers V1 must match S to NG (S ). By the induction hypothesis, there is a matching M covering S on the bipartite subgraph G of G induced by S NG (S ). And furthermore, the bipartite subgraph G of G induced by (V1 S ) (V2 NG (S )) also satises the condition, and contains a matching M that covers (V1 S ). This is because if A V1 S has |A| > |NG (A)|, this implies |A S | > |NG (A S )|, which violates the assumption about G. Letting M = M M , M denes a matching in G that covers V1 .
8/9
More on Matchings
Corollary A bipartite graph G = (V , E ) with bipartition (V1 , V2 ) has a perfect matching if and only if |V1 | = |V2 | and S V1 , |S | |NG (S )|. Unfortunately, the proof we have given is not constructive enough: it doesnt yield an (efcient) algorithm to compute a maximum matching in a bipartite graph. An alternative proof of Halls theorem (which we do not give) based on alternating paths and augmenting paths, is constructive & yields an efcient (polynomial time) algorithm for computing a maximum matching.
9/9
New Graphs from 8ld

Definition: (he union of two simple graphs G5 # !V5, E5" and G7 # !V7, E7" is the simple graph with /erte3 set V5 J V7 and edge set E5 J E7. (he union of G5 and G7 is denoted by G5 J G7. Example:
&epresenting Graphs: Ad9acency ?ists

Definition: An ad1acenc$ list represents a graph !with no multiple edges" by specifying the /ertices that are ad9acent to each /erte3.
Example:
Example:
&epresentation of Graphs:
Ad9acency Matrices
Definition: +uppose that G ; !V E" is a simple graph where DVD ; n. Arbitrarily list the /ertices of G as v(5 v(7 K v(n. (he ad1acenc$ matrix ' of G with respect to this listing of /ertices is the n 2 n 6-5 matri3 with its !i 1"th entry ; 5 when v(i and v(1 are ad9acent and ;6 when they are not ad9acent.
*n other words:
A =[ aij ]
and:
Ad9acency Matrices !continued"

Example: 3he vertex ordering is is a b c d. A sparse graph has few edges relati/e to the number of possible edges. +parse graphs are more efficient to represent using an ad9acency list than an ad9acency matri3. .ut for a dense graph an ad9acency matri3 is often preferable.
(ote: (he ad9acency matri3 of an undirected graph is symmetric:
aij = a ji , i , j
Also since there are no loops each diagonal entry is 4ero5
aii =0 , i
Ad9acency Matrices !continued"

Ad9acency matrices can also be used to
represent graphs with loops and multi-edges. When multiple edges connect /ertices vi and v1 !or if multiple loops present at the same /erte3" the !i 1"th entry eFuals the number of edges connecting the pair of /ertices. Example: Ad9acency matri3 of a pseudograph using /erte3 ordering a b c d5
Adjacency Matrices (continued)

Adjacency matrices can represent directed
graphs in exactly the same way. The matrix A for a directed graph G = (V, E) has a 1 in its (i, j)th position if there is an edge from vi to vj, where v1, v2, vn is a list of the !ertices. "n other words,
aij =1 if (i , j ) E aij = 0 if (i , j ) E
#ote$ the adjacency matrix for a directed
graph need not %e symmetric.
Isomorphism of Graphs
Definition: Two (undirected) graphs G1 = (V1, E1) and G2 = (V2, E2) are isomorphic if there is a bijection, f : V 1 V 2 , with the property that for all ertices a , b V
{a ,b } E 1
if and only if
{f ( a ) , f ( b )} E 2
!uch a function f is called an isomorphism" Intuiti ely, isomorphic graphs are #T$% !&'%(, e)cept for #renamed( ertices"
*somorphism of Graphs !cont&"

Example: +how that the graphs G ;!V E" and H ; !* +" are isomorphic.
Solution: (he function f with f!u5" ; v5 f!u7" ; v< f!uB" ; vB and f!u<" ; v7 is a one-to-one correspondence between V and *.

*t is difficult to determine whether two graphs are isomorphic by brute force: there are nN bi9ections between /ertices of two n-/erte3 graphs. 8ften we can show two graphs are not isomorphic by finding a property that only one of the two graphs has. +uch a property is called graph invariant: e.g. number of /ertices of gi/en degree the degree seFuence !list of the degrees" .....

Example: Are these graphs are isomorphic:
Solution: NoN +ince deg!a" ; 7 in G a must correspond to t u x or $, since these are the /ertices of degree 7 in ). .ut each of these /ertices is ad9acent to another /erte3 of degree 7 in H which is not true for a in G. +o G and ) can not be isomorphic.
Example: 4etermine whether these two graphs are isomorphic.
Solution: (he function f is defined by: f!u5" ; vA f!u7" ; vB f!uB" ; v< f!u<" ; vC f!uC" ; v5 and f!uA" ; v7 is a bi9ection.
Algorithms for Graph *somorphism

(he best algorithms known for determining
whether two graphs are isomorphic ha/e e3ponential worst-case time comple3ity !in the number of /ertices of the graphs". )owe/er there are algorithms with good time comple3ity in many practical cases. +ee e.g. a publicly a/ailable software called NA1(' for graph isomorphism.
Applications of Graph *somorphism

(he Fuestion whether graphs are isomorphic plays an important role in applications of graph theory. 2or e3ample: Chemists use molecular graphs to model chemical compounds. %ertices represent atoms and edges represent chemical bonds. When a new compound is synthesi0ed a database of molecular graphs is checked to determine whether the new compound is isomorphic to the graph of an already known one.
Section Summary
Paths Connectedness in Undirected Graphs (strong) Connectedness in Directed Graphs
1 / 20
Paths (in undirected graphs)
Informally, a path is a sequence of edges connecting vertices. Formally:
Denition: For an undirected graph G = (V , E ), an integer n 0, and vertices u , v V , a path (or walk) of length n from u to v in G is a sequence: x0 , e1 , x1 , e2 , . . . , xn1 , en , xn of interleaved vertices xj V and edges ei E , such that x0 = u and xn = v , and such that ei = {xi 1 , xi } E for all i {1, . . . , n}. Such a path starts at u and ends at v . A path of length n 1 is called a circuit (or cycle) if n 1 and the path starts and ends at the same vertex, i.e., u = v . A path or circuit is called simple if it does not contain the same edge more than once.
More on paths
When G = (V , E ) is a simple undirected graph a path x0 , e1 , . . . , en , xn is determined uniquely by the sequence of vertices x0 , x1 , . . . , xn . So, for simple undirected graphs we can denote a path by its sequence of vertices x0 , x1 , . . . , xn .
Note 1: The word simple is overloaded. Dont confuse a simple undirected graph with a simple path. There can be a simple path in a non-simple graph, and a non-simple path in a simple graph. Note 2: The terms path and simple path used in Rosens book are not entirely standard. Other books use the terms walk and trail to denote path and simple path, respectively. Furthermore, others use path itself to mean a walk that doesnt re-visit any vertex, except possibly the rst and last in case it is a circuit. To stick to Rosens terminology, we shall use the non-standard term tidy path to refer to such a walk.
Example
Here is a simple undirected graph:

d g f
d , a, b, c , f is a simple (and tidy) path of length 4. d , g , c , b, a, d is a simple (and tidy) circuit of length 5. a, b, g , f is not a path, because {b, g } is not an edge. d , a, b, c , f , b, a, g is a path, but it is not a simple path, because the edge {a, b} occurs twice in it. c , g , a, d , g , f is a simple path, but it is not a tidy path, because vertex g occurs twice in it.
Example: an acquantance graph
Ellen David
Fred
Bob
Alice
Kathy
5 / 20
Paths in directed graphs (same denitions)

Denition: For an directed graph G = (V , E ), an integer n 0, and vertices u , v V , a path (or walk) of length n from u to v in G is a sequence of vertices and edges x0 , e1 , x1 , e2 , . . . , xn , en , such that x0 = u and xn = v , and such that ei = (xi 1 , xi ) E for all i {1, . . . , n}. When there are no multi-edges in the directed graph G, the path can be denoted (uniquely) by its vertex sequence x0 , x1 , . . . , xn . A path of length n 1 is called a circuit (or cycle) if the path starts and ends at the same vertex, i.e., u = v . A path or circuit is called simple if it does not contain the same edge more than once. (And we call it tidy if it does not contain the same vertex more than once, except possibly the rst and last in case u = v and the path is a circuit (cycle).)
Connectness in undirected graphs

Denition: An undirected graph G = (V , E ) is called connected, if there is a path between every pair of distinct vertices. It is called disconnnected otherwise.
d b e
This graph is connected
7 / 20
Connectness in undirected graphs

Denition: An undirected graph G = (V , E ) is called connected, if there is a path between every pair of distinct vertices. It is called disconnnected otherwise.
d b e
This graph is connected

d b e
This graph is not connected

Proposition
There is always a simple, and tidy, path between any pair of vertices u , v of a connected undirected graph G. Proof: By denition of connectedness, for every pair of vertices u , v , there must exist a shortest path x0 , e1 , x1 , . . . , en , xn in G such that x0 = u and xn = v . Suppose this path is not tidy, and n 1. (If n = 0, the Proposition is trivial.) Then xj = xk for some 0 j < k n. But then x0 , e1 , x1 , . . . , xj , ek +1 , xk +1 , . . . , en , xn is a shorter path from u to v , contradicting the assumption that the original path was shortest.
8 / 20
connected components of undirected graphs

Denition: A connected component H = (V , E ) of a graph G = (V , E ) is a maximal connected subgraph of G, meaning H is connected and V V and E E , but H is not a proper subgraph of a larger connected subgraph R of G.
d b e g a c f h
This graph, G = (V , E ), has 3 connected components. (It is thus a disconnected graph.) One connected component of G is H1 = (V1 , E1 ), where V1 = {d , a, b} and E1 = {{d , a}, {d , b}}.
Connectedness in directed graphs

Denition: A directed graph G = (V , E ) is called strongly connected, if for every pair of vertices u and v in V , there is a (directed) path from u to v , and a directed path from v to u . (G = (V , E ) is weakly connected if there is a path between every pair of vertices in V in the underlying undirected graph (meaning when we ignore the direction of edges in E .) A strongly connected component (SCC) of a directed graph G, is a maximal strongly connected subgraph H of G which is not contained in a larger strongly connected subgraph of G.
10 / 20
Example
d b e
This digraph, G, is not strongly connected, because, for example, there is no directed path from b to c . Question: what are the strongly connected components (SCCs) of G?
11 / 20
Example
d b e
This digraph, G, is not strongly connected, because, for example, there is no directed path from b to c . Question: what are the strongly connected components (SCCs) of G? One strongly connected component (SCC) of G is H1 = (V1 , E1 ), where V1 = {d , a, b} and E1 = {(d , a), (a, b), (b, d )}. Another SCC of G is H2 = (V2 , E2 ), where V2 = {e, c , f } and E2 = {(e, c ), (c , f ), (f , e)}. There are no other SCCs in G.
Directed Acyclic Graphs

A Directed Acyclic Graph (DAG), is a directed graph that contains no circuits or loops. Example:
f e c b a d a d c b f e
This is a DAG
This is NOT a DAG

Discrete Mathematics (Chapter 6) 12 / 21
Euler Paths and Euler Circuits Hamiltonian Paths and Hamiltonian Circuits
12 / 20
The Knigsberg Bridge Problem
Leonard Euler (1707-1783) was asked to solve the following:
Question: Can you start a walk somewhere in Knigsberg, walk across each of the 7 bridges exactly once, and end up back where you started from?
13 / 20
The Knigsberg Bridge Problem
Leonard Euler (1707-1783) was asked to solve the following:
Question: Can you start a walk somewhere in Knigsberg, walk across each of the 7 bridges exactly once, and end up back where you started from? Euler (in 1736) used graph theory to answer this question.
13 / 20
Euler paths and Euler Circuits

Recall that an (undirected) multigraph does not have any loops, but can have multiple edges between the same pair of vertices. Denition: An Euler path in a multigraph G is a simple path that contains every edge of G. (So, every edge occurs exactly once in the path.) An Euler circuit in an multigraph G is a simple circuit that contains every edge of G. (So, every edge occurs exactly once in the circuit.) Question: Is there a simple criterion for determining whether a multigraph G has an Euler path (an Euler circuit)?
14 / 20
Euler paths and Euler Circuits

Recall that an (undirected) multigraph does not have any loops, but can have multiple edges between the same pair of vertices. Denition: An Euler path in a multigraph G is a simple path that contains every edge of G. (So, every edge occurs exactly once in the path.) An Euler circuit in an multigraph G is a simple circuit that contains every edge of G. (So, every edge occurs exactly once in the circuit.) Question: Is there a simple criterion for determining whether a multigraph G has an Euler path (an Euler circuit)? Answer: Yes.
Eulers Theorem
Eulers Theorem (1736)
A connected undirected multigraph with at least two vertices has an Euler circuit if and only if each of its vertices has even degree. Proof: Only if direction: Suppose a multigraph G = (V , E ) has an Euler circuit, x0 e1 x1 e2 . . . em xm , where x0 = xm = u . For every vertex v V , v = u , each time we enter v via an edge ei , we must leave v via a different edge ei +1 . So, in total, since we see all edges incident to v exactly once, all such vertices v must have even degree. Likewise, the initial (and nal) vertex u = x0 = xm , must also have even degree, because the edges e1 and em pair up in the same way.
Proof of Eulers Theorem (continued)

The (harder) if direction: Suppose G = (V , E ) is connected and every vertex in V has even degree. We give a constructive proof that, given such a multigraph G, shows how to construct an Euler circuit (efciently). Start a walk at any vertex v , never re-using an edge, walking for as long as possible until you can not do so any more. Claim: Such a walk (simple path), w1 , must end at the vertex v where it started (i.e., it must be a circuit). Reason: For any vertex z other than v , whenever the walk enters z via an edge, there must be an odd number of edges incident to z remaining. Note: zero is not an odd number! After leaving z , there must be an even number of edges of z remaining. If the simple circuit w1 covers every edge of G, we are done. If not, .....
Proof of Eulers theorem (nal part)

Note that every vertex has even degree remaining after the edges of the simple circuit w1 are removed. If the simple circuit w1 does not cover every edge of G, since G is connected, there must be some vertex x on the circuit w1 which is incident to an edge not in w1 . So, w1 = w1 xw1 We start a new walk at the vertex x , on the remaining graph without the edges of w1 . This yields a new circuit w2 that must start and end at x . We can then then splice w2 inside w1 (at the point where x occurs) in order to get a new longer Euler circuit: w1 w2 w1 . We can do this same process repeatedly until there are no edges remaining. Note: this also yields a reasonably efcient algorithm for computing an Euler circuit in a connected multigraph where every vertex has even degree.
Eulers theorem for paths

Eulers Theorem for paths
A connected undirected multigraph G has an Euler path which is not an Euler circuit if and only if G has exactly two vertices of odd degree. The proof is very similar to the case of Euler circuits: just start the initial walk at one of the vertices of odd degree. The proof is thus similarly constructive, and yields an efcient algorithm to construct an Euler path, if one exists.
18 / 20
Hamiltonian Paths
Denition: A Hamiltonian path in a (undirected) graph G is a simple path that visits every vertex exactly once. (In other words, it is a tidy path that visits every vertex.) A Hamiltonian circuit in a (undirected) graph G is a simple circuit that passes through every vertex exactly once (except for the common start and end vertex, which is seen exactly twice). Question: Is there a simple criterion for determining whether a (simple undirected) graph has a Hamiltonian path, or Hamiltonian circuit?
19 / 20
Hamiltonian Paths
Denition: A Hamiltonian path in a (undirected) graph G is a simple path that visits every vertex exactly once. (In other words, it is a tidy path that visits every vertex.) A Hamiltonian circuit in a (undirected) graph G is a simple circuit that passes through every vertex exactly once (except for the common start and end vertex, which is seen exactly twice). Question: Is there a simple criterion for determining whether a (simple undirected) graph has a Hamiltonian path, or Hamiltonian circuit? Answer: No. Nobody knows any efcient algorithm for determining whether a given (arbitrary) graph G has a Hamiltonian path/circuit. The problem is NP-complete.
More on Hamiltonian paths/circuits

There are sufcient criteria that guarantee existence of a Hamiltonian circuit. For example:
Ores Theorem
Every simple undirected graph, G = (V , E ), with n 3 vertices, in which deg (u ) + deg (v ) n for every two non-adjacent vertices u and v in V , has a Hamiltonian circuit.
Corollary (Diracs Theorem)

Every simple undirected graph, G = (V , E ), with n 3 vertices, in which deg (u ) n/2 for all vertices u V , has a Hamiltonian circuit. We will NOT prove these theorems, and we will NOT assume that you know these theorems.

Graph Theory Notes

Uploaded by

Copyright:

Available Formats

Graph Theory Notes

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

Graph Theory Notes

Uploaded by

Copyright:

Available Formats

Discrete Mathematics & Mathematical Reasoning Chapter 10: Graphs

Kousha Etessami (U. of Edinburgh, UK)

Discrete Mathematics (Chapter 6)

Kousha Etessami (U. of Edinburgh, UK)

Discrete Mathematics (Chapter 6)

Kousha Etessami (U. of Edinburgh, UK)

Discrete Mathematics (Chapter 6)

A (simple undirected) graph:

A directed graph (digraph) (with loops):

A simple directed graph:

Graph Terminology Zoo (ridiculous)

differs from book.

Kousha Etessami (U. of Edinburgh, UK)

Formal Dention of Directed Graphs

Kousha Etessami (U. of Edinburgh, UK)

Discrete Mathematics (Chapter 6)

Denition of (Undirected) Graphs

Kousha Etessami (U. of Edinburgh, UK)

Discrete Mathematics (Chapter 6)

Graph Models: Computer Networks

diagnostic self-links at data centers: we use a pseudograph.

network with multiple one-way links we use a directed !multi"graph.

Graph Models: +ocial Networks

represent relationships between them.

1seful graph models of social networks include:

Graph Models: +ocial Networks

Example: An influence graph

by /ertices and links are represented by directed edges. *n a citation network:

from the /erte3 representing paper A to the /erte3 representing paper ..

&oad networks modeled using graphs

Degree and neighborhood of a vertex

4egrees and Neighborhoods of %ertices

4egree of %ertices !continued"

must be e/en since deg!v" is e/en for each v @ V5

4irected Graphs !continued"

+pecial (ypes of Graphs: Complete Graphs

+pecial (ypes of Graphs: Cycles

+pecial (ypes of +imple Graphs: n-Cubes

.ipartite Graphs !continued"

Complete .ipartite Graphs

.ipartite Graphs and Matchings

Kousha Etessami (U. of Edinburgh, UK)

Discrete Mathematics (Chapter 6)

Example of a Bipartite Graph

Kousha Etessami (U. of Edinburgh, UK)

Discrete Mathematics (Chapter 6)

Matchings in Bipartite Graphs

Kousha Etessami (U. of Edinburgh, UK)

Discrete Mathematics (Chapter 6)

Kousha Etessami (U. of Edinburgh, UK)

Discrete Mathematics (Chapter 6)

Halls Marriage Theorem

Kousha Etessami (U. of Edinburgh, UK)

Discrete Mathematics (Chapter 6)

proof of Halls Theorem, continued...

Kousha Etessami (U. of Edinburgh, UK)

Discrete Mathematics (Chapter 6)

proof of Halls Theorem (continued)

Kousha Etessami (U. of Edinburgh, UK)