Slides credited from Hsueh-I Lu, Hsu-Chun Hsiao, & Michael Tsai

▪ Mini-HW 7 released
▪ Due on 11/30 (Thur) 17:20

▪ Homework 3 released tonight

▪ Due on 12/14 (Thur) 17:20 (three weeks)

▪ Class change!!
▪ Start from 12/07 (Thur) to forever (NOT next week!!)
▪ Location: R103

Frequently check the website for the updated information!

2
3
▪ Graph Basics
▪ Graph Theory
▪ Graph Representations
▪ Graph Traversal
▪ Breadth-First Search (BFS)
▪ Depth-First Search (DFS)

▪ DFS Applications
▪ Connected Components
▪ Strongly Connected Components
▪ Topological Sorting
4
▪ A graph G is defined as
▪ V: a finite, nonempty set of vertices
▪ E: a set of edges / pairs of vertices
1 2

3 4 5

5
▪ Graph type
▪ Undirected: edge 𝑢, 𝑣 = 𝑣, 𝑢
▪ Directed: edge 𝑢, 𝑣 goes from vertex 𝑢 to vertex 𝑣; 𝑢, 𝑣 ≠ 𝑣, 𝑢
▪ Weighted: edges associate with weights

1 2 1 2

3 4 5 3 4 5

6
How many edges at most can a undirected (or directed) graph have?
▪ Adjacent (相鄰)
▪ If there is an edge 𝑢, 𝑣 , then 𝑢 and 𝑣 are adjacent.
▪ Incident (作用)
▪ If there is an edge 𝑢, 𝑣 , the edge 𝑢, 𝑣 is incident from 𝑢 and
is incident to 𝑣.
▪ Subgraph (子圖)
▪ If a graph 𝐺 ′ = 𝑉 ′ , 𝐸′ is a subgraph of 𝐺 = 𝑉, 𝐸 , then 𝑉 ′ ⊆
𝑉 and 𝐸 ′ ⊆ 𝐸

7
▪ Degree
▪ The degree of a vertex 𝑢 is the number of edges incident on 𝑢
▪ In-degree of 𝑢: #edges 𝑥, 𝑢 in a directed graph
▪ Out-degree of 𝑢: #edges 𝑢, 𝑥 in a directed graph
▪ Degree = in-degree + out-degree
▪ Isolated vertex: degree = 0

σ𝑖 𝑑𝑖
𝐸 =
2

8
▪ Path
▪ a sequence of edges that connect a sequence of vertices
▪ If there is a path from 𝑢 (source) to 𝑣 (target), there are a sequence
of edges 𝑢, 𝑖1 , 𝑖1 , 𝑖2 , … , 𝑖𝑘−1 , 𝑖𝑘 , (𝑖𝑘 , 𝑣)
▪ Reachable: 𝑣 is reachable from 𝑢 if there exists a path from 𝑢 to 𝑣

▪ Simple Path
▪ All vertices except for 𝑢 and 𝑣 are all distinct

▪ Cycle
▪ A simple path where 𝑢 and 𝑣 are the same

▪ Subpath
▪ A subsequence of the path 9
▪ Connected
▪ Two vertices are connected if there is a path between them
▪ A connected graph has a path from every vertex to every other

▪ Tree
▪ a connected, acyclic, undirected graph

▪ Forest
▪ an acyclic, undirected but possibly disconnected graph

1 2 1 2 1 2

3 4 5 3 4 5 3 4 5
10
▪ Theorem. Let 𝐺 be an undirected graph. The following
statements are equivalent:
▪ 𝐺 is a tree
▪ Any two vertices in 𝐺 are connected by a unique simple path
▪ 𝐺 is connected, but if any edge is removed from 𝐸, the
resulting graph is disconnected.
▪ 𝐺 is connected and 𝐸 = 𝑉 − 1
▪ 𝐺 is acyclic, and 𝐸 = 𝑉 − 1
▪ 𝐺 is acyclic, but if any edge is added to 𝐸, the resulting graph
contains a cycle

Proofs in Textbook Appendix B.5

11
12
▪ How to traverse all bridges where each one can only be
passed through once

A A

B C
B C
D

D 13
▪ Euler path
▪ Can you traverse each edge in a connected graph exactly once
without lifting the pen from the paper?
▪ Euler tour
▪ Can you finish where you started?

A A A

B C B C B C

D D D
Euler path Euler path Euler path 14
Euler tour Euler tour Euler tour
 Is it possible to determine whether a graph has an Euler path or an Euler
tour, without necessarily having to find one explicitly?

▪ Solved by Leonhard Euler in 1736

▪ 𝐺 has an Euler path iff 𝐺 has exactly 0 or 2 odd vertices
▪ 𝐺 has an Euler tour iff all vertices must be even vertices

Even vertices = vertices with even degrees

Odd vertices = vertices with odd degrees

15
▪ Hamiltonian Path
▪ A path that visits each vertex exactly once

▪ Hamiltonian Cycle
▪ A Hamiltonian path where the start and destination are the same

▪ Both are NP-complete

16
▪ Modeling applications using graph theory
▪ What do the vertices represent?
▪ What do the edges represent?
▪ Undirected or directed?

17
Social Network Knowledge Graph
18
▪ How to represent a graph in computer programs?
▪ Two standard ways to represent a graph 𝐺 = 𝑉, 𝐸
▪ Adjacency matrix
▪ Adjacency list

19
 Matrix
▪ Adjacency matrix = 𝑉 × 𝑉 matrix 𝐴 with 𝐴[𝑢][𝑣] = 1 if
(𝑢, 𝑣) is an edge
1 2 3 4 5 6
1 1 1 1
3 2 1 1 1
3 1 1 1
2
4 1 1 1
6 5 1
5 6 1 1
4
• For undirected graphs, 𝐴 is symmetric; i.e., 𝐴 = 𝐴𝑇
• If weighted, store weights instead of bits in 𝐴
20
 Matrix
▪ Space:
▪ Time for querying an edge:
▪ Time for inserting an edge:
▪ Time for deleting an edge:
▪ Time for listing all neighbors of a vertex:
▪ Time for identifying all edges:
▪ Time for finding in-degree and out-degree of a vertex?

21
 List
▪ Adjacency lists = vertex indexed array of lists
▪ One list per vertex, where for 𝑢 ∈ 𝑉, 𝐴[𝑢] consists of all
vertices adjacent to 𝑢
1
1 2 3
3 2 1 4 5

2 3 1 4 6
4 2 3 6
6 5 2
5 6 3 4
4
If weighted, store weights also in adjacency lists
22
 List
▪ Space:
▪ Time for querying an edge:
▪ Time for inserting an edge:
▪ Time for deleting an edge:
▪ Time for listing all neighbors of a vertex:
▪ Time for identifying all edges:
▪ Time for finding in-degree and out-degree of a vertex?

23
 ▪ Matrix representation is suitable for dense graphs
▪ List representation is suitable for sparse graphs
▪ Besides graph density, you may also choose a data structure
based on the performance of other operations

List a
Query an Insert an Delete an Identify
Space vertex’s
edge edge edge all edges
neighbors
Adjacency Matrix

Adjacency List

24
25

Textbook Chapter 22 – Elementary Graph Algorithms

▪ From a source vertex, systematically follow the edges of a
graph to visit all reachable vertices of the graph
▪ Useful to discover the structure of a graph
▪ Standard graph-searching algorithms
▪ Breadth-First Search (BFS, 廣度優先搜尋)
▪ Depth-First Search (DFS, 深度優先搜尋)

26
27

Textbook Chapter 22.2 – Breadth-first search

 Layer 2

Layer 1

Source 𝒔

28
▪ Input: directed/undirected graph 𝐺 = (𝑉, 𝐸) and source 𝑠
▪ Output: a breadth-first tree with root 𝑠 (𝑇BFS ) that contains
all reachable vertices
▪ 𝑣. 𝑑: distance from 𝑠 to 𝑣, for all 𝑣 ∈ 𝑉
▪ Distance is the length of a shortest path in G
▪ 𝑣. 𝑑 = ∞ if 𝑣 is not reachable from 𝑠
▪ 𝑣. 𝑑 is also the depth of 𝑣 in 𝑇BFS
▪ 𝑣. 𝜋 = 𝑢 if (𝑢, 𝑣) is the last edge on shortest path to 𝑣
▪ 𝑢 is 𝑣’s predecessor in 𝑇BFS

29
 BFS(G, s)
▪ Initially 𝑇BFS contains only 𝑠 for each vertex u in G.V-{s}
u.color = WHITE
▪ As 𝑣 is discovered from 𝑢, 𝑣 and u.d = ∞
(𝑢, 𝑣) are added to 𝑇BFS u.pi = NIL
▪ 𝑇BFS is not explicitly stored; can be s.color = GRAY
reconstructed from 𝑣. 𝜋 s.d = 0
s.pi = NIL
▪ Implemented via a FIFO queue Q = {}
ENQUEUE(Q, s)
▪ Color the vertices to keep track of while Q! = {}
progress: u = DEQUEUE(Q)
▪ GRAY: discovered (first time for each v in G.Adj[u]
encountered) if v.color == WHITE
v.color = GRAY
▪ BLACK: finished (all adjacent v.d = u.d + 1
vertices discovered) v.pi = u
▪ WHITE: undiscovered ENQUEUE(Q,v)
u.color = BLACK

30
𝑠 𝑤 𝑟
0 1 1

𝑟 𝑡 𝑥 𝑡 𝑥 𝑣
1 2 2 2 2 2

𝑥 𝑣 𝑢 𝑣 𝑢 𝑦
2 2 3 2 3 3
31
𝑢 𝑦 𝑦
3 3 3

32
▪ Definition of 𝛿(𝑠, 𝑣): the shortest-path distance from 𝑠 to 𝑣 = the
minimum number of edges in any path from 𝑠 to 𝑣
▪ If there is no path from 𝑠 to 𝑣, then 𝛿 𝑠, 𝑣 = ∞

▪ The BFS algorithm finds the shortest-path distance to each reachable

vertex in a graph 𝐺 from a given source vertex 𝑠 ∈ 𝑉.

33
Lemma 22.1
Let 𝐺 = 𝑉, 𝐸 be a directed or undirected graph, and let 𝑠 ∈ 𝑉 be an
arbitrary vertex. Then, for any edge 𝑢, 𝑣 ∈ 𝐸, 𝛿 𝑠, 𝑣 ≤ 𝛿 𝑠, 𝑢 + 1.
𝑠-𝑣的最短路徑一定會小於等於𝑠-𝑢的最短路徑距離+1
▪ Proof
▪ Case 1: 𝑢 is reachable from 𝑠
▪ 𝑠- 𝑢- 𝑣 is a path from 𝑠 to 𝑣 with length 𝛿 𝑠, 𝑢 + 1 s
▪ Hence, 𝛿 𝑠, 𝑣 ≤ 𝛿 𝑠, 𝑢 + 1 v
▪ Case 2: 𝑢 is unreachable from 𝑠
▪ Then 𝑣 must be unreachable too. 𝛿 𝑠, 𝑢 u
▪ Hence, the inequality still holds.

34
Lemma 22.2
Let 𝐺 = 𝑉, 𝐸 be a directed or undirected graph, and suppose BFS is run
on 𝐺 from a given source vertex 𝑠 ∈ 𝑉. Then upon termination, for each
vertex 𝑣 ∈ 𝑉, the value 𝑣. 𝑑 computed by BFS satisfies 𝑣. 𝑑 ≥ 𝛿 𝑠, 𝑣 .
▪ Proof by induction BFS算出的d值必定大於等於真正距離

Inductive hypothesis: 𝑣. 𝑑 ≥ 𝛿 𝑠, 𝑣 after 𝑛 ENQUEUE ops

▪ Holds when 𝑛 = 1: 𝑠 is in the queue and 𝑣. 𝑑 = ∞ for all 𝑣 ∈ 𝑉 𝑠
▪ After 𝑛 + 1 ENQUEUE ops, consider a white vertex 𝑣 that is
discovered during the search from a vertex 𝑢
(by induction hypothesis)
(by Lemma 22.1)
▪ Vertex 𝑣 is never enqueued again, so 𝑣. 𝑑 never changes again
35
Lemma 22.3
Suppose that during the execution of BFS on a graph 𝐺 = 𝑉, 𝐸 , the
queue 𝑄 contains the vertices 𝑣1 , 𝑣2 , … , 𝑣𝑟 , where 𝑣1 is the head of 𝑄
and 𝑣𝑟 is the tail. Then, 𝑣𝑟 . 𝑑 ≤ 𝑣1 . 𝑑 + 1 and 𝑣𝑖 . 𝑑 ≤ 𝑣𝑖+1 . 𝑑 for 1 ≤ 𝑖 < 𝑟.
• Q中最後一個點的d值 ≤ Q中第一個點的d值+1
▪ Proof by induction • Q中第i個點的d值 ≤ Q中第i+1點的d值
Inductive hypothesis:𝑣𝑟 . 𝑑 ≤ 𝑣1 . 𝑑 + 1 and 𝑣𝑖 . 𝑑 ≤ 𝑣𝑖+1 . 𝑑 after 𝑛 queue ops
▪ Holds when 𝑄 = 𝑠 .
▪ Consider two operations for inductive step:
▪ Dequeue op: when 𝑄 = 𝑣1 , 𝑣2 , … , 𝑣𝑟 and dequeue 𝑣1
▪ Enqueue op: when 𝑄 = 𝑣1 , 𝑣2 , … , 𝑣𝑟 and enqueue 𝑣𝑟+1

36
 Inductive H1 (Q中最後一個點的d值 ≤ Q中第一個點的d值+1)
hypothesis: H2 (Q中第i個點的d值 ≤ Q中第i+1點的d值)

▪ Dequeue op
𝑣1 𝑣2 … 𝑣𝑟−1 𝑣𝑟 (induction hypothesis H1)
(induction hypothesis H2)  H1 holds
𝑣2 … 𝑣𝑟−1 𝑣𝑟
 H2 holds
▪ Enqueue op
Let 𝑢 be 𝑣𝑟+1 ’s predecessor,
𝑢 𝑣1 𝑣2 … 𝑣𝑟−1 𝑣𝑟
Since 𝑢 has been removed from 𝑄, the new head
𝑢 𝑣1 𝑣2 … 𝑣𝑟−1 𝑣𝑟 𝑣𝑟+1 𝑣1 satisfies (induction hypothesis H2)
 H1 holds
(induction hypothesis H1)

 H2 holds 37
Corollary 22.4
Suppose that vertices 𝑣𝑖 and 𝑣𝑗 are enqueued during the execution of BFS,
and that 𝑣𝑖 is enqueued before 𝑣𝑗 . Then 𝑣𝑖 . 𝑑 ≤ 𝑣𝑗 . 𝑑 at the time that 𝑣𝑗 is
enqueued.
▪ Proof 若𝑣𝑖 比𝑣𝑗 早加入queue  𝑣𝑖 . 𝑑 ≤ 𝑣𝑗 . 𝑑

▪ Lemma 22.3 proves that 𝑣𝑖 . 𝑑 ≤ 𝑣𝑖+1 . 𝑑 for 1 ≤ 𝑖 < 𝑟

▪ Each vertex receives a finite 𝑑 value at most once during the course of BFS
▪ Hence, this is proved.

38
Theorem 22.5 – BFS Correctness
Let 𝐺 = 𝑉, 𝐸 be a directed or undirected graph, and and suppose that BFS is
run on 𝐺 from a given source vertex 𝑠 ∈ 𝑉.
1) BFS discovers every vertex 𝑣 ∈ 𝑉 that is reachable from the source 𝑠
2) Upon termination, 𝑣. 𝑑 = 𝛿 𝑠, 𝑣 for all 𝑣 ∈ 𝑉
3) For any vertex 𝑣 ≠ 𝑠 that is reachable from 𝑠, one of the shortest paths
from 𝑠 to 𝑣 is a shortest path from 𝑠 to 𝑣. 𝜋 followed by the edge 𝑣. 𝜋, 𝑣
▪ Proof of (1)
▪ All vertices 𝑣 reachable from 𝑠 must be discovered; otherwise they
would have 𝑣. 𝑑 = ∞ > 𝛿 𝑠, 𝑣 . (contradicting with Lemma 22.2)

39
(2)

▪ Proof of (2) by contradiction

▪ Assume some vertices receive 𝑑 values not equal to its shortest-path
distance
▪ Let 𝑣 be the vertex with minimum 𝛿 𝑠, 𝑣 that receives such an incorrect
𝑑 value; clearly 𝑣 ≠ 𝑠
▪ By Lemma 22.2, 𝑣. 𝑑 ≥ 𝛿 𝑠, 𝑣 , thus 𝑣. 𝑑 > 𝛿 𝑠, 𝑣 (𝑣 must be reachable)
▪ Let 𝑢 be the vertex immediately preceding 𝑣 on a shortest path from 𝑠 to
𝑣, so 𝛿 𝑠, 𝑣 = 𝛿 𝑠, 𝑢 + 1
▪ Because 𝛿 𝑠, 𝑢 < 𝛿 𝑠, 𝑣 and 𝑣 is the minimum 𝛿 𝑠, 𝑣 , we have 𝑢. 𝑑 =
𝛿 𝑠, 𝑢
▪ 𝑣. 𝑑 > 𝛿 𝑠, 𝑣 = 𝛿 𝑠, 𝑢 + 1 = 𝑢. 𝑑 + 1
40
(2)

▪ Proof of (2) by contradiction (cont.)

▪ 𝑣. 𝑑 > 𝛿 𝑠, 𝑣 = 𝛿 𝑠, 𝑢 + 1 = 𝑢. 𝑑 + 1
▪ When dequeuing 𝑢 from 𝑄, vertex 𝑣 is either WHITE, GRAY, or BLACK
▪ WHITE: 𝑣. 𝑑 = 𝑢. 𝑑 + 1, contradiction
▪ BLACK: it was already removed from the queue
▪ By Corollary 22.4, we have 𝑣. 𝑑 ≤ 𝑢. 𝑑, contradiction
▪ GRAY: it was painted GRAY upon dequeuing some vertex 𝑤
▪ Thus 𝑣. 𝑑 = 𝑤. 𝑑 + 1 (by construction)
▪ 𝑤 was removed from 𝑄 earlier than 𝑢, so 𝑤. 𝑑 ≤ 𝑢. 𝑑 (by Corollary 22.4)
▪ 𝑣. 𝑑 = 𝑤. 𝑑 + 1 ≤ 𝑢. 𝑑 + 1, contradiction
▪ Thus, (2) is proved.
41
(3) For any vertex 𝑣 ≠ 𝑠 that is reachable from 𝑠, one of the shortest paths
from 𝑠 to 𝑣 is a shortest path from 𝑠 to 𝑣. 𝜋 followed by the edge 𝑣. 𝜋, 𝑣
▪ Proof of (3)
▪ If 𝑣. 𝜋 = 𝑢, then 𝑣. 𝑑 = 𝑢. 𝑑 + 1. Thus, we can obtain a shortest path
from 𝑠 to 𝑣 by taking a shortest path from 𝑠 to 𝑣. 𝜋 and then traversing
the edge 𝑣. 𝜋, 𝑣 .

42
▪ BFS(G, s) forms a BFS tree with all reachable 𝑣 from 𝑠
▪ We can extend the algorithm to find a BFS forest that contains every
vertex in 𝐺
BFS-Visit(G, s)
//explore full graph and builds up s.color = GRAY
a collection of BFS trees s.d = 0
BFS(G) s.π = NIL
for u in G.V Q = empty
u.color = WHITE ENQUEUE(Q, s)
u.d = ∞ while Q ≠ empty
u.π = NIL u = DEQUEUE(Q)
for s in G.V for v in G.adj[u]
if(s.color == WHITE) if v.color == WHITE
// build a BFS tree v.color = GRAY
BFS-Visit(G, s) v.d = u.d + 1
v.π = u
ENQUEUE(Q, v)
u.color = BLACK
43
44

Textbook Chapter 22.3 – Depth-first search

▪ Search as deep as possible and then backtrack until finding a
new path
12 13
9 10
8 11
1 14

27
45
36

Timestamps: discovery time / finishing time 45

// Explore full graph and builds up DFS-Visit(G, u)
a collection of DFS trees time = time + 1
DFS(G) u.d = time // discover time
for each vertex u in G.V u.color = GRAY
u.color = WHITE for each v in G.Adj[u]
u.pi = NIL
if v.color == WHITE
time = 0 // global timestamp
v.pi = u
for each vertex u in G.V
if u.color == WHITE DFS-VISIT(G, v)
DFS-VISIT(G, u) u.color = BLACK
time = time + 1
u.f = time // finish time

▪ Implemented via recursion (stack)

▪ Color the vertices to keep track of progress:
▪ GRAY: discovered (first time encountered)
▪ BLACK: finished (all adjacent vertices discovered)
46
▪ WHITE: undiscovered
▪ Parenthesis Theorem
▪ Parenthesis structure: represent the discovery of vertex 𝑢 with a left
parenthesis “(𝑢” and represent its finishing by a right parenthesis “𝑢)”. In
DFS, the parentheses are properly nested.
▪ White Path Theorem
▪ In a DFS forest of a directed or undirected graph 𝐺 = 𝑉, 𝐸 ,
▪ vertex 𝑣 is a descendant of vertex 𝑢 in the forest  at the time 𝑢. 𝑑
that the search discovers 𝑢, there is a path from 𝑢 to 𝑣 in 𝐺 consisting
entirely of WHITE vertices
▪ Classification of Edges in 𝐺
▪ Tree Edge
▪ Back Edge
▪ Forward Edge
▪ Cross Edge
47
▪ Parenthesis Theorem
▪ Parenthesis structure: represent the discovery of vertex 𝑢 with a left
parenthesis “(𝑢” and represent its finishing by a right parenthesis
“𝑢)”. In DFS, the parentheses are properly nested.
Properly nested: (x (y y) x)
Not properly nested: (x (y x) y)

Proof in textbook p. 608

48
▪ White Path Theorem
▪ In a DFS forest of a directed or undirected graph 𝐺 = 𝑉, 𝐸 ,
▪ vertex 𝑣 is a descendant of vertex 𝑢 in the forest  at the time
𝑢. 𝑑 that the search discovers 𝑢, there is a path from 𝑢 to 𝑣 in 𝐺
consisting entirely of WHITE vertices
▪ Proof.
▪
▪ Since 𝑣 is a descendant of 𝑢, 𝑢. 𝑑 < 𝑣. 𝑑
▪ Hence, 𝑣 is WHITE at time 𝑢. 𝑑
▪ In fact, since 𝑣 can be any descendant of 𝑢, any vertex on the path from 𝑢
to 𝑣 are WHITE at time 𝑢. 𝑑
▪  (textbook p. 608)

49
▪ Classification of Edges in 𝐺
▪ Tree Edge (GRAY to WHITE)
▪ Edges in the DFS forest
▪ Found when encountering a new vertex 𝑣 by exploring 𝑢, 𝑣
▪ Back Edge (GRAY to GRAY)
▪ 𝑢, 𝑣 , from descendant 𝑢 to ancestor 𝑣 in a DFS tree
▪ Forward Edge (GRAY to BLACK)
▪ 𝑢, 𝑣 , from ancestor 𝑢 to descendant 𝑣. Not a tree edge.
▪ Cross Edge (GRAY to BLACK)
▪ Any other edge between trees or subtrees. Can go between vertices in
same DFS tree or in different DFS trees
In an undirected graph, back edge = forward edge.
To avoid ambiguity, classify edge as the first type in the list that applies.
50
▪ Edge classification by the color of 𝑣 when visiting 𝑢, 𝑣
▪ WHITE: tree edge
▪ GRAY: back edge
▪ BLACK: forward edge or cross edge
▪ 𝑢. 𝑑 < 𝑣. 𝑑  forward edge
▪ 𝑢. 𝑑 > 𝑣. 𝑑  cross edge

Theorem 22.10
In DFS of an undirected graph, there are only tree edges and back edges
without forward and cross edge.

Why?
51
▪ Connected Components
▪ Strongly Connected Components
▪ Topological Sort

52
53
▪ Input: a graph 𝐺 = 𝑉, 𝐸
▪ Output: a connected component of 𝐺
▪ a maximal subset 𝑈 of 𝑉 s.t. any two nodes in 𝑈 are connected in 𝐺

54
Why must the connected components of a graph be disjoint?
 1 7
5

2 8 9

6
3
4

10
Time Complexity:

BFS and DSF both find the connected components with the same complexity
55
56
57

Textbook Chapter 22.5 – Strongly connected components

▪ Input: a directed graph 𝐺 = 𝑉, 𝐸
▪ Output: a connected component of 𝐺
▪ a maximal subset 𝑈 of 𝑉 s.t. any two nodes in 𝑈 are reachable in 𝐺

Why must the strongly

1 connected components
7 of a graph be disjoint?
4

2
8
5
6
3
58
▪ Step 1: Run DFS on 𝐺 to obtain the finish time 𝑣. 𝑓 for 𝑣 ∈ 𝑉.
▪ Step 2: Run DFS on the transpose of 𝐺 where the vertices 𝑉 are
processed in the decreasing order of their finish time.
▪ Step 3: output the vertex partition by the second DFS

59
1 1

4 4

2 2

5 5
6 6
3 3

60
 1

2 1

3 4 3

4
6
5
5
6

61
 G 𝑣

𝑤
C

Lemma
Let 𝐶 be the strongly connected component of 𝐺 (and 𝐺 𝑇 ) that contains
the node 𝑢 with the largest finish time 𝑢. 𝑓. Then 𝐶 cannot have any
incoming edge from any node of 𝐺 not in 𝐶.
▪ Proof by contradiction
▪ Assume that 𝑣, 𝑤 is an incoming edge to 𝐶.
▪ Since 𝐶 is a strongly connected component of 𝐺, there cannot be any
path from any node of 𝐶 to 𝑣 in 𝐺.
▪ Therefore, the finish time of 𝑣 has to be larger than any node in 𝐶,
including 𝑢.  𝑣. 𝑓 > 𝑢. 𝑓, contradiction

62
Theorem
By continuing the process from the vertex 𝑢∗ whose finish time 𝑢∗ . 𝑓 is
the largest excluding those in 𝐶, the algorithm returns the strongly
connected components.
▪ Practice to prove using induction

G GT

𝑢 𝑢

C C
63
1

4
6
5

64
1

4
6
5

65
▪ Step 1: Run DFS on 𝐺 to obtain the finish time 𝑣. 𝑓 for 𝑣 ∈ 𝑉.
▪ Step 2: Run DFS on the transpose of 𝐺 where the vertices 𝑉 are
processed in the decreasing order of their finish time.
▪ Step 3: output the vertex partition by the second DFS

Time Complexity:

66
67
68

Textbook Chapter 22.4 – Topological sort

1 1

3 3

2 2

6 6
5 5
4 4

69
▪ Definition
▪ a directed graph without any directed cycle

6
5
4

70
▪ Taking courses should follow the specific order
▪ How to find a course taking order?

微積分上 微積分下 機率
計算機網路

計概 計組 作業系統

計程 資料結構 演算法

71
▪ Input: a directed acyclic graph 𝐺 = (𝑉, 𝐸)
▪ Output: a linear order of 𝑉 s.t. all edges of 𝐺 going from lower-
indexed nodes to higher-indexed nodes

a
f a b c e d d

a f b c e d e
f
c
72
▪ Run DFS on the input DAG G.
▪ Output the nodes in decreasing order of their finish time.
DFS(G) DFS-Visit(G, u)
for each vertex u in G.V time = time + 1
u.color = WHITE u.d = time
u.pi = NIL u.color = GRAY
time = 0 for each v in G.Adj[u] (outgoing)
for each vertex u in G.V if v.color == WHITE
if u.color == WHITE v.pi = u
DFS-VISIT(G, u) DFS-VISIT(G, v)
u.color = BLACK
time = time + 1
u.f = time // finish time

73
Example Illustration
5

f a b c e d a
1
d

4 b 2

e
f
c
3
6

74
Example Illustration
6

a
1
d

4 b 2

e
a f b c e d f
c
3
5

75
▪ Run DFS on the input DAG G.
▪ Output the nodes in decreasing order of their finish time.
▪ As each vertex is finished, insert it onto the front of a linked list
▪ Return the linked list of vertices
DFS-Visit(G, u)
Time Complexity: time = time + 1
u.d = time
u.color = GRAY
DFS(G) for each v in G.Adj[u]
for each vertex u in G.V
if v.color == WHITE
u.color = WHITE
u.pi = NIL v.pi = u
time = 0 DFS-VISIT(G, v)
for each vertex u in G.V u.color = BLACK
if u.color == WHITE time = time + 1
DFS-VISIT(G, u) u.f = time // finish time
76
Lemma 22.11
A directed graph is acyclic  a DFS yields no back edges.
▪ Proof
▪ : suppose there is a back edge 𝑢, 𝑣
▪ 𝑣 is an ancestor of 𝑢 in DFS forest
▪ There is a path from 𝑣 to 𝑢 in 𝐺 and 𝑢, 𝑣 completes the cycle
▪  : suppose there is a cycle 𝑐
▪ Let 𝑣 be the first vertex in 𝑐 to be discovered and 𝑢 is a predecessor of 𝑣 in 𝑐
▪ Upon discovering 𝑣 the whole cycle from 𝑣 to 𝑢 is WHITE
▪ At time 𝑣. 𝑑, the vertices of 𝑐 form a path of white vertices from 𝑣 to 𝑢
▪ By the white-path theorem, vertex 𝑢 becomes a descendant of 𝑣 in the DFS forest
▪ Therefore, 𝑢, 𝑣 is a back edge

77
Theorem 22.12
The algorithm produces a topological sort of the input DAG. That is, if
𝑢, 𝑣 is a directed edge (from 𝑢 to 𝑣) of 𝐺, then 𝑢. 𝑓 > 𝑣. 𝑓.
▪ Proof
▪ When 𝑢, 𝑣 is being explored, 𝑢 is GRAY and there are three cases for 𝑣:
▪ Case 1 – GRAY
▪ 𝑢, 𝑣 is a back edge (contradicting Lemma 22.11), so 𝑣 cannot be GRAY
▪ Case 2 – WHITE
▪ 𝑣 becomes descendant of 𝑢
▪ 𝑣 will be finished before 𝑢
▪ Case 3 – BLACK
▪ 𝑣 is already finished
78
79
▪ Since cycle detection becomes back edge detection (Lemma
22.11), DFS can be used to test whether a graph is a DAG
▪ Is there a topological order for cyclic graphs?
▪ Given a topological order, is there always a DFS traversal
that produces such an order?

80
81
82

Important announcement will be sent to @ntu.edu.tw mailbox

& post to the course website

Course Website: http://ada17.csie.org

Email: ada-ta@csie.ntu.edu.tw