DATA STRUCTURE AND

ALGORITHMS FIT324
Summer 2014
Lecture #12
Nguyen Xuan Thang
[email protected]
1
Pathf!"!#
! a #ra$h
2
Content
Pathfinding on a graph
Shortest path proble
!i"#stra algorith
$ellan%&or algorith
&loyd%'arshall algorith
3
Pathf!"!#%! a #ra$h
Pathfinding problem
(n a graph )*+,-./- find a path fro vertex s to
vertex t.
&ind a path
&ind all possible paths
)raph traversal algoriths can be used
BFS
DFS
4
v
1
(s)
v
2
v
4
v
3
v
5
(t)
Pathf!"!#u&!# '(S )1*
5
Algorithm BFS(V,E,s,t): Boolean
Input: A graph with V & E are the set of vertices and
edges, s is the source vertex, t is the destination vertex.
Output: A path from s to t if exist, otherwise return false.
For each v in V do
Color[v]Black
Path[v] -1
Color[ s]Yellow
Q new empty queue
Q.Enqueue(s)

Code
Pathf!"!#u&!# '(S )2*
6
while Q is not empty do
u Q.Dequeue()
If u equals to t then
showPath(t)
return true
Color[u] Red
For each v adjacent to u do
If Color[v] is Black then
Color[v] Yellow
Q.Enqueue(v)
path[v] s
endwhile
return false
Code
Pathf!"!#u&!# '(S )3*
&ind path fro 0 to !
Start at 0. Put 0 to the 1ueue +ar#ed 2ello3/
4ueue5 0
7
F
A
B
G C
E
D
0
0
0 0
0
0
0
Pathf!"!#u&!# '(S )4*
Ta#e 0 fro 4ueue- visit 0 +ar#ed 6ed/- 0#!
&ound t3o vertices $ and .- that are ad"acent to 0
Put $ 7 . to 1ueue- update path8$9 7 path8.9
4ueue5 0 $ .
8
A
B
G
C
E
D
F
0
A
0 0
A
0
0
Pathf!"!#u&!# '(S )+*
Ta#e $ fro 4ueue- visit $ +$#!/- put ) and : to
1ueue
;pdate path8)9 and path8:9
4ueue5 0 $ . ) :
9
A
B
G
C
E
D
F
0
A
B B
A
0
0
Pathf!"!#u&!# '(S ),*
Ta#e . fro 1ueue- visit . +.#!/. Put !7& to the
1ueue
;pdate path8!9 7 path8&9
4ueue5 0 $ . ) : ! &
10
A
B
G
C
E
D
F
0
A
B B
A
E
E
Pathf!"!#u&!# '(S )-*
Ta#e ) fro the 1ueue- visit )- nothing putted to the
1ueue.
Ta#e : fro the 1ueue- visit :. & is an ad"acent node to
:- but & is ar#ed as 2ello3 already.
'e 3on<t put & to the 1ueue.
4ueue5 0 $ . ) : ! &
11
A
B
G
C
E
D
F
0
A
B B
A
E
E
Pathf!"!#u&!# '(S ).*
Ta#e ! fro the 1ueue- visit !. ! is the destination
vertex. Sho3 path and return true.
4ueue5 0 $ . ) : ! &
12
A
B
G
C
E
D
F
0
A
B B
A
E
E
Pathf!"!#u&!# '(S )/*
13
Algorithm showPath(t)
Input: Destination vertex t.
Output: print the path from s to t if exist.
u t
while u !=0 do
Print u
u path[u]
Code
(!" m0 1%ffee #ame )1*
Introduction
Start fro a position in a ap
&ind the cup of coffee
14
(!" m0 1%ffee #ame )2*
Introduction
$&S is a solution
15
(!" m0 1%ffee #ame )3*
Data representation
The ap
;sing a atrix =xN += ro3s- N coluns/
16
0 1 2 3 4
0
G X X G G
1
Y G X G C
2
G G G G G
(!" m0 1%ffee #ame )4*
Data representation
The graph
.ach cell in the ap is a node of the graph. Nodes
are indexed fro > to =?N @ A.
'ith node v- the corresponding ro3 and colun are5
ro3 * v B NC colun*v D NC
17
0 1 2 3 4
0
0 1 2 3 4
1
5 6 7 8 9
2
10 11 12 13 14
(!" m0 1%ffee #ame )+*
Data representation
.ach node has a list of ad"acent nodes.
Node >5
EFG
Node F5
E>- H- A>G
Node I5
EJ- K- AKG
18
0 1 2 3 4
0
G X X G G
1
Y G X G C
2
G G G G G
0 1 2 3 4
0
0 1 2 3 4
1
5 6 7 8 9
2
10 11 12 13 14
(!" m0 1%ffee #ame ),*
erte! "D#
19
$erte!
% inde! & int ''Position in the (ertices list
% ro)* column & int ''+hich cell in the map
% mar,ed & boolean ''-sed in BFS
% num.f"d/erte! & int
% ad/erte! & int0 1 ''2ist of ad/acent (ertices
LLLLLLLLLLLLLLLLLLLLL
3 $erte!4inde!* map* m* n5 & (oid
3 getInde!45 & int
3 get6o)45 & int
3 getColumn45 & int
3 777
(!" m0 1%ffee #ame )-*
.ther (ariables
char8989 ap +The ap/
),ertex 89 list,ertex +0rray contains all =?N vertices/
),ertex start,ertex- end,ertex
2our position and position of the coffee cup.
0rray4ueue 1
,ertex 1ueue used in $&S +each 1ueue<s ite is a
vertex/
),ertex 89 path
Path inforation
20
(!" m0 1%ffee #ame ).*
Implementation
See 3ee#A2LtutorialLinstruction for the detail.
21
0 1 2 3 4 8 9 :
0 $ $ $ $ $ ; $ $
1 $ $ $ $ $ ; $ $
2 $ $ $ $ $ ; $ C
3 $ $ ; $ $ $ $ $
4 < $ ; $ $ $ $ $
(4,0) (3,0) (2,0) (2,1) (2,2) (2,3) (2,4)
(3,4) (3,5) (3,6) (2,6) (2,7)
Sh%rte&t $ath
22
Re1a22 a 3e#hte" #ra$h )1*
(n a 3eighted graph- each
edge has an associated
nuerical value- called
the 3eight of the edge.
.dge 3eights ay represent
distances- costs- etc.
.xaple5
(n a train route graph-
the 3eight of an edge
represents the distance
bet3een the endpoint
cities
23
Re1a22 a 3e#hte" #ra$h )2*
24
S#- F6" ="> C.P B?6 2?I
S#- 0 210 4@0
F6" 210 0 380 848 3A8
="> 380 0 4:8 2A0
C.P 4:8 0 438
B?6 848 2A0 438 0 1A0
2?I 4@0 3A8 1A0 0
w[i][i]=0
w[i][j]=w[j][i] if G is un-direced
w[i][j]= if !ere is n" ed#e
c"nnec $ere% i " $ere% j
Sh%rte&t $ath $r%42em )1*
Definition
(n a graph $B4*?5- consider to a path from s to t.
P
,
BC
1

2

3
D
,
E.
Let +4P
,
5 is the )eight of the path P
#
3e have5
(f ) is an un%)eighted graph- then5
+4P
,
5*The nuber of vertices in P
#
*,
(f ) is a )eighted graph- then5
w P
k
= w(I

, I
+1
)
k-1
=1
&ind a P
#
so that +4P
,
5 is minimum.
25
Sh%rte&t $ath $r%42em )2*
Shortest path properties
0 sub%path of a shortest path is itself a shortest
path.
&or exaple5
(f P
sp
*E,
A
,
M
,
H
,
K
,
F
,
N
G is the shortest
path fro ,
A
to ,
N
- then the path E,
M
,
H
,
K
G is
the shortest path fro ,
M
to ,
K
.
There is a tree of shortest path fro a vertex to all
other vertices that connect to this vertex.
26
Sh%rte&t $ath $r%42em )3*
Shortest path in an un%3eighted graph
BFS is a solution
Shortest path in a 3eighted graph
.very edge has positi(e )eight
.dge ay have negati(e )eight
)raph ay have negati(e )eight cFcle
27
A B s d
-3
1
8
W(s, d)=6
A B s d
-3
1
8
W(s, d)= -
1
Sh%rte&t $ath $r%42em )4*
Shortest path algorith types5
Single%pair shortest path +SPSP/
&ind shortest paths fro a given vertex to a given
vertex.
Single%source shortest path 4SSSP5
&ind shortest paths fro a given vertex to all other
vertices.
"ll%pairs shortest path 4"PSP5
&ind shortest paths for every pair of vertices.
28
D56&tra a2#%rthm )1*
0 SSSP algorithm
&ind shortest paths fro (erte! s to all other
vertices.
"ssumption
The graph is connected.
The edge )eights are non%negati(e.
!efinition of the distance fro ( to s
!8v9 is the total 3eight of the shortest path fro s to
v.
'ith a given vertex s- !i"#stra algorith 3ill
copute the !8v9 for all v in ,.
29
D56&tra a2#%rthm )2*
Di/,stra algorithm idea
'e gro3 a GcloudH of (ertices- beginning 3ith s
and eventually covering all the vertices.
'e store 3ith each vertex ( a label d4(5
representing the distance of ( from s in the sub%
graph consisting of the cloud and its ad"acent
vertices
0t each step5
'e add to the cloud the vertex u outside the cloud
3ith the sallest distance label- d+u/
'e update the labels of the vertices ad"acent to u
30
D56&tra a2#%rthm )3*
-pdate 4?dge rela!ation5
:onsider an edge
e B +u*I/ such that5
u is the vertex ost
recently added to the cloud
I is not in the cloud
The relaxation of edge e
updates distance d+I/
as follo3s5
d+I/ inEd+I/*d+u/ O )eight+u*I/G
31
d(z) = = = = 75
d(u) = = = = 50
z
s
u
d(z) = = = = 60
d(u) = = = = 50
z
s
u
e
e
D56&tra a2#%rthm )4*
?!ample
32
C B
A
E
D
F
4
8
7
2 5
2
3 9
1
C B
A
E
D
F
4
8
7
2 5
2
3 9
1
C B
A
E
D
F
4
8
7
2 5
2
3 9
1
0
8 2 4

C B
A
E
D
F
4
8
7
2 5
2
3 9
1
0
8 2 3
5
11
D56&tra a2#%rthm )+*
?!ample
33
C B
A
E
D
F
4
8
7
2 5
2
3 9
1
C B
A
E
D
F
4
8
7
2 5
2
3 9
1
0
7 2 3
5
8
0
7 2 3
5
8
C B
A
E
D
F
4
8
7
2 5
2
3 9
1
0
8 2 3
5
8
C B
A
E
D
F
4
8
7
2 5
2
3 9
1
0
7 2 3
5
8
D56&tra a2#%rthm ),*
Di/,stra algorithm
34
Algorithm Dijkstra(V,E,s):
Input: A graph G=(V,E), w is the weighted matrix, s is the
source vertex.
Output: The shortest paths from s to other vertices.
For each v in V do
D[v]
path[v] -1
D[s] 0
Create empty S set //The cloud set
Do
Find vertex u in V \ S so that D[u] is minimum
Move u from V to S
For each z in V \ S do
D[z] min{D[z], D[u]+w[u][z]}
path[z] u //Only update if D[u]+w[u][z]<D[z]
While could not find any u
Code
D56&tra a2#%rthm )-*
Di/,stra algorithm demonstration
35
s
3
8
2
6
7
4
5
24
18
2
9
14
15
5
30
20
44
16
11
6
19
6







0
distance D
& = ' (
) = ' s* 2* 3* 4* 5* 6* 7* 8(
+ = '0* * , , , , , (
D56&tra a2#%rthm ).*
Di/,stra algorithm demonstration
36
s
3
8
2
6
7
4
5
24
18
2
9
14
15
5
30
20
44
16
11
6
19
6

15
9



14
0
& = 's(
) = ' -* 2* 3* 4* 5* 6* 7* 8(
+ = '0* 9* , , , 14, 15, (



D56&tra a2#%rthm )/*
Di/,stra algorithm demonstration
37
s
3
8
2
6
7
4
5
24
18
2
9
14
15
5
30
20
44
16
11
6
19
6

15
9
33


14
0
& = 's* 2(
) = ' -* -* 3* 4* 5* 6* 7* 8(
+ = '0* 9* 33, , , 14, 15, (



D56&tra a2#%rthm )10*
Di/,stra algorithm demonstration
38
s
3
8
2
6
7
4
5
24
18
2
9
14
15
5
30
20
44
16
11
6
19
6

15
9
32

44
14
0
& = 's* 2* 6(
) = ' -* -* 3* 4* 5* -* 7* 8(
+ = '0* 9* 32, , 44, 14, 15, , (


33



D56&tra a2#%rthm )11*
Di/,stra algorithm demonstration
39
s
3
8
2
6
7
4
5
24
18
2
9
14
15
5
30
20
44
16
11
6
19
6

15
9
32
59
35
14
0
& = 's* 2* 6* 7(
) = ' -* -* 3* 4* 5* -* -* 8(
+ = '0* 9* 32, , 35, 14, 15, 59(


33



44

D56&tra a2#%rthm )12*
Di/,stra algorithm demonstration
40
s
3
8
2
6
7
4
5
24
18
2
9
14
15
5
30
20
44
16
11
6
19
6

15
9
32
51
34
14
0
& = 's* 2* 6* 7* 3(
) = ' -* -* -* 4* 5* -* -* 8(
+ = '0* 9* 32, , 34, 14, 15, 51(


33



44

35
59
D56&tra a2#%rthm )13*
Di/,stra algorithm demonstration
41
s
3
8
2
6
7
4
5
24
18
2
9
14
15
5
30
20
44
16
11
6
19
6
45
15
9
32
50
34
14
0
& = 's* 2* 6* 7* 3* 5(
) = ' -* -* -* 4* -* -* -* 8(
+ = '0* 9* 32, 45, 34, 14, 15, 50(


33



44

35
59

51
D56&tra a2#%rthm )14*
Di/,stra algorithm demonstration
42
s
3
8
2
6
7
4
5
24
18
2
9
14
15
5
30
20
44
16
11
6
19
6
45
15
9
32
50
34
14
0
& = 's* 2* 6* 7* 3* 5* 4(
) = ' -* -* -* -* -* -* -* 8(
+ = '0* 9* 32, 45, 34, 14, 15, 50(


33



44

35
59

51
D56&tra a2#%rthm )1+*
Di/,stra algorithm demonstration
43
s
3
8
2
6
7
4
5
24
18
2
9
14
15
5
30
20
44
16
11
6
19
6
45
15
9
32
50
34
14
0
& = 's* 2* 6* 7* 3* 5* 4* 8(
) = ' -* -* -* -* -* -* -* -(
+ = '0* 9* 32, 45, 34, 14, 15, 50(


33



44

35
59

51
D56&tra a2#%rthm )1,*
Di/,stra algorithm analFsis
Textboo# page 3:A- suppleentary textboo# page
98@.
The tie coplexity of !i"#stra algorith is5
.4J?J 3 JJ
2
5 bF using arraF to store and D0(17
.4J?Jlog 3 JJlog5 bF using a prioritF Kueue to store
and D0(17
!i"#stra algorith doesn<t 3or# 3ith negative%
3eighted edge.
)raph can be directed or un%directed.
44
'e22ma!7(%r" a2#%rthm )1*
0 SSSP algorithm
&ind shortest paths fro (erte! s to all other
vertices.
'or#s even 3ith negati(e%)eight edges.
:an detect the existence of negati(e%)eight cFcle
reachable fro s.
"ssumption
The graph is connected.
The edges are directed.
45
'e22ma!7(%r" a2#%rthm )2*
!efinition of D0i10(15
!8i98v9 is the total 3eight of the shortest path that
use i or fe)er edges- fro s to (.
6ecall that the distance fro vertex u to vertex v is
the total 3eight of the shortest path fro u to v.
Then !8i98v9 is the distance fro s to ( using i or
fe)er edges.
!8i98s9*>
!8i98v9*if you can<t get to v 3ithin i edges.
$ellan%&ord algorith computes the D0i10(1 for
anF ( in and iB1 77 JJ.
46
'e22ma!7(%r" a2#%rthm )3*
"lgorithm idea
To copute !8i98v9- 3e calculate i
th
ro) from i%1
th
ro).
:onsider a vertex u so that there is an edge fro u
to v.
'e #no3 the shortest path fro s to u using i%A or
fe3er edges- !8i%A98u9.
Then5
D | u = mtn
w u |u]=
{D | 1 u +w|u]|u]]
47
'e22ma!7(%r" a2#%rthm )4*
?!ample5
48
4
S
3
1
5
2
30
-20
10
50
30
15
30
& 1 2 3 4 5
0 0
1 0 50 15
2 0 50 80 15 45
3 0 25 80 15 45 75
4 0 25 55 15 45 75
5 0 25 55 15 45 65
D | u = mtn
w u |u]=
{D | 1 u + w|u]|u]]
'e22ma!7(%r" a2#%rthm )+*
Bellman%Ford algorithm
49
Algorithm BellmanFord(V,E,s): boolean
Input: A graph G=(V,E), w is the weighted matrix, s is the
source vertex.
Output: The shortest paths from s to other vertices or
false if G has a negative-weighted cycle.
For i 0 to |V| do
For each v in V do
D[i][v]
path[v] -1
D[i][s] 0
For i 1 to |V| do
For each v in V do
For each e=(u,v) in E do
D[i][v] Min{D[i-1][u]+w[u][v]}
path[v] u
min
// D[i][v]=D[i-1][u
min
]+w[u
min
][v]
Code
'e22ma!7(%r" a2#%rthm ),*
"lgorithm analFsis
(f the graph contains no negative%3eight cycles
reachable fro the source vertex s- after JJ % 1
iterations all distance estiates represent shortest
paths.
(f the graph contains a negative%3eight cycles
reachable fro s.
50
4
S
3
1
5
2
30
-100
10
50
30
15
30
'e22ma!7(%r" a2#%rthm )-*
"lgorithm analFsis
51
4
S
3
1
5
2
30
-100
10
50
30
15
30
& 1 2 3 4 5
0 0
1 0 50 15
2 0 50 80 15 45
3 0 -55 80 15 45 75
4 -5 -55 -25 15 45 75
5 -5 -55 -25 10 45 -15
'e22ma!7(%r" a2#%rthm ).*
Bellman%Ford algorithm
52
Algorithm BellmanFord(V,E,s):
Input: A graph G=(V,E), w is the weighted matrix, s is the
source vertex.
Output: The shortest paths from s to other vertices or
false if G has a negative-weighted cycle.

For i 1 to |V| do
For each v in V do
For each e=(u,v) in E do
If D[i][v] > D[i][u] + w[u][v] then
return false //There is a negative-weighted cycle
return true
Code
'e22ma!7(%r" a2#%rthm )/*
"lgorithm analFsis
Tie coplexity
.4JJ7J?J5
53
(2%0"78ar&ha22&a2#%rthm )1*
0 "PSP 4all%pairs shortest path5 algorithm
&ind shortest paths fro e(erF pair of (ertices u
i
* u
/
+i-"*A..P,P and i # "/
"ssumption
The graph is connected.
The edges are non%negati(e )eighted.
54
(2%0"78ar&ha22&a2#%rthm )2*
"lgorithm idea
:onstruct a atrix D
4,5
0i10/1 that is the 3eight of the
shortest path fro u
i
to u
/
3ith all interediate
vertices in an initial subset
,
BCu
1
* u
2
* D u
,
E
55
u
i
u
j
p
V
k
(2%0"78ar&ha22&a2#%rthm )3*
"lgorithm idea
D
405
0i10/1 is the direct%path from u
i
to u
/
D
405
0i10/1 B )0i10/1
D
415
0i10/1 is the shortest path fro u
i
to u
/
through
one intermediate vertex.
D
415
0i10/1BminCD
405
0i10/1*D
405
0i10,13D
405
0,10/1E
for each , in JJ
D4,50i10/1 is the shortest path fro u
i
to u
/
through ,
intermediate vertices.
D
4,5
0i10/1BminCD
4,%15
0i10/1*D
4,%15
0i10,13D
4,%15
0,10/1E for
each , in JJ
56
(2%0"78ar&ha22&a2#%rthm )4*
FloFd%+arshalls algorithm
57
Algorithm FloydWarshalls(V,E):
Input: A graph G=(V,E), w is the weighted matrix.
Output: The shortest paths of any pair u,v vertices in G.
For (i,j 1 to |V| ) do
!8i98"9*38i98"9
path8i98"9*%A
For (k 1 to |V|)
For (i,j 1 to |V|)
!8i98"9 inE!8i98"9- !8i98#9O!8#98"9G
path8i98"9 # BBQnly update if
BB !8i98"9*!8i98#9O!8#98"9
Code
Tut%ra2 9 !e:t t%$1
Preparing for the tutorial&
Practice 3ith exaples and exercises in
3ee#A2LtutorialLinstruction
Preparing for ne!t topic&
:ourse revie3
58
59
Any
Questions?