Algorithms & data structures

Graph
Damjan Strnad
 2

Graf
●
a graph is a data structure G=(V,E) consisting of a set of
nodes V and a set E of edges (i.e., node pairs) between
nodes
●
example: G=({a,b,c,d,e},{(a,b),(a,c),(b,d),(b,e),(c,d),(d,e)})
a b

c d

●
linked lists and trees are special cases of graphs
a

a b d c b c

d e
 3

Graf
●
a graph can be undirected (edges are unordered pairs of
nodes) or directed (first node of the edge is the source and
the second node is the destination)
●
example: G=({a,b,c,d,e},{(a,b),(a,c),(c,a),(d,b),(b,e),(c,d),e,d)})
– the order of nodes in edge
a b
notation is now important
e

●
a complete graph is a graph in c d

which every node is connected to every other node:

– there are n(n+1)/2 edges in a complete graph with n nodes
●
the degree of a node is the number of its neighbors:
– in a directed graph (digraph) we distinguish node
in-degree and out-degree
 4

Graph
● a path in a graph is every sequence of edges (v1,v2),
(v2,v3), … , (vr-1,vr)
●
a simple path is a path that visits each intermediate node
only once
●
a closed simple path is a cycle
●
a cycle going through all nodes of a graph is a
Hamiltonian cycle
●
an acyclic graph is a graph that contains no cycles
●
a graph is connected, if there is a path between every
pair of nodes
●
a graph is weighted, if edges have assigned weights or
costs; w(u,v) is a weight of edge from u to v
 5

Graph representation
●
representation with adjacency lists:
– suitable for sparse graphs for which |E|≪|V|2
– consists of an array Adj which for every graph node u∈V
stores a pointer to a linked list of node‘s neighbors**
– for weighted graphs the edge weight is stored in the nodes
of the linked list
– space complexity for a graph with n nodes is
S(n)=O(|V|+|E|)
A
1 2 1 2 5
2 1 3 4 5
3
3 2 4 5
4 2 3 5
5 4
5 1 2 3 4
 6

Graph representation
●
representation with adjacency lists:
– suitable for sparse graphs for which |E|≪|V|2
– consists of an array Adj which for every graph node u∈V
stores a pointer to a linked list of node‘s neighbors**
– for weighted graphs the edge weight is stored in the nodes
of the linked list
– space complexity for a graph with n nodes is
S(n)=O(|V|+|E|)
A
1 2 1 2
2 3
3
3 5
4 2 3
5 4
5 1 2 4
 7

Graph representation
●
representation with adjacency lists:
– suitable for sparse graphs for which |E|≪|V|2
– consists of an array Adj which for every graph node u∈V
stores a pointer to a linked list of node‘s neighbors**
– for weighted graphs the edge weight is stored in the nodes
of the linked list
– space complexity for a graph with n nodes is
S(n)=O(|V|+|E|)
3 A
1 2 1 2 3
1
7 2 3 1
1 4 3
3 3 5 3
2 4 2 4 3 2
5 4
8 5 1 1 2 7 4 8
 8

Graph representation
●
representation with adjacency matrix:
– suitable for dense graphs for which |E|≈|V|2
– graph nodes are enumerated using indices from 1 to |V|
– adjacency matrix A is of size |V|×|V|, where element aij
determines if there is an edge going from node i to node j:

{
aij = 1 , if (i , j)∈E
0 , otherwise
– space complexity is S(n)=Θ(|V|2)

[ ]
1 2 0 1 0 0 1
1 0 1 1 1
3 A= 0 1 0 1 1
0 1 1 0 1
5 4 1 1 1 1 0
 9

Graph representation
●
representation with adjacency matrix:
– suitable for dense graphs for which |E|≈|V|2
– graph nodes are enumerated using indices from 1 to |V|
– adjacency matrix A is of size |V|×|V|, where element aij
determines if there is an edge going from node i to node j:

{
aij = 1 , if (i , j)∈E
0 , otherwise
– space complexity is S(n)=Θ(|V|2)

[ ]
1 2
0 1 0 0 0
0 0 1 0 0
3 A= 0 0 0 0 1
0 1 1 0 0
5 4 1 1 0 1 0
 10

Graph representation
●
representation with adjacency matrix:
– for weighted graphs we use:

{
aij = w (i , j), if (i , j)∈E
∞ , otherwise

[ ]
3
1 2 ∞ 3 ∞ ∞ ∞
1
∞ ∞ 1 ∞ ∞
1 7 4 A= ∞ ∞ ∞ ∞ 3
3
3 ∞ 4 2 ∞ ∞
2
5 4 1 7 ∞ 8 ∞
8
 11

Graph representation
●
representation with edge list:
– suitable for very sparse graphs and algorithms that iterate over
edges (Prim’s algorithm)
– the list stores edges as pairs (u,v)
– for weighted graphs the list elements are triplets (u,v,w)
– the space complexity is S(n)=O(|E|)
1 2

5 4

1 2 1 5 2 3 2 4 2 5 3 4 3 5 4 5
 12

Graph representation
●
representation with edge list:
– suitable for very sparse graphs and algorithms that iterate over
edges (Prim’s algorithm)
– the list stores edges as pairs (u,v)
– for weighted graphs the list elements are triplets (u,v,w)
– the space complexity is S(n)=O(|E|)
3
1 2
1

1 7 4 3
3
2
5 4
8

1 2 3 5 1 1 4 2 4 5 4 8 2 3 1 5 2 7 3 5 3 4 3 2
 13

Breadth-first search
●
breadth-first search (BFS) is one of basic algorithms for
searching in graphs:
– we are given a graph G=(V,E) with source node s
– the algorithm systematically follows graph edges and
discovers nodes which are reachable from s
– the first found path from s to arbitrary node v is also the
shortest one if the cost of all edges is the same
– the basic operation is node expansion, where we record
all node‘s successors which have not been recorded before
– the node‘s color shows its status:
●
white nodes have not yet been visited (initial state of all
nodes except source node s)
●
gray nodes have already been visited
●
black nodes have already been expanded (i.e., all
successors have already been visited)
 14

Breadth-first search
● for each node v we store its color (color[v]), pointer to
parent (parent[v]) and distance (i.e., the number of edges,
d[v]) from source s
●
expansion sequence is determined by position in a queue, into
which the newly generated (i.e., gray) nodes are added
BREADTH-FIRST-SEARCH(G,s)
for each node u ∈ V – {s} do
color[u] ← WHITE, d[u] ← ∞, parent[u] ← NIL
color[s] ← GRAY, d[s] ← 0, parent[s] ← NIL
Q ← ∅
ENQUEUE(Q,s)
while Q ≠ ∅ do % while queue Q is not empty
u ← DEQUEUE(Q) % remove first node in queue
for each node v ∈ Adj[u] do % check all neighbours of u
if color[v] = WHITE then
color[v] ← GRAY, d[v] ← d[u] + 1, parent[v] ← u
ENQUEUE(Q,v)
color[u] ← BLACK % node u is expanded
 15

Breadth-first search – example

●
let the source node be 2 1 7
●
assume the neighbors are 6
2 8
enqueued by increasing node
labels when some node is 4
5
3
expanded
●
inicialization: Q 2
– all nodes are colored white,
their distance from source set u color[u] parent[u] d[u]
to ∞, and pointers to parents 1 w NIL ∞
set to NIL 2 g NIL 0
3 w NIL ∞
– the source node is colored 4 w NIL ∞
gray, assigned distance 0 and 5 w NIL ∞
enqueued in Q 6 w NIL ∞
7 w NIL ∞
8 w NIL ∞
 16

Breadth-first search – example

●
1. iteration: expand node 2* 1 7
6
2 8
5
3 4

Q 2
u color[u] parent[u] d[u]
1 w NIL ∞
2 g NIL 0
3 w NIL ∞
4 w NIL ∞
5 w NIL ∞
6 w NIL ∞
7 w NIL ∞
8 w NIL ∞
 17

Breadth-first search – example

●
1. iteration: expand node 2* 1 7
6
2 8
5
3 4

Q 1 3 8
u color[u] parent[u] d[u]
1 g 2 1
2 b NIL 0
3 g 2 1
4 w NIL ∞
5 w NIL ∞
6 w NIL ∞
7 w NIL ∞
8 g 2 1
 18

Breadth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 1* 6
2 8
5
3 4

Q 1 3 8
u color[u] parent[u] d[u]
1 g 2 1
2 b NIL 0
3 g 2 1
4 w NIL ∞
5 w NIL ∞
6 w NIL ∞
7 w NIL ∞
8 g 2 1
 19

Breadth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 1* 6
2 8
5
3 4

Q 3 8 7
u color[u] parent[u] d[u]
1 b 2 1
2 b NIL 0
3 g 2 1
4 w NIL ∞
5 w NIL ∞
6 w NIL ∞
7 g 1 2
8 g 2 1
 20

Breadth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 1 6
2 8
●
3. iteration: expand node 3*
5
3 4

Q 3 8 7
u color[u] parent[u] d[u]
1 b 2 1
2 b NIL 0
3 g 2 1
4 w NIL ∞
5 w NIL ∞
6 w NIL ∞
7 g 1 2
8 g 2 1
 21

Breadth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 1 6
2 8
●
3. iteration: expand node 3*
5
3 4

Q 8 7 4
u color[u] parent[u] d[u]
1 b 2 1
2 b NIL 0
3 b 2 1
4 g 3 2
5 w NIL ∞
6 w NIL ∞
7 g 1 2
8 g 2 1
 22

Breadth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 1 6
2 8
●
3. iteration: expand node 3
5
●
4. iteration: expand node 8* 3 4

Q 8 7 4
u color[u] parent[u] d[u]
1 b 2 1
2 b NIL 0
3 b 2 1
4 g 3 2
5 w NIL ∞
6 w NIL ∞
7 g 1 2
8 g 2 1
 23

Breadth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 1 6
2 8
●
3. iteration: expand node 3
5
●
4. iteration: expand node 8* 3 4

Q 7 4
u color[u] parent[u] d[u]
1 b 2 1
2 b NIL 0
3 b 2 1
4 g 3 2
5 w NIL ∞
6 w NIL ∞
7 g 1 2
8 b 2 1
 24

Breadth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 1 6
2 8
●
3. iteration: expand node 3
5
●
4. iteration: expand node 8 3 4
●
5. iteration: expand node 7*
Q 7 4
u color[u] parent[u] d[u]
1 b 2 1
2 b NIL 0
3 b 2 1
4 g 3 2
5 w NIL ∞
6 w NIL ∞
7 g 1 2
8 b 2 1
 25

Breadth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 1 6
2 8
●
3. iteration: expand node 3
5
●
4. iteration: expand node 8 3 4
●
5. iteration: expand node 7*
Q 4 6
u color[u] parent[u] d[u]
1 b 2 1
2 b NIL 0
3 b 2 1
4 g 3 2
5 w NIL ∞
6 g 7 3
7 b 1 2
8 b 2 1
 26

Breadth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 1 6
2 8
●
3. iteration: expand node 3
5
●
4. iteration: expand node 8 3 4
●
5. iteration: expand node 7
●
6. iteration: expand node 4* Q 4 6
u color[u] parent[u] d[u]
1 b 2 1
2 b NIL 0
3 b 2 1
4 g 3 2
5 w NIL ∞
6 g 7 3
7 b 1 2
8 b 2 1
 27

Breadth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 1 6
2 8
●
3. iteration: expand node 3
5
●
4. iteration: expand node 8 3 4
●
5. iteration: expand node 7
●
6. iteration: expand node 4* Q 6 5
u color[u] parent[u] d[u]
1 b 2 1
2 b NIL 0
3 b 2 1
4 b 3 2
5 g 4 3
6 g 7 3
7 b 1 2
8 b 2 1
 28

Breadth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 1 6
2 8
●
3. iteration: expand node 3
5
●
4. iteration: expand node 8 3 4
●
5. iteration: expand node 7
●
6. iteration: expand node 4 Q 6 5
●
7. iteration: expand node 6* u color[u] parent[u] d[u]
1 b 2 1
2 b NIL 0
3 b 2 1
4 b 3 2
5 g 4 3
6 g 7 3
7 b 1 2
8 b 2 1
 29

Breadth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 1 6
2 8
●
3. iteration: expand node 3
5
●
4. iteration: expand node 8 3 4
●
5. iteration: expand node 7
●
6. iteration: expand node 4 Q 5
●
7. iteration: expand node 6* u color[u] parent[u] d[u]
1 b 2 1
2 b NIL 0
3 b 2 1
4 b 3 2
5 g 4 3
6 b 7 3
7 b 1 2
8 b 2 1
 30

Breadth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 1 6
2 8
●
3. iteration: expand node 3
5
●
4. iteration: expand node 8 3 4
●
5. iteration: expand node 7
●
6. iteration: expand node 4 Q 5
●
7. iteration: expand node 6 u color[u] parent[u] d[u]
●
8. iteration: expand node 5* 1 b 2 1
2 b NIL 0
3 b 2 1
4 b 3 2
5 g 4 3
6 b 7 3
7 b 1 2
8 b 2 1
 31

Breadth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 1 6
2 8
●
3. iteration: expand node 3
5
●
4. iteration: expand node 8 3 4
●
5. iteration: expand node 7
●
6. iteration: expand node 4 Q
●
7. iteration: expand node 6 u color[u] parent[u] d[u]
●
8. iteration: expand node 5* 1 b 2 1
2 b NIL 0
3 b 2 1
4 b 3 2
5 b 4 3
6 b 7 3
7 b 1 2
8 b 2 1
 32

Breadth-first search – example

●
the working of BFS can be summarized in a table
iteration expanded node u parent[u] d[u]
1 2 NIL 0
2 1 2 1
3 3 2 1
4 8 2 1
5 7 1 2
6 4 3 2
7 6 7 3
8 5 4 3
 33

Breadth-first search – example

●
with the help of pointers to parents we can build the search
tree of BFS
u color[u] parent[u] d[u]
2 1 b 2 1
2 b NIL 0
1 3 8 3 b 2 1
4 b 3 2
7 4 5 b 4 3
6 b 7 3
6 5 7 b 1 2
8 b 2 1
 34

Depth-first search
●
depth-first search (DFS) is similar to breadth-first search
but uses a stack instead of a queue for storing gray nodes
DEPTH-FIRST-SEARCH(G,s)
for each node u ∈ V – {s} do
color[u] ← WHITE, d[u] ← ∞, parent[u] ← NIL
color[s] ← GRAY, d[s] ← 0, parent[s] ← NIL
Q ← ∅
PUSH(S,s)
while S ≠ ∅ do % while stack S is not empty
u ← POP(S) % remove node on top of stack
for each node v ∈ Adj[u] do % check all neighbors of u
if color[v] = WHITE then
color[v] ← GRAY, d[v] ← d[u] + 1, parent[v] ← u
PUSH(S,v)
color[u] ← BLACK % node u is expanded
 35

Depth-first search – example

●
let the source node be 2 1 7
●
assume the neighbors are 6
2 8
pushed onto stack by
increasing node labels when 4
5
3
some node is expanded
●
inicialization:
u color[u] parent[u] d[u]
– all nodes are colored white,
1 w NIL ∞
their distance from source set
2 g NIL 0
to ∞, and pointers to parents 3 w NIL ∞
set to NIL 4 w NIL ∞
– the source node is colored 5 w NIL ∞ 2
gray, assigned distance 0 and 6 w NIL ∞ S
enqueued in Q 7 w NIL ∞
8 w NIL ∞
 36

Depth-first search – example

●
1. iteration: expand node 2* 1 7
6
2 8
5
3 4

u color[u] parent[u] d[u]

1 w NIL ∞
2 g NIL 0
3 w NIL ∞
4 w NIL ∞
5 w NIL ∞ 2
6 w NIL ∞ S
7 w NIL ∞
8 w NIL ∞
 37

Depth-first search – example

●
1. iteration: expand node 2* 1 7
6
2 8
5
3 4

u color[u] parent[u] d[u]

1 g 2 1
2 b NIL 0
8
3 g 2 1
4 w NIL ∞ 3
5 w NIL ∞ 1
6 w NIL ∞ S
7 w NIL ∞
8 g 2 1
 38

Depth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 8* 6
2 8
5
3 4

u color[u] parent[u] d[u]

1 g 2 1
2 b NIL 0
8
3 g 2 1
4 w NIL ∞ 3
5 w NIL ∞ 1
6 w NIL ∞ S
7 w NIL ∞
8 g 2 1
 39

Depth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 8* 6
2 8
5
3 4

u color[u] parent[u] d[u]

1 g 2 1 7
2 b NIL 0
4
3 g 2 1
4 g 8 2 3
5 w NIL ∞ 1
6 w NIL ∞ S
7 g 8 2
8 b 2 1
 40

Depth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 8 6
2 8
●
3. iteration: expand node 7*
5
3 4

u color[u] parent[u] d[u]

1 g 2 1 7
2 b NIL 0
4
3 g 2 1
4 g 8 2 3
5 w NIL ∞ 1
6 w NIL ∞ S
7 g 8 2
8 b 2 1
 41

Depth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 8 6
2 8
●
3. iteration: expand node 7*
5
3 4

u color[u] parent[u] d[u]

1 g 2 1 6
2 b NIL 0
4
3 g 2 1
4 g 8 2 3
5 w NIL ∞ 1
6 g 7 3 S
7 b 8 2
8 b 2 1
 42

Depth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 8 6
2 8
●
3. iteration: expand node 7
5
●
4. iteration: expand node 6* 3 4

u color[u] parent[u] d[u]

1 g 2 1 6
2 b NIL 0
4
3 g 2 1
4 g 8 2 3
5 w NIL ∞ 1
6 g 7 3 S
7 b 8 2
8 b 2 1
 43

Depth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 8 6
2 8
●
3. iteration: expand node 7
5
●
4. iteration: expand node 6* 3 4

u color[u] parent[u] d[u]

1 g 2 1 5
2 b NIL 0
4
3 g 2 1
4 g 8 2 3
5 g 6 4 1
6 b 7 3 S
7 b 8 2
8 b 2 1
 44

Depth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 8 6
2 8
●
3. iteration: expand node 7
5
●
4. iteration: expand node 6 3 4
●
5. iteration: expand node 5*
u color[u] parent[u] d[u]
1 g 2 1 5
2 b NIL 0
4
3 g 2 1
4 g 8 2 3
5 g 6 4 1
6 b 7 3 S
7 b 8 2
8 b 2 1
 45

Depth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 8 6
2 8
●
3. iteration: expand node 7
5
●
4. iteration: expand node 6 3 4
●
5. iteration: expand node 5*
u color[u] parent[u] d[u]
1 g 2 1
2 b NIL 0
4
3 g 2 1
4 g 8 2 3
5 b 6 4 1
6 b 7 3 S
7 b 8 2
8 b 2 1
 46

Depth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 8 6
2 8
●
3. iteration: expand node 7
5
●
4. iteration: expand node 6 3 4
●
5. iteration: expand node 5
●
6. iteration: expand node 4* u color[u] parent[u] d[u]
1 g 2 1
2 b NIL 0
4
3 g 2 1
4 g 8 2 3
5 b 6 4 1
6 b 7 3 S
7 b 8 2
8 b 2 1
 47

Depth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 8 6
2 8
●
3. iteration: expand node 7
5
●
4. iteration: expand node 6 3 4
●
5. iteration: expand node 5
●
6. iteration: expand node 4* u color[u] parent[u] d[u]
1 g 2 1
2 b NIL 0
3 g 2 1
4 b 8 2 3
5 b 6 4 1
6 b 7 3 S
7 b 8 2
8 b 2 1
 48

Depth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 8 6
2 8
●
3. iteration: expand node 7
5
●
4. iteration: expand node 6 3 4
●
5. iteration: expand node 5
●
6. iteration: expand node 4 u color[u] parent[u] d[u]
●
7. iteration: expand node 3* 1 g 2 1
2 b NIL 0
3 g 2 1
4 b 8 2 3
5 b 6 4 1
6 b 7 3 S
7 b 8 2
8 b 2 1
 49

Depth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 8 6
2 8
●
3. iteration: expand node 7
5
●
4. iteration: expand node 6 3 4
●
5. iteration: expand node 5
●
6. iteration: expand node 4 u color[u] parent[u] d[u]
●
7. iteration: expand node 3* 1 g 2 1
2 b NIL 0
3 b 2 1
4 b 8 2
5 b 6 4 1
6 b 7 3 S
7 b 8 2
8 b 2 1
 50

Depth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 8 6
2 8
●
3. iteration: expand node 7
5
●
4. iteration: expand node 6 3 4
●
5. iteration: expand node 5
●
6. iteration: expand node 4 u color[u] parent[u] d[u]
●
7. iteration: expand node 3 1 g 2 1
2 b NIL 0
●
8. iteration: expand node 1* 3 b 2 1
4 b 8 2
5 b 6 4 1
6 b 7 3 S
7 b 8 2
8 b 2 1
 51

Depth-first search – example

●
1. iteration: expand node 2 1 7
●
2. iteration: expand node 8 6
2 8
●
3. iteration: expand node 7
5
●
4. iteration: expand node 6 3 4
●
5. iteration: expand node 5
●
6. iteration: expand node 4 u color[u] parent[u] d[u]
●
7. iteration: expand node 3 1 b 2 1
2 b NIL 0
●
8. iteration: expand node 1* 3 b 2 1
4 b 8 2
5 b 6 4
6 b 7 3 S
7 b 8 2
8 b 2 1
 52

Depth-first search – example

●
the working of DFS can be summarized in a table
iteration expanded node u parent[u] d[u]
1 2 NIL 0
2 8 2 1
3 7 8 2
4 6 7 3
5 5 6 4
6 4 8 2
7 3 2 1
8 1 2 1
 53

Depth-first search – example

●
with the help of pointers to parents we can build the search
tree of DFS
u color[u] parent[u] d[u]
2
1 b 2 1
2 b NIL 0
1 3 8
3 b 2 1
4 b 8 2
7 4 5 b 6 4
6 b 7 3
6 7 b 8 2
8 b 2 1
5
 54

Time complexity of BFS and DFS

●
initialization requires O(|V|) time
●
inserting and removing one node from a queue requires
O(1) time, for all nodes O(|V|) time
●
if the graph is described using adjacency lists, the list is
traversed once for each node, but requires O(|E|) time
for all nodes; total time complexity is in this case
T(n) = O(|V| + |E|)
●
if the graph is described using adjacency matrix, the
algorithm time complexity is T(n) = O(|V|2)