Graph Algorithms – Part 4

CPT212 – Design & Analysis of

Algorithms

Teh Je Sen (2019)

Abllah Zawawi Talib (2023) 1
Learning Objectives
 In this topic, we will learn about
algorithms that can be used to find the
following:
 Shortest path
 Spanning trees
 Note:
 : Number of vertices
 : Number of edges

2
SHORTEST PATH

3
Scenario
 So far, we have been able to identify if
two vertices can reach other.
 What if we want to find the shortest
path between two vertices?
 Example:
 Find the fastest way to travel from one city
to another.

4
Weighted Graph
 The shortest path problem is applicable
to weighted graphs.
Each edge is associated
 Example: to a weight.

849 PVD
1843 ORD 2
SFO 14

802

1205
43 LGA
17
337

7
HNL 2555 138 10
99
1233
LAX DFW 1120
MIA
5
Shortest Path Problem
 Given a weighted graph and two
vertices, and , find a path with
minimal weight between and .
 Weights can refer to distance, cost,
time, etc.
 Applications:
 Internet packet routing
 Flight reservations
 Driving directions 6
Shortest Path Properties
 Property 1: A subpath of a shortest
path is itself a shortest path

849 PVD
1843 ORD 2
SFO 14

802

1205
43 LGA
17
337

7
HNL 2555 138 10
99
1233
LAX DFW 1120
MIA
7
Verify of the highlighted paths
are the shortest paths
 Shortest path from PVD to HNL
 Shortest path from PVD to LAX
 Shortest path from PVD to ORD
849 PVD
1843 ORD 2
SFO 14

802

1205
43 LGA
17
337

7
HNL 2555 138 10
99
1233
LAX DFW 1120
MIA
8
Shortest Path Properties
 Property 2: There is a tree of shortest
paths from a start vertex to all the
other vertices
 Tree of shortest paths with PVD as
starting vertex:
849 PVD
1843 ORD 2
SFO 802 14

1205
7 43 LGA
337

1 87 10
2555 13
HNL 1233 99
LAX DFW 1120
MIA 9
Algorithms to Find the
Shortest Path
 Dijkstra’s Algorithm
 A greedy method
 Cannot handle negative edges
 Bellman-Ford Algorithm
 Slower than Djikstra
 Can handle negative edges

10
Dijkstra’s Algorithm
 Can be seen as a “weighted” BFS
starting from a source vertex,
 Iteratively grow a “cloud” of vertices
out of
 In each iteration, the closest vertex to the
cloud is chosen to be included
 Algorithm terminates when there are no
more vertices outside the cloud, or the
vertices are not connected to the cloud
11
Dijkstra’s Algorithm
 Notations:
 For every vertex, we assign a label
 will store the length of the best path from
at the time (will be updated with each
iteration)
 At each iteration, a vertex with the
smallest will be pulled into the cloud
 Then update the label of every vertex
adjacent to and is outside the cloud (edge
relaxation) 12
Dijkstra’s Algorithm
Psuedocode

13
Dijkstra’s Algorithm Simple
Example (Undirected Graph)

𝐷[𝑎]=0
a
9 2

b 6
c
𝐷[ [𝑐𝑐]=∞
𝐷 ]=2
𝐷 [𝑏
𝐷 ] =∞
𝐶] =∞
=9
4 1

d
𝐷 [ 𝑑 ] =∞

14
 Dijkstra’s Algorithm Simple
Example (Undirected Graph)

𝐷[𝑎]=0
a
9 2

b 6
c
𝐷[ [𝑐𝑐]=∞
𝐷 ]=2
𝐷[[𝑏𝐶] =8
𝐷 ] =∞
=∞
=9
4 1
has been removed from
Complete the remaining
the queue and placed into
iterations on your own! d
the cloud.
remain in the queue 𝐷[[𝑑𝑑] =∞
𝐷 =3

15
Task
 Perform Dijkstra’s Algorithm on the
digraph shown below:

b 1 d
10
2 3 9
a 4 6
source
5 7
c 2 e
16
Analysis of Complexity
 Dijkstra’s algorithm uses a priority queue
 The algorithm involves at most:
 insertions into Q
 calls for removing the minimum value
 calls for changing the key value (weight)
 The above operations run in [while loop, ∑ v∈G log n
removal from Q, ∑ v∈G deg(v) log n relaxation]
 So what is the worst case complexity for
Dijkstra’s algorithm?
 because worst case, the graph is fully connected,

17
Negative Edges
 Dijkstra’s algorithm does not work when
negative edges exist because it is based
on the greedy method
 Vertices are added based on increasing
distance
 A negative edge will mess up the distances

18
Negative Edges
 Example – Try Djikstra’s algorithm for
the following graph from source, :

8 A 4
6
7 1
B C D
0 -8
2 5
E F

19
Bellman-Ford Algorithm
 Iteratively update the distance from the
source by checking each edge in the
graph
 Algorithm continues iterating as long as
there is an edge such that

20
Bellman-Ford Algorithm
 Pseudocode:
Algorithm ShortestPath(G,s)
Input: Weighted simple digraph, and a source vertex,
Output: Length shortest paths from to each vertex of
for all vertices

while there is an edge() such that

> + weight(edge())
do
= + weight(edge())

What is the worst case complexity of this algorithm?

For every vertex, you need to check all edges.
21
Bellman-Ford Example
a 2 b (a,2)
c 5 (b,7)

2 2 -3 9

d 4 e f 5
(a,2) (b,-1)
(g,-2)
(a,2) (e,4)
(e,3)
1g
-5 iteration: init 1 2
a 0
(d,3)
b  2
c  7
edges: ab ad ae bc be ce de dg ef ge d 2

e  2 -1 -2
f  4 3
g  3
22
SPANNING TREES

23
Re-cap
 Spanning tree
 A connected sub-graph that includes all
vertices from the original graph and is
also a tree
 A tree is a sub-graph that has no
cycles

24
Scenario
 Bob wants to connect all computers in
his office building using the least
amount of cable
 This problem can be represented as an
undirected weighted graph
 Rather than computing the shortest
path from some vertex, Bob needs to
find a tree that contains all vertices
and has a minimum total weight
25
Minimum Spanning Tree
(MST)
 Spanning tree with minimum weight

6 6 10
10

1
7 1
20

5 5

26
Algorithms for MST Problem
 Both algorithms based on Greedy
method
 Prim-Jarnik algorithm
 Grows MST from a single root vertex
 Kruskal’s algorithm
 Grows MST in clusters

27
MST Properties
 Properties that are related to MSTs:
 Light edge property
 Cycle property
 Partition property

28
Light Edge Property
 The Greedy method for MST relies on the light-edge
property.
 If a connected undirected weighted graph is divided
into two subsets, the minimum-weight edge
that connects the subset is included in the MST of
Any minimum-weight “bridge” edge can
be immediately added into the MST.

29
Other MST Properties
 Cycle Property
 Let be a MST of a weighted graph,
 Let be an edge of that is not in
 Let be a cycle formed by when included
in
f 8 For all edges, in
4
𝑻 C 9 Else, we can get a spanning
2 6
3 e tree with smaller weight by
8 7 replacing with
7
30
Other MST PropertiesU V
f 7
4
 Partition property 2 5
9
8
 If a “bridge” edge 8 e 3
has the same 7
weight as another Replacing f with e yields
“bridge” edge , another MST

replacing with U
f 7
V

will yield another 4

9
MST 2 5
8
8 e 3
7
31
Prim-Jarnik Algorithm
 Similar to Dijkstra’s algorithm
 Grow a MST starting from a “root”
vertex,
 An initial “cloud” will be defined which
includes
 Each iteration, a minimum-weight edge
will be brought into the cloud
 By choosing the minimum-weight edge
each time, a valid edge is added into the
MST (light edge property)
32
Prim-Jarnik Algorithm

What is the difference

between Prim-Jarnik and
Dijkstra?

Update the value for each

vertex with minimum
distance from the tree,
rather than distance from
source vertex.
33
 Example
D[D]= D[D]=7
D[B]=2 7 D D[B]=2 7 D
B 4 B 4
D[C]=8 D[F]= D[F]=
9 D[C]=5 9
2 5 F 2 5 F
C C
8 8
8 3 8 3
E E
A A
D[A]=0
7 D[E]=7 7 D[E]=7
D[A]=0
D[D]=7
D[D]=7
D[B]=2 7 D D
D[B]=2 7
B 4 B
D[C]=5 D[F]= 4
9 D[C]=5 9
D[F]=4
2 5 F 5
C 2 F
8 C
3 8
8 8 3
E E
A D[E]=7
A
D[A]=0
7 7 D[E]=7
D[A]=0

34
 Example
D[D]=7
D[B]=2 7 D
B 4
D[F]=4
D[C]=5 9
2 5 F
C
8
8 3
E
A D[D]=7
D[A]=0 7 D[E]=3
D[B]=2 7 D
B 4
D[F]=4
D[C]=5 9
2 5 F
C
8
8 3
E
A
D[A]=0 7 D[E]=3

35
Analysis of Complexity
 Complexity of Prim-Jarnik is the same
as Dijkstra’s algorithm
 Involves insertion (n), removal of
minimum value (n), updating values of
priority queue (m), each requiring
 because the graph is when a graph is
connected,

36
Kruskal’s Algorithm
 Basic concept:
 Maintains multiple smaller trees in a forest
 Repeatedly merge until there is a single
tree
 Steps:
 Each vertex starts in its own cluster
 Consider each edge ordered by increasing
weight
 If an edge connects two clusters, add the edge
to the MST
 If the edge is within a cluster, discard it
37
Kruskal’s Algorithm

38
Kruskal’s Algorithm
a b4 c 9
5 6 8 13
7
d e 8
6
7
f

sorted edges: ab ad bd df ce ef de bf bc cd

4
a b c
6
5
7
has 5 edges (n-1 edges) d e
6
𝑀𝑆𝑇 𝑐𝑜𝑚𝑝𝑙𝑒𝑡𝑒 7
f 39
Example
G G
B 8 8
4 B 4
9 E E
5 6 9 6
1 F 1 5 F
C 11 3 C 11 3
2 2
7 D H 7 H
D
A 10 A 10

G G
B 8 B 8
4 4
9 E 9 E
5 6 5 6
1 F 1 F
C 11 3 C 11 3
2 2
7 D H 7 H
D
A 10 A 10
40
Example
G G
B 8 B 8
4 4
9 E 9 E
5 6 6
1 F 1 5 F
C 11 3 C 11 3
2 2
7 D H 7 H
D
A 10 A 10
four steps

s
ep
st
G o G
8 8
tw
B 4 B 4
9 E 9 E
5 6 5 6
1 F 1 F
C 11 3 C 11 3
2 2
7 D H 7 D H
A 10 A 10
41
Analysis of Complexity
 Ordering of edges =
 Remove minimum value for each run is
constant
 For connected graph,
 Thus, the complexity of Kruskal’s
algorithm is equivalent to
 How is this calculated?
 m log n^2 = 2m log n = O(m log n)
42
Task
 Perform Kruskal’s algorithm on the
graph below to find the MST

43
End of Part 4

44