Design and Analysis of Algorithms

Shortest Path
Dr. Lê Nguyên Khôi
The VNU University of Engineering and Technology
Contents

 Path definition
 Shortest path properties
 Single-source shortest path
 All-pairs shortest path

Design and Analysis of Algorithms 1

Path – Definition
Graph
 Path
 a sequence of vertices
 2 consecutive vertices , connected by
an edge in
 is a path from to
 Cycle
 a path with
 whereas, the first vertices are distinct and

Design and Analysis of Algorithms 2

Path – Weight
Consider a directed graph with
edge-weight function . The weight
of path is defined as:

Example:

Design and Analysis of Algorithms 3

Shortest Path – Definition

A shortest path from to is a path of

minimum weight from to .
The shortest-path weight from to is
defined as:

Note: if no path from to

exists

Design and Analysis of Algorithms 4

Shortest Path – Properties

Optimal substructure
 A sub-path of a shortest path is the
shortest path
 is the shortest path
from to
 is the sub-path of
from to với
 is the shortest path from to

Design and Analysis of Algorithms 5

Shortest Path – Properties

Optimal substructure – Proof

 Divide into sub-paths
 Then,
 Consider path from to , with weight

 Then, is a path from to

with weight smaller than
 Contradiction with the hypothesis is the
shortest path from to

Design and Analysis of Algorithms 6

Shortest Path – Properties

Triangle inequality
 For all vertices , we have

Design and Analysis of Algorithms 7

Shortest Path – Properties

Negative weight
 If a graph contains a negative-weight cycle,
then some shortest paths may not exist

Design and Analysis of Algorithms 8

Single-Source Shortest Path

 Problem. From a given source vertex

, find the shortest-path weights
for all
 If all edge weights are
nonnegative, all shortest-path weights
must exist
 Greedy strategy to find the optimal
solution
 As shortest path has the optimal substructure
property

Design and Analysis of Algorithms 9

Single-Source Shortest Path

 Greedy strategy:
1. Maintain a set of vertices whose shortest-
path distances from are known
2. At each step add to the vertex
whose distance estimate from is minimal
3. Update the distance estimates of vertices
adjacent to

Design and Analysis of Algorithms 10

Greedy Strategy

Initialization
0
2 9

3 5 2

1 4 1

1 4
8

Design and Analysis of Algorithms 11

Greedy Strategy

Add to
0
2 9

3 5 2

1 4 1

1 4
8

Design and Analysis of Algorithms 12

Greedy Strategy

Add to
0
2 9

3 5 2

1 4 1

1 4
8

Design and Analysis of Algorithms 13

Greedy Strategy

Add to
0
2 9

3 5 2

1 4 1

1 4
8

Design and Analysis of Algorithms 14

Greedy Strategy

Add to
0
2 9

3 5 2

1 4 1

1 4
8

store the weight of the shortest path

from to for all
Design and Analysis of Algorithms 15
Dijkstra’s Algorithm
 Consider special path from to only
consist of vertices in
 : store the path weight from to
 Initialization , ,
with

Design and Analysis of Algorithms 16

Dijkstra’s Algorithm

: store the path weight from to

 Initialization , ,
with
 Each step:
 Choose vertex such that minimal
 Add to , then store the weight of the shortest
path from to
 Then, recalculate for all

 Repeat, until contains all vertices in

Design and Analysis of Algorithms 17

Dijkstra’s Algorithm – Exercise

Design and Analysis of Algorithms 18

Dijkstra’s Algorithm – Pseudocode

for each do

while do times

for each do times

if then

Design and Analysis of Algorithms 19

Unweighted Graph
Suppose that for all .
Can Dijkstra’s algorithm be improved?
 Use a simple FIFO queue instead of a priority queue
Breadth First Search
while do

for each do
if then

Complexity Analysis:
Design and Analysis of Algorithms 20
Breadth First Search

Design and Analysis of Algorithms 21

All-Pairs Shortest Path

Single-Source Shortest Path

 Dijkstra’s algorithm
All-Pairs Shortest Path
 Dijkstra’s Algorithms times
Greedy strategy
 Floyd-Warshall’s Algorithms
Dynamic programming

Design and Analysis of Algorithms 22

Floyd-Warshall’s Algorithm
, set empty,
15
1 4
5 1 2 3 4

5 1 0 5 ∞ ∞
50 30 15
5
2 50 0 15 5
3 30 ∞ 0 15
15
2 3 4 15 ∞ 5 0

Design and Analysis of Algorithms 23

Floyd-Warshall’s Algorithm
1 2 3 4
1 0 5 ∞ ∞

15 2 50 0 15 5
1 4 3 30 ∞ 0 15
5 4 15 ∞ 5 0
5 15
5 50 30
1 2 3 4
1 0 5 ∞ ∞
15
2 3 2 50 0 15 5
3 30 35 0 15
4 15 20 5 0

Design and Analysis of Algorithms 24

Floyd-Warshall’s Algorithm
1 2 3 4
1 0 5 ∞ ∞

15 2 50 0 15 5
1 4 3 30 35 0 15
5 4 15 20 5 0
5 15
5 50 30
1 2 3 4
1 0 5 20 10
15
2 3 2 50 0 15 5
3 30 35 0 15
4 15 20 5 0

Design and Analysis of Algorithms 25

Floyd-Warshall’s Algorithm
1 2 3 4
1 0 5 20 10

15 2 50 0 15 5
1 4 3 30 35 0 15
5 4 15 20 5 0
5 15
5 50 30
1 2 3 4
1 0 5 20 10
15
2 3 2 45 0 15 5
3 30 35 0 15
4 15 20 5 0

Design and Analysis of Algorithms 26

Floyd-Warshall’s Algorithm
1 2 3 4
1 0 5 20 10

15 2 45 0 15 5
1 4 3 30 35 0 15
5 4 15 20 5 0
5 15
5 50 30
1 2 3 4
1 0 5 15 10
15
2 3 2 20 0 10 5
3 30 35 0 15
4 15 20 5 0

Design and Analysis of Algorithms 27

Floyd-Warshall’s Algorithm

Define weight of a shortest path

from to with intermediate vertices
belong to the set

Thus,
Initialization

Design and Analysis of Algorithms 28

Floyd-Warshall’s Algorithm
 If vertex is on the shortest path from to , then the
path from to and the path from to is the shortest
path
 If is the weight of the path not containing , then
this path only consist of vertices in , thus

 If is the weight of the path not containing , then

the sub-path from to has weight , and the
sub-path from to has weight
 Thus

Design and Analysis of Algorithms 29

Floyd-Warshall Algorithm – Pseudocode

for to do
for to do
for to do
if then

 Run in time
 Simple to code
 Efficient in practice
Design and Analysis of Algorithms 30