Digraphs

 A digraph is a graph
Directed Graphs BOS
whose edges are all
directed
E

ORD
 Short for “directed graph” D
JFK

SFO
 Applications C
 one-way streets
LAX
DFW
 flights B
MIA
 task scheduling
A

© 2010 Goodrich, Tamassia Directed Graphs 1 © 2010 Goodrich, Tamassia Directed Graphs 2

Digraph Properties D Digraph Application

C  Scheduling: edge (a,b) means task a must be
 A graph G=(V,E) such that B completed before b can be started
 Each edge goes in one direction: A ics21 ics22 ics23
 Edge (a,b) goes from a to b, but not b to a
 If G is simple, m < n⋅(n − 1)
ics51 ics53 ics52
 If we keep in-edges and out-edges in separate ics161
adjacency lists, we can perform listing of
ics131 ics141 ics121 ics171
incoming edges and outgoing edges in time
proportional to their size
ics151 The good life

© 2010 Goodrich, Tamassia Directed Graphs 3 © 2010 Goodrich, Tamassia Directed Graphs 4
 Directed DFS Reachability
 We can specialize the traversal
algorithms (DFS and BFS) to
E
 DFS tree rooted at v: vertices reachable
digraphs by traversing edges
only along their direction from v via directed paths
D E D
 In the directed DFS algorithm,
we have four types of edges
E D C
 discovery edges C
 back edges
A
 forward edges B C F
cross edges
E D


A directed DFS starting at a A A B


vertex s determines the vertices
C F
reachable from s
A B
© 2010 Goodrich, Tamassia Directed Graphs 5 © 2010 Goodrich, Tamassia Directed Graphs 6

Strong Connectivity
Strong Connectivity Algorithm
 Each vertex can reach all other vertices  Pick a vertex v in G
a
 Perform a DFS from v in G G: c
g

If there’s a w not visited, print “no”

a 
d
g  Let G’ be G with edges reversed e
c  Perform a DFS from v in G’ f b

 If there’s a w not visited, print “no”

 Else, print “yes” a
d g
 Running time: O(n+m) G’: c
e
d
e
b
f f b

© 2010 Goodrich, Tamassia Directed Graphs 7 © 2010 Goodrich, Tamassia Directed Graphs 8
 Strongly Connected
Components Transitive Closure
 Maximal subgraphs such that each vertex can reach  Given a digraph G, the D E
all other vertices in the subgraph transitive closure of G is the
digraph G* such that B
 Can also be done in O(n+m) time using DFS, but is G
 G* has the same vertices C
more complicated (similar to biconnectivity).
as G
A
a  if G has a directed path
g {a,c,g} from u to v (u ≠ v), G*
c
has a directed edge from D E
u to v
d
e  The transitive closure B
{f,d,e,b} provides reachability
b information about a digraph C
f
A G*
© 2010 Goodrich, Tamassia Directed Graphs 9 © 2010 Goodrich, Tamassia Directed Graphs 10

Computing the Floyd-Warshall

Transitive Closure Transitive Closure
If there's a way to get
 We can perform from A to B and from  Idea #1: Number the vertices 1, 2, …, n.
DFS starting at B to C, then there's a  Idea #2: Consider paths that use only
each vertex way to get from A to C.
vertices numbered 1, 2, …, k, as
 O(n(n+m)) intermediate vertices:
Uses only vertices numbered 1,…,k
(add this edge if it’s not already in)
i
Alternatively ... Use
dynamic programming:
The Floyd-Warshall Uses only vertices j
Algorithm numbered 1,…,k-1 Uses only vertices
k numbered 1,…,k-1
© 2010 Goodrich, Tamassia Directed Graphs 11 © 2010 Goodrich, Tamassia Directed Graphs 12
 Floyd-Warshall’s Algorithm Floyd-Warshall Example v7
 Number vertices v1 , …, vn Algorithm FloydWarshall(G) BOS
Input digraph G
 Compute digraphs G0, …, Gn Output transitive closure G* of G ORD v4
 G0=G i←1
 Gk has directed edge (vi, vj) for all v ∈ G.vertices() JFK
if G has a directed path denote v as vi v2
i←i+1
v6
from vi to vj with
intermediate vertices in G0 ← G SFO
{v1 , …, vk} for k ← 1 to n do
 We have that Gn = G* G k ← Gk − 1
for i ← 1 to n (i ≠ k) do
 In phase k, digraph Gk is for j ← 1 to n (j ≠ i, k) do DFW
computed from Gk − 1 if Gk − 1.areAdjacent(vi, vk) ∧ LAX
Running time: O(n3), Gk − 1.areAdjacent(vk, vj)
v3
 v1
assuming areAdjacent is O(1) if ¬Gk.areAdjacent(vi, vj) MIA
(e.g., adjacency matrix) Gk.insertDirectedEdge(vi, vj , k)
return Gn v5
© 2010 Goodrich, Tamassia Directed Graphs 13 © 2010 Goodrich, Tamassia Directed Graphs 14

Floyd-Warshall, Iteration 1 v7 Floyd-Warshall, Iteration 2 v7

BOS BOS

ORD v4 ORD v4
JFK JFK
v2 v6 v2 v6
SFO SFO

DFW DFW
LAX LAX
v3 v3
v1 v1
MIA MIA

v5 v5
© 2010 Goodrich, Tamassia Directed Graphs 15 © 2010 Goodrich, Tamassia Directed Graphs 16
 Floyd-Warshall, Iteration 3 v7 Floyd-Warshall, Iteration 4 v7
BOS BOS

ORD v4 ORD v4
JFK JFK
v2 v6 v2 v6
SFO SFO

DFW DFW
LAX LAX
v3 v3
v1 v1
MIA MIA

v5 v5
© 2010 Goodrich, Tamassia Directed Graphs 17 © 2010 Goodrich, Tamassia Directed Graphs 18

Floyd-Warshall, Iteration 5 v7 Floyd-Warshall, Iteration 6 v7

BOS BOS

ORD v4 ORD v4
JFK JFK
v2 v6 v2 v6
SFO SFO

DFW DFW
LAX LAX
v3 v3
v1 v1
MIA MIA

v5 v5
© 2010 Goodrich, Tamassia Directed Graphs 19 © 2010 Goodrich, Tamassia Directed Graphs 20
 Floyd-Warshall, Conclusion v7
DAGs and Topological Ordering
BOS
 A directed acyclic graph (DAG) is a D E
v4 digraph that has no directed cycles
ORD
 A topological ordering of a digraph B
JFK is a numbering
v2 v1 , …, vn
C
v6
of the vertices such that for every A DAG G
SFO edge (vi , vj), we have i < j
 Example: in a task scheduling v4 v5
digraph, a topological ordering a D E
DFW task sequence that satisfies the v2
precedence constraints
LAX
v3 B
Theorem v3
v1
MIA A digraph admits a topological v1 C
ordering if and only if it is a DAG Topological
v5 A ordering of G
© 2010 Goodrich, Tamassia Directed Graphs 21 © 2010 Goodrich, Tamassia Directed Graphs 22

Topological Sorting Algorithm for Topological Sorting

 Number vertices, so that (u,v) in E implies u < v  Note: This algorithm is different than the
1
wake up A typical student day one in the book
2 3
eat
study computer sci.
Algorithm TopologicalSort(G)
4 5 H←G // Temporary copy of G
nap more c.s. n ← G.numVertices()
7
play while H is not empty do
8 Let v be a vertex with no outgoing edges
write c.s. program 6 Label v ← n
9 work out n←n−1
bake cookies Remove v from H
10
11
sleep
 Running time: O(n + m)
dream about graphs
© 2010 Goodrich, Tamassia Directed Graphs 23 © 2010 Goodrich, Tamassia Directed Graphs 24
 Implementation with DFS Topological Sorting Example
 Simulate the algorithm by Algorithm topologicalDFS(G, v)
using depth-first search Input graph G and a start vertex v of G
 O(n+m) time. Output labeling of the vertices of G
in the connected component of v
v.setLabel(VISITED)
Algorithm topologicalDFS(G) for all e ∈ v.outEdges()
Input dag G { outgoing edges }
Output topological ordering of G w ← e.opposite(v)
n ← G.numVertices() if w.getLabel() = UNEXPLORED
for all u ∈ G.vertices() { e is a discovery edge }
u.setLabel(UNEXPLORED) topologicalDFS(G, w)
for all v ∈ G.vertices() else
if v.getLabel() = UNEXPLORED { e is a forward or cross edge }
topologicalDFS(G, v) Label v with topological number n
n←n-1

© 2010 Goodrich, Tamassia Directed Graphs 25 © 2010 Goodrich, Tamassia Directed Graphs 26

Topological Sorting Example Topological Sorting Example

9 9

© 2010 Goodrich, Tamassia Directed Graphs 27 © 2010 Goodrich, Tamassia Directed Graphs 28
 Topological Sorting Example Topological Sorting Example

7 7
8 8

9 9

© 2010 Goodrich, Tamassia Directed Graphs 29 © 2010 Goodrich, Tamassia Directed Graphs 30

Topological Sorting Example Topological Sorting Example

4
6 5 6 5

7 7
8 8

9 9

© 2010 Goodrich, Tamassia Directed Graphs 31 © 2010 Goodrich, Tamassia Directed Graphs 32
 Topological Sorting Example Topological Sorting Example
2

3 3
4 4
6 5 6 5

7 7
8 8

9 9

© 2010 Goodrich, Tamassia Directed Graphs 33 © 2010 Goodrich, Tamassia Directed Graphs 34

Topological Sorting Example

2
1

3
4
6 5

7
8

© 2010 Goodrich, Tamassia Directed Graphs 35