Data Structures and
Algorithms
Live Class 7
Heikki Peura
[email protected]
Announcements
Homework 3 is on the Hub
• Algorithms, libraries, . . .
• Complete in your assigned study groups
Today
1. Representing networks as graphs
2. Graph search
2 / 21
Graph problems
Suppose you’re travelling around Europe
• Know flights and their prices
• Goal: fewest stops between cities, cheapest price, ...?
3 / 21
Defining graphs
70
30 50
OSL ARN HEL
80
120
50
190
110
LON AMS RIX
40
90
A set of nodes (or vertices) connected by edges
4 / 21
Defining graphs
70
30 50
OSL ARN HEL
80
120
50
190
110
LON AMS RIX
40
90
A set of nodes (or vertices) connected by edges
Edges may or may not be directed → directed or undirected graph
(Twitter/X vs Facebook?)
4 / 21
Defining graphs
70
30 50
OSL ARN HEL
80
120
50
190
110
LON AMS RIX
40
90
A set of nodes (or vertices) connected by edges
Edges may or may not be directed → directed or undirected graph
(Twitter/X vs Facebook?)
Edges may have weights (eg prices): weighted graph
4 / 21
Pics: Gmaps, Uber, Hautsch et al.
Graphs are everywhere
Represent situations with pairwise relationships
Many applications, some not obvious:
• Roads, social networks, the web, financial markets,
complex projects etc
6 / 21
Graph: collection of vertices and edges
Graph G = (V , E): V = set of n vertices, E = set of m edges
• Edges can be directed or undirected
• Connected graph: there is a path between any two vertices
7 / 21
Graph: collection of vertices and edges
Graph G = (V , E): V = set of n vertices, E = set of m edges
• Edges can be directed or undirected
• Connected graph: there is a path between any two vertices
How many edges can there be in a connected graph with n vertices?
7 / 21
Graph: collection of vertices and edges
Graph G = (V , E): V = set of n vertices, E = set of m edges
• Edges can be directed or undirected
• Connected graph: there is a path between any two vertices
How many edges can there be in a connected graph with n vertices?
7 / 21
Graph: collection of vertices and edges
Graph G = (V , E): V = set of n vertices, E = set of m edges
• Edges can be directed or undirected
• Connected graph: there is a path between any two vertices
How many edges can there be in a connected graph with n vertices?
Sparse vs. dense graphs:
• Between n − 1 vs n(n − 1)/2 edges
7 / 21
Graph: collection of vertices and edges
Graph G = (V , E): V = set of n vertices (nodes), E = set of m edges
8 / 21
Graph: collection of vertices and edges
Graph G = (V , E): V = set of n vertices (nodes), E = set of m edges
How to represent a graph?
• Adjacency matrix: size n × n
• Avw = 1 if there is an edge between vertices v , w
(unweighted)
• n2 space required
8 / 21
Graph: collection of vertices and edges
Graph G = (V , E): V = set of n vertices (nodes), E = set of m edges
How to represent a graph?
• Adjacency matrix: size n × n
• Avw = 1 if there is an edge between vertices v , w
(unweighted)
• n2 space required
LON OSL ARN HEL AMS RIX
LON 0 1 1 1 1 1
OSL 1 0 1 1 0 0
ARN 1 1 0 1 0 0
HEL 1 1 1 0 1 1
AMS 1 0 0 1 0 0
RIX 1 0 0 1 0 0
8 / 21
Graph: adjacency lists
How to represent a graph?
• Adjacency list: connectivity represented using lists of edges
and vertices
• List of vertices (each vertex points to edges starting from it);
list of edges (each edge points to its endpoints)
• Linear space requirement in n (vertices), m (edges)
• In practice: a dictionary in Python
9 / 21
Graph: adjacency lists
How to represent a graph?
• Adjacency list: connectivity represented using lists of edges
and vertices
• List of vertices (each vertex points to edges starting from it);
list of edges (each edge points to its endpoints)
• Linear space requirement in n (vertices), m (edges)
• In practice: a dictionary in Python
• LON: {OSL, ARN, HEL, AMS, RIX}
• OSL: {LON, ARN, HEL}
• ARN: {LON, OSL, HEL}
• HEL: {LON, OSL, ARN, AMS, RIX}
• AMS: {LON, HEL}
• RIX: {LON, HEL}
9 / 21
Graph: adjacency lists
How to represent a graph?
• Adjacency list: connectivity represented using lists of edges
and vertices
• List of vertices (each vertex points to edges starting from it);
list of edges (each edge points to its endpoints)
• Linear space requirement in n (vertices), m (edges)
• In practice: a dictionary in Python
• LON: {OSL, ARN, HEL, AMS, RIX}
• OSL: {LON, ARN, HEL}
• ARN: {LON, OSL, HEL}
• HEL: {LON, OSL, ARN, AMS, RIX}
• AMS: {LON, HEL}
• RIX: {LON, HEL}
Adjacency matrices vs lists:
• Dense vs sparse graphs
• We’ll be using adjacency lists
9 / 21
Graph data structure
Class Digraph – (directed graph) methods:
• __init__: creates a new (empty) Digraph
• add_node(node): adds a node
• add_edge(edge): adds an edge
• children_of(node): returns all edges from node
• has_node(node): returns True if it is in the graph, False otherwise
• __str__: returns a string with all the edges of the graph
Class Graph(Digraph) – undirected graph:
• add_edge(edge): adds an edge and its reverse edge to the graph
• Example of inheritance: Graph does everything that Digraph
does, except that add_edge method is different
10 / 21
Today
1. Representing networks as graphs
2. Graph search
11 / 21
Searching a graph
12 / 21
Searching a graph
Small worlds phenomenon:
• Six degrees of separation (Marconi, Milgram, Guare)
• Hollywood: Kevin Bacon numbers
12 / 21
Example: social network
How to find the fewest steps to someone on a social network?
Idea:
• Me?
• My friends?
• Friends’ friends?
• Friends’ friends’ friends?
• ...
13 / 21
Example: social network
How to find the fewest steps to someone on a social network?
Idea:
• Me?
• My friends?
• Friends’ friends?
• Friends’ friends’ friends?
• ...
John Helena Paul Donald
Jared
Chris
Vicki
13 / 21
Example: social network
How to find the fewest steps to someone on a social network?
Idea:
• Me?
• My friends?
• Friends’ friends?
• Friends’ friends’ friends?
• ...
John Helena Paul Donald
Jared
Chris
Vicki
13 / 21
Example: social network
How to find the fewest steps to someone on a social network?
Idea:
• Me?
• My friends?
• Friends’ friends?
• Friends’ friends’ friends?
• ...
John Helena Paul Donald
Jared
Chris
Vicki
13 / 21
Example: social network
How to find the fewest steps to someone on a social network?
Idea:
• Me?
• My friends?
• Friends’ friends?
• Friends’ friends’ friends?
• ...
John Helena Paul Donald
Jared
Chris
Vicki
13 / 21
Example: social network
How to find the fewest steps to someone on a social network?
Idea:
• Me?
• My friends?
• Friends’ friends?
• Friends’ friends’ friends?
• ...
John Helena Paul Donald
Jared
Chris
Vicki
13 / 21
Breadth-first search (BFS)
• Explore nodes in layers
Plan:
1. How to explore a graph in layers?
2. How much work is it to go through the entire
graph?
3. How to get shortest distances and paths?
14 / 21
Exploring in layers: BFS algorithm
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
15 / 21
Exploring in layers: BFS algorithm
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
What is a queue?
• First in, first out (just like in a cafe)
• Remove from front, insert to back in constant time O(1)
• Here we keep a queue of nodes to visit next
15 / 21
Exploring in layers: BFS algorithm
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
What is a queue?
• First in, first out (just like in a cafe)
• Remove from front, insert to back in constant time O(1)
• Here we keep a queue of nodes to visit next
15 / 21
Exploring in layers: BFS algorithm
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
What is a queue?
• First in, first out (just like in a cafe)
• Remove from front, insert to back in constant time O(1)
• Here we keep a queue of nodes to visit next
15 / 21
Exploring in layers: BFS algorithm
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
What is a queue?
• First in, first out (just like in a cafe)
• Remove from front, insert to back in constant time O(1)
• Here we keep a queue of nodes to visit next
15 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: John
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Helena
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Helena Chris
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Chris
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Chris
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Chris Paul
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Paul
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Paul
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Paul
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Paul Vicki
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Vicki
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Vicki
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Vicki Jared
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Jared
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Jared
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices:
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices:
16 / 21
BFS example
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
John Helena Paul Donald
Jared
Chris
Vicki
Queue of vertices: Donald
16 / 21
How much work is it?
Complete exploration: At the end, any given node v is explored if
and only if there is a path from s to v
17 / 21
How much work is it?
Complete exploration: At the end, any given node v is explored if
and only if there is a path from s to v
Running time?
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
17 / 21
How much work is it?
Complete exploration: At the end, any given node v is explored if
and only if there is a path from s to v
Running time?
Algorithm: BFS(graph G, start node s)
• Initialise: mark all nodes unexplored except s explored
• Use Q = queue data structure; add s to Q
• Main loop: While Q is not empty:
• Remove the first node of Q and call it v
• For each edge (v , w): if w unexplored:
• Mark w explored
• Add w to Q
Work for each of n nodes Work for each of m edges
• •
• •
• •
17 / 21
How to get shortest distances?
• Fewest edges to cross to get from s to v
• This is the same as how many “layers of friends” we have to cross, or
degrees of separation
18 / 21
How to get shortest distances?
• Fewest edges to cross to get from s to v
• This is the same as how many “layers of friends” we have to cross, or
degrees of separation
To get shortest distance, we need to add:
Initialise: dist(v ) = 0 for v = s, ∞ for other v
Main loop: when going through edge (v , w): if w unexplored, set
dist(w) = dist(v ) + 1
18 / 21
How to get shortest distances?
• Fewest edges to cross to get from s to v
• This is the same as how many “layers of friends” we have to cross, or
degrees of separation
To get shortest distance, we need to add:
Initialise: dist(v ) = 0 for v = s, ∞ for other v
Main loop: when going through edge (v , w): if w unexplored, set
dist(w) = dist(v ) + 1
John Helena Paul Donald
Jared
Chris
Vicki
18 / 21
We can also check if the graph is connected
Connected components = the “pieces” of (undirected) graph G
• Check if graph is connected
• Visualise graphs
• Clustering
How to do it:
• If BFS finishes without exploring all nodes, restart it from an
unexplored node
• O(m + n) time (the size of the entire graph)
19 / 21
Graphs are useful
Many problems revolve around transitions from one system state
to another
• Graph search often useful
• Examples: travelling, scheduling, etc.
For unweighted graphs, many problems solved by BFS or DFS -
depth-first search
• Linear time
• DFS explores like a maze — optional problem
• Weighted graphs require different methods
20 / 21
Review
Many problems can be modelled using graphs
• Graph representation: adjacency list or adjacency
matrix
• Graphs can be (un)directed, (un)weighted
• Breadth-first search for shortest paths in unweighted
graphs
Review exercises:
• BFS implementation
• Optional: apply BFS to a Hollywood dataset
• Optional: Dijkstra’s algorithm for weighted graphs
21 / 21
Exercises
Session zip file ses07.zip on the Hub
• HTML instructions
• At some point, you’ll need the ses07.py-file with
skeleton code (open in Spyder)
