DFS Applications

Strongly connected components

Instructor: Ashok Singh Sairam

Strongly Connected Components
• G is strongly connected if every pair (u, v) of vertices
in G is reachable from one another.
• A strongly connected component (SCC) of G is a
maximal set of vertices C  V such that for all u, v 
C, both u v and v u exist.

• Aim: Find SCC of a DAG

MA512: Data Structures and Algorithms
2
SCC: Applications
• Packaging software modules
 Construct directed graph of which modules call which
other modules
• Social networking sites
 Find people who are friends of one another and have
common interest

MA512: Data Structures and Algorithms

3
Transpose of a Directed Graph
• GT = transpose of a
G
b c d
directed G.
T T
 G = (V, E ),

 ET = {(u, v):(v, u)  E }.
T e f g h
 G is G with all edges

reversed. a GT b c d

e f g h
MA512: Data Structures and Algorithms
4
Transpose of a Directed Graph
• Transpose: Adjacency list representation
 Adj’[1..|V|]: new adjacency list representation
 For each vertex 𝑢 ∈ 𝐺

• For each vertex 𝑣 ∈ 𝑎𝑑𝑗 𝑢

• Insert (Adj’[v], u)
• Can create GT in Θ(V + E) time if using adjacency lists.
• G and GT have the same SCC’s. (u and v are reachable from
each other in G if and only if reachable from each other in GT.)

MA512: Data Structures and Algorithms

5
Component Graph
• GSCC = (VSCC, ESCC).
• VSCC has one vertex for each SCC in G.
• ESCC has an edge if there’s an edge between the
corresponding SCC’s in G.

• We will show that the component graph is a DAG

MA512: Data Structures and Algorithms

6
Component Graph: example

• GSCC for the example considered:

MA512: Data Structures and Algorithms

7
SCCs is a DAG: example
• If there is a path from u to u’, there cannot be a path
from v’ to v

u u’

v v’

C C’

MA512: Data Structures and Algorithms

8
 GSCC are a DAG
Lemma 22.13
Let C and C be distinct SCC’s in G, let u, v  C, u, v  C, and
suppose there is a path u u in G. Then there cannot also be a path
v v in G.
Proof:
• Suppose there is a path v v in G.
• Then there are paths u u v and v v u in G.
• Therefore, u and v are reachable from each other, so
they are not in separate SCC’s

MA512: Data Structures and Algorithms

9
 SCC: Algorithm
SCC(G)
1. call DFS(G) to compute finishing times f [u] for all u
2. compute GT
3. call DFS(GT), but in the main loop, consider vertices in order of
decreasing f [u] (as computed in first DFS)
4. output the vertices in each tree of the depth-first forest formed
in second DFS as a separate SCC

MA512: Data Structures and Algorithms

10
SCC: Example
• Compute finishing time of all vertices
G a b c d
13/14 11/16 1/10 8/9

12/15 3/4 2/7 5/6

e f g h

MA512: Data Structures and Algorithms

11
 a b c d

13/14 11/16 1/10 8/9

12/15 3/4 2/7 5/6

• Compute GT e f g h

• Call DFS, consider vertices in decreasing order f(u)

a b c d

e f g h

MA512: Data Structures and Algorithms

12
 a b c d a b c d
13/14 11/16 1/10 8/9

12/15 3/4 2/7 5/6

e f g h
e f g h

• Output vertices in each tree in second DFS …

cd
abe

fg h

MA512: Data Structures and Algorithms

13
Exercise

MA512: Data Structures and Algorithms

14
How does it work?
• Idea:
 Considering vertices in 2nd DFS in decreasing order f(u) 
visiting vertices of component graph in topological order.
 Because we are running DFS on GT, we will not be visiting

any v from a u, where v and u are in different components.

• Notation:
 d[u] and f [u] always refer to first DFS.

 Extend notation for d and f to sets of vertices U  V:

 d(U) = minuU{d[u]} (earliest discovery time)

 f (U) = maxuU{ f [u]} (latest finishing time)

MA512: Data Structures and Algorithms

15
Lemma 22.14
Let C and C be distinct SCC’s in G = (V, E). Suppose there is an
edge (u, v)  E such that u  C and v C. Then f (C) > f (C).
Proof: Case 1: d(C) < d(C)
 Let x be the first vertex discovered in C.
 At time d[x], all vertices in C and C are white. Thus, there
exist paths of white vertices from x to all vertices in C and C.
 By the white-path theorem, all vertices in C and C are
descendants of x in depth-first tree.
 By the parenthesis theorem, f [x] = f (C) > f(C).
C
C

u v

x
MA512: Data Structures and Algorithms
16
Proof:Case 2: d(C) > d(C)
 Let y be the first vertex discovered in C.
 At time d[y], all vertices in C are white and there is a white
path from y to each vertex in C  all vertices in C become
descendants of y. Again, f [y] = f (C).
 At time d[y], all vertices in C are also white.

 By earlier lemma, since there is an edge (u, v), we cannot

have a path from C to C.
 So no vertex in C is reachable from y.
 Therefore, at time f [y], all vertices in C are still white.
 Therefore, for all w  C, f [w] > f [y], which implies that f (C)
> f (C). C C

u v

x y

MA512: Data Structures and Algorithms

17
Corollary 22.15
Let C and C be distinct SCC’s in G = (V, E). Suppose there is an
edge (u, v)  ET, where u  C and v  C. Then f(C) < f(C).
Proof:
• (u, v)  ET  (v, u)  E.
• Since SCC’s of G and GT are the same, f(C) > f (C),
by Lemma 22.14.

MA512: Data Structures and Algorithms

18
Correctness of SCC
• When we do the second DFS, on GT, start with SCC
C such that f(C) is maximum.
 The second DFS starts from some x  C, and it visits all
vertices in C.
 Corollary 22.15 says that since f(C) > f (C) for all C  C,
there are no edges from C to C in GT.
 Therefore, DFS will visit only vertices in C.
 Which means that the depth-first tree rooted at x
contains exactly the vertices of C.

MA512: Data Structures and Algorithms

19
• The next root chosen in the second DFS is in SCC
C such that f (C) is maximum over all SCC’s other
than C.
 DFS visits all vertices in C, but the only edges out of C
go to C, which we’ve already visited.
 Therefore, the only tree edges will be to vertices in C.
• We can continue the process.
• Each time we choose a root for the second DFS, it
can reach only
 vertices in its SCC—get tree edges to these,
 vertices in SCC’s already visited in second DFS—get no
tree edges to these.

MA512: Data Structures and Algorithms

20
Acknowledgement
David A. Plaisted
Department of Computer Science
University of North Carolina at Chapel Hill

Analysis of Algorithms
21