NPTEL MOOC,JAN-FEB 2015

Week 3, Module 4

DESIGN AND ANALYSIS

OF ALGORITHMS
Depth first search (DFS)

MADHAVAN MUKUND, CHENNAI MATHEMATICAL INSTITUTE

http://www.cmi.ac.in/~madhavan
Depth first search
Start from i, visit a neighbour j

Suspend the exploration of i and explore j instead

Continue till you reach a vertex with no unexplored

neighbours

Backtrack to nearest suspended vertex that still has an

unexplored neighbour

Suspended vertices are stored in a stack

Last in, first out: most recently suspended is checked

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

9 Stack of suspended vertices

8

10
 Depth first search 1
2
Start at 4
1 3
4
5
3 Visited
6
2 5 7
8
4 9
6 7 10

9 Stack of suspended vertices

8

10
 Depth first search 1
2
Start at 4
1 3
4 1
5
3 Visited
6
2 5 7
8
4 9
6 7 10

9 Stack of suspended vertices

8

10
 Depth first search 1
2
1
Start at 4
1 3
4 1
5
3 Visited
6
2 5 7
8
4 9
6 7 10

9 Stack of suspended vertices

8
4
10
 Depth first search 1
2
1
1
Start at 4
1 3
4 1
5
3 Visited
6
2 5 7
8
4 9
6 7 10

9 Stack of suspended vertices

8
4 1
10
 Depth first search 1
2
1
1
Start at 4
1 3 1
4 1
5
3 Visited
6
2 5 7
8
4 9
6 7 10

9 Stack of suspended vertices

8
4 1 2
10
 Depth first search 1
2
1
1
Start at 4
1 3 1
4 1
5
3 Visited
6
2 5 7
8
4 9
6 7 10

9 Stack of suspended vertices

8
4 1
10
 Depth first search 1
2
1
1
Start at 4
1 3 1
4 1
5
3 Visited
6
2 5 7
8
4 9
6 7 10

9 Stack of suspended vertices

8
4
10
 Depth first search 1
2
1
1
Start at 4
1 3 1
4 1
5
3 Visited
6
2 5 7
8
4 9
6 7 10

9 Stack of suspended vertices

8

10
 Depth first search 1
2
1
1
Start at 4
1 3 1
4 1
5 1
3 Visited
6
2 5 7
8
4 9
6 7 10

9 Stack of suspended vertices

8
4
10
 Depth first search 1
2
1
1
Start at 4
1 3 1
4 1
5 1
3 Visited
6 1
2 5 7
8
4 9
6 7 10

9 Stack of suspended vertices

8
4 5
10
 Depth first search 1
2
1
1
Start at 4
1 3 1
4 1
5 1
3 Visited
6 1
2 5 7 1
8
4 9
6 7 10

9 Stack of suspended vertices

8
4 5 6
10
 Depth first search 1
2
1
1
Start at 4
1 3 1
4 1
5 1
3 Visited
6 1
2 5 7 1
8
4 9
6 7 10

9 Stack of suspended vertices

8
4 5
10
 Depth first search 1
2
1
1
Start at 4
1 3 1
4 1
5 1
3 Visited
6 1
2 5 7 1
8 1
4 9
6 7 10

9 Stack of suspended vertices

8
4 5 6
10
 Depth first search 1
2
1
1
Start at 4
1 3 1
4 1
5 1
3 Visited
6 1
2 5 7 1
8 1
4 9 1
6 7 10

9 Stack of suspended vertices

8
4 5 6 8
10
 Depth first search 1
2 1
1
Start at 4
1 3 1
4 1
5 1
3 Visited
6 1
2 5 7 1
8 1
4 9 1
6 7 10 1

9 Stack of suspended vertices

8
4 5 6 8 9
10
 Depth first search 1
2 1
1
Start at 4
1 3 1
4 1
5 1
3 Visited
6 1
2 5 7 1
8 1
4 9 1
6 7 10 1

9 Stack of suspended vertices

8
4 5 6 8
10
 Depth first search 1
2 1
1
Start at 4
1 3 1
4 1
5 1
3 Visited
6 1
2 5 7 1
8 1
4 9 1
6 7 10 1

9 Stack of suspended vertices

8
4 5 6
10
 Depth first search 1
2 1
1
Start at 4
1 3 1
4 1
5 1
3 Visited
6 1
2 5 7 1
8 1
4 9 1
6 7 10 1

9 Stack of suspended vertices

8
4 5
10
 Depth first search 1
2 1
1
Start at 4
1 3 1
4 1
5 1
3 Visited
6 1
2 5 7 1
8 1
4 9 1
6 7 10 1

9 Stack of suspended vertices

8
4
10
 Depth first search 1
2 1
1
Start at 4
1 3 1
4 1
5 1
3 Visited
6 1
2 5 7 1
8 1
4 9 1
6 7 10 1

9 Stack of suspended vertices

8

10
Depth first search
Depth first search

DFS is most natural to implement recursively

For each unvisited neighbour j of i, call DFS(j)

Depth first search

DFS is most natural to implement recursively

For each unvisited neighbour j of i, call DFS(j)

No need to explicitly maintain a stack

Stack is maintained implicitly by recursive calls

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

function DFS(i) // DFS starting from vertex i

//Mark i as visited
visited[i] = 1

//Explore each neighbour of i recursively

for each (i,j) in E
if visited[j] == 0
parent[j] = i
DFS(j)
Complexity of DFS
Complexity of DFS
Each vertex marked and explored exactly once
Complexity of DFS
Each vertex marked and explored exactly once

DFS(j) need to examine all neighbours of j

Complexity of DFS
Each vertex marked and explored exactly once

DFS(j) need to examine all neighbours of j

In adjacency matrix, scan row j: n entries

Overall O(n2)
Complexity of DFS
Each vertex marked and explored exactly once

DFS(j) need to examine all neighbours of j

In adjacency matrix, scan row j: n entries

Overall O(n2)

With adjacency list, scanning takes O(m) time

across all vertices

Total time is O(m+n), like BFS

Properties of DFS
Properties of DFS
Paths discovered by DFS are not shortest paths,
unlike BFS
Properties of DFS
Paths discovered by DFS are not shortest paths,
unlike BFS

Why use DFS at all?

Properties of DFS
Paths discovered by DFS are not shortest paths,
unlike BFS

Why use DFS at all?

Many useful features can be extracted from recording

the order in which DFS visited vertices
Properties of DFS
Paths discovered by DFS are not shortest paths,
unlike BFS

Why use DFS at all?

Many useful features can be extracted from recording

the order in which DFS visited vertices

DFS numbering
Properties of DFS
Paths discovered by DFS are not shortest paths,
unlike BFS

Why use DFS at all?

Many useful features can be extracted from recording

the order in which DFS visited vertices

DFS numbering

Maintain a counter
Properties of DFS
Paths discovered by DFS are not shortest paths,
unlike BFS

Why use DFS at all?

Many useful features can be extracted from recording

the order in which DFS visited vertices

DFS numbering

Maintain a counter

Increment and record counter value when entering

and leaving a vertex.
Depth first search
//Initialization
for j = 1..n {visited[j] = 0; parent[j] = -1}
count = 0

function DFS(i) // DFS starting from vertex i

//Mark i as visited
visited[i] = 1; pre[i] = count; count++

//Explore each neighbours of i recursively

for each (i,j) in E
if visited[j] == 0
parent[j] = i
DFS(j)
post[i] = count; count++
DFS numbering
1

2 5

4
6 7

8 9

10
DFS numbering
1

pre[i] and post[i] can be used

3
to find
2 5
if the graph has a cycle —
i.e., a loop 4
6 7
cut vertex — removal
disconnects the graph
8 9
…
10
Summary
BFS and DFS are two systematic ways to explore a
graph

Both take time linear in the size of the graph with

adjacency lists

Recover paths by keeping parent information

BFS can compute shortest paths, in terms of

number of edges

DFS numbering can reveal many interesting features