NPTEL MOOC,JAN-FEB 2015

Week 3, Module 3

DESIGN AND ANALYSIS

OF ALGORITHMS
Breadth first search (BFS)

MADHAVAN MUKUND, CHENNAI MATHEMATICAL INSTITUTE

http://www.cmi.ac.in/~madhavan
Graphs, formally
G = (V,E)

Set of vertices V

Set of edges E

E is a subset of pairs (v,v’): E ⊆ V × V

Undirected graph: (v,v’) and (v’,v) are the same edge

Directed graph:

(v,v’) is an edge from v to v’

Does not guarantee that (v’,v) is also an edge

Finding a
route
Find a v0
sequence of
vertices v0, v1,
…, vk such that

v0 is source
v1
Each (vi,vi+1) v3
is an edge in
E
v2 v4
vk is target
v5
 Adjacency matrix
1
1 2 3 4 5 6 7 8 9 10
3 1 0 1 1 1 0 0 0 0 0 0
2 1 0 1 0 0 0 0 0 0 0
2 5 3 1 1 0 0 0 0 0 0 0 0
4 1 0 0 0 1 0 0 1 0 0
4 5 0 0 0 1 0 1 1 0 0 0
6 7 6 0 0 0 0 1 0 1 1 1 0
7 0 0 0 0 1 1 0 0 0 0
8 0 0 0 1 0 1 0 0 1 0
8 9 9 0 0 0 0 0 1 0 1 0 1
10 0 0 0 0 0 0 0 0 1 0
10
 Adjacency list
For each vertex, maintain a
1 list of its neighbours

3 1 2,3,4
2 1,3
2 5 3 1,2
4 1,5,8
4 5 4,6,7
6 7 6 5,7,8,9
7 5,6
8 9 8 4,6,9
9 6,8,10
10 10 9
Finding a path

Mark vertices that have been visited

Keep track of vertices whose neighbours have

already been explored

Avoid going round indefinitely in circles

Two fundamental strategies: breadth first and

depth first
Breadth first search
Explore the graph level by level

First visit vertices one step away

Then two steps away

Remember which vertices have been visited

Also keep track of vertices visited, but whose

neighbours are yet to be explored
Breadth first search

Recall that V = {1,2,…,n}

Array visited[i] records whether i has been visited

When a vertex is visited for the first time, add it to

a queue

Explore vertices in the order they reach the

queue
Breadth first search
Exploring a vertex i:

for each edge (i,j)

if visited[j] == 0
visited[j] = 1
append j to queue

Initially, queue contains only source vertex

At each stage, explore vertex at the head of the

queue

Stop when the queue becomes empty

 Breadth first search 1
2
1 3
4
5
3 Visited
6
2 5 7
8
4 9
6 7 10

Queue
8 9

10
 Breadth first search 1
2
1 3
4
5
3 Visited
6
2 5 7
8
4 9
6 7 10

Queue
8 9

10 head tail
 Breadth first search 1
2
1

1 3
4
5
3 Visited
6
2 5 7
8
4 9
6 7 10

Queue
8 9
1
10
 Breadth first search 1
2
1

1 3
4
5
3 Visited
6
2 5 7
8
4 9
6 7 10

Queue
8 9

10
 Breadth first search 1
2
1
1
1 3
4
5
3 Visited
6
2 5 7
8
4 9
6 7 10

Queue
8 9
2
10
 Breadth first search 1
2
1
1
1 3 1
4
5
3 Visited
6
2 5 7
8
4 9
6 7 10

Queue
8 9
2 3
10
 Breadth first search 1
2
1
1
1 3 1
4 1
5
3 Visited
6
2 5 7
8
4 9
6 7 10

Queue
8 9
2 3 4
10
 Breadth first search 1
2
1
1
1 3 1
4 1
5
3 Visited
6
2 5 7
8
4 9
6 7 10

Queue
8 9
3 4
10
 Breadth first search 1
2
1
1
1 3 1
4 1
5
3 Visited
6
2 5 7
8
4 9
6 7 10

Queue
8 9
4
10
 Breadth first search 1
2
1
1
1 3 1
4 1
5
3 Visited
6
2 5 7
8
4 9
6 7 10

Queue
8 9

10
 Breadth first search 1
2
1
1
1 3 1
4 1
3 Visited 5 1
6
2 5 7
8
4 9
6 7 10

Queue
8 9
5
10
 Breadth first search 1
2
1
1
1 3 1
4 1
3 Visited 5 1
6
2 5 7
8 1
4 9
6 7 10

Queue
8 9
5 8
10
 Breadth first search 1
2
1
1
1 3 1
4 1
3 Visited 5 1
6
2 5 7
8 1
4 9
6 7 10

Queue
8 9
8
10
 Breadth first search 1
2
1
1
1 3 1
4 1
3 Visited 5 1
6 1
2 5 7
8 1
4 9
6 7 10

Queue
8 9
8 6
10
 Breadth first search 1
2
1
1
1 3 1
4 1
3 Visited 5 1
6 1
2 5 7 1
8 1
4 9
6 7 10

Queue
8 9
8 6 7
10
 Breadth first search 1
2
1
1
1 3 1
4 1
3 Visited 5 1
6 1
2 5 7 1
8 1
4 9
6 7 10

Queue
8 9
6 7
10
 Breadth first search 1
2
1
1
1 3 1
4 1
3 Visited 5 1
6 1
2 5 7 1
8 1
4 9 1
6 7 10

Queue
8 9
6 7 9
10
 Breadth first search 1
2
1
1
1 3 1
4 1
3 Visited 5 1
6 1
2 5 7 1
8 1
4 9 1
6 7 10

Queue
8 9
7 9
10
 Breadth first search 1
2
1
1
1 3 1
4 1
3 Visited 5 1
6 1
2 5 7 1
8 1
4 9 1
6 7 10

Queue
8 9
9
10
 Breadth first search 1
2
1
1
1 3 1
4 1
3 Visited 5 1
6 1
2 5 7 1
8 1
4 9 1
6 7 10

Queue
8 9

10
 Breadth first search 1
2 1
1

1 3 1
4 1
5 1
3 Visited
6 1
2 5 7 1
8 1
4 9 1
6 7 10 1

Queue
8 9
10
10
 Breadth first search 1
2 1
1

1 3 1
4 1
5 1
3 Visited
6 1
2 5 7 1
8 1
4 9 1
6 7 10 1

Queue
8 9

10
Breadth first search
function BFS(i) // BFS starting from vertex i

//Initialization
for j = 1..n {visited[j] = 0}; Q = []

//Start the exploration at i

visited[i] = 1; append(Q,i)

//Explore each vertex in Q

while Q is not empty
j = extract_head(Q)
for each (j,k) in E
if visited[k] == 0
visited[k] = 1; append(Q,k)
Complexity of BFS
Each vertex enters Q exactly once

If graph is connected, loop to process Q iterated n

times

For each j extracted from Q, need to examine all

neighbours of j

In adjacency matrix, scan row j: n entries

Hence, overall O(n2)

Complexity of BFS
Let m be the number of edges in E. What if m << n2?

Adjacency list: scanning neighbours of j takes time

proportional to number of neighbours (degree of j)

Across the loop, each edge (i,j) is scanned twice,

once when exploring i and again when exploring j

Overall, exploring neighbours takes time O(m)

Marking n vertices visited still takes O(n)

Overall, O(n+m)
Complexity of BFS

For graphs, O(m+n) is considered the best

possible

Need to see each edge and vertex at least once

O(m+n) is considered to be linear in the size of the

graph
Enhancements to BFS

If BFS(i) sets visited[j] = 1, we know that i and j are

connected

How do we identify a path from i to j

When we mark visited[k] = 1, remember the

neighbour from which we marked it

If exploring edge (j,k) visits k, set parent[k] = j

Breadth first search
function BFS(i) // BFS starting from vertex i

//Initialization
for j = 1..n {visited[j] = 0; parent[j] = -1}
Q = []

//Start the exploration at i

visited[i] = 1; append(Q,i)

//Explore each vertex in Q

while Q is not empty
j = extract_head(Q)
for each (j,k) in E
if visited[k] == 0
visited[k] = 1; parent[k] = j; append(Q,k);
Reconstructing the path

BFS(i) sets visited[j] = 1

visited[j] = 1, so parent[j] = j’ for some j’

visited[j’] = 1, so parent[j’] = j” for some j’’

Eventually, trace back path to k with parent[k] = i

Recording distances

BFS can record how long the path is to each

vertex

Instead of binary array visited[ ], keep integer array

level[ ]

level[j] = -1 initially

level[j] = p means j is reached in p steps from i

Breadth first search
function BFS(i) // BFS starting from vertex i

//Initialization
for j = 1..n {level[j] = -1; parent[j] = -1}
Q = []

//Start the exploration at i, level[i] set to 0

level[i] = 0; append(Q,i)

//Explore each vertex in Q, increment level for each new vertex

while Q is not empty
j = extract_head(Q)
for each (j,k) in E
if level[k] == -1
level[k] = 1+level[j]; parent[k] = j;
append(Q,k);
 L P
Breadth first search 1
2
1 L : Level 3
P : Parent 4
5
3
6
2 5 7
8
4 9
6 7 10

Queue
8 9

10
 L P
Breadth first search 1
2
1 L : Level 3
P : Parent 4
5
3
6
2 5 7
8
4 9
6 7 10

Queue
8 9

10 head tail
 L P
Breadth first search 1
2
0 -

1 L : Level 3
P : Parent 4
5
3
6
2 5 7
8
4 9
6 7 10

Queue
8 9
1
10
 L P
Breadth first search 1
2
0 -

1 L : Level 3
P : Parent 4
5
3
6
2 5 7
8
4 9
6 7 10

Queue
8 9

10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3
P : Parent 4
5
3
6
2 5 7
8
4 9
6 7 10

Queue
8 9
2
10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4
5
3
6
2 5 7
8
4 9
6 7 10

Queue
8 9
2 3
10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
5
3
6
2 5 7
8
4 9
6 7 10

Queue
8 9
2 3 4
10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
5
3
6
2 5 7
8
4 9
6 7 10

Queue
8 9
3 4
10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
5
3
6
2 5 7
8
4 9
6 7 10

Queue
8 9
4
10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
5
3
6
2 5 7
8
4 9
6 7 10

Queue
8 9

10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
3 5 2 4
6
2 5 7
8
4 9
6 7 10

Queue
8 9
5
10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
3 5 2 4
6
2 5 7
8 2 4
4 9
6 7 10

Queue
8 9
5 8
10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
3 5 2 4
6
2 5 7
8 2 4
4 9
6 7 10

Queue
8 9
8
10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
3 5 2 4
6 3 5
2 5 7
8 2 4
4 9
6 7 10

Queue
8 9
8 6
10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
3 5 2 4
6 3 5
2 5 7 3 5
8 2 4
4 9
6 7 10

Queue
8 9
8 6 7
10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
3 5 2 4
6 3 5
2 5 7 3 5
8 2 4
4 9
6 7 10

Queue
8 9
6 7
10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
3 5 2 4
6 3 5
2 5 7 3 5
8 2 4
4 9 3 8
6 7 10

Queue
8 9
6 7 9
10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
3 5 2 4
6 3 5
2 5 7 3 5
8 2 4
4 9 3 8
6 7 10

Queue
8 9
7 9
10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
3 5 2 4
6 3 5
2 5 7 3 5
8 2 4
4 9 3 8
6 7 10

Queue
8 9
9
10
 L P
Breadth first search 1
2
0
1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
3 5 2 4
6 3 5
2 5 7 3 5
8 2 4
4 9 3 8
6 7 10

Queue
8 9

10
 L P
Breadth first search 1 0
2 1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
3 5 2 4
6 3 5
2 5 7 3 5
8 2 4
4 9 3 8
6 7 10 4 9
Queue
8 9
10
10
 L P
Breadth first search 1 0
2 1
-
1
1 L : Level 3 1 1
P : Parent 4 1 1
3 5 2 4
6 3 5
2 5 7 3 5
8 2 4
4 9 3 8
6 7 10 4 9
Queue
8 9

10
Recording distances
BFS with level[ ] gives us the shortest path to each
node in terms of number of edges

In general, edges are labelled by a cost (money,

time, distance …)

Min cost path not same as fewest edges

Will look at shortest paths in weighted graphs later

BFS computes shortest paths if all costs are 1