Upload
chelsea-roth
View
73
Download
0
Tags:
Embed Size (px)
DESCRIPTION
SI 614 Basic network concepts and intro to Pajek. Lecture 2 Instructor: Lada Adamic. Outline. Basic network metrics Bipartite graphs Graph theory in math Pajek. Network elements: edges. Directed (also called arcs) A -> B A likes B, A gave a gift to B, A is B’s child Undirected - PowerPoint PPT Presentation
Citation preview
School of InformationUniversity of Michigan
SI 614Basic network concepts and intro to Pajek
Lecture 2
Instructor: Lada Adamic
Network elements: edges
Directed (also called arcs) A -> B
A likes B, A gave a gift to B, A is B’s child
Undirected A <-> B or A – B
A and B like each other A and B are siblings A and B are co-authors
Edge attributes weight (e.g. frequency of communication) ranking (best friend, second best friend…) type (friend, relative, co-worker) properties depending on the structure of the rest of the graph:
e.g. betweenness
Directed networks
2
1
1
2
1
2
1
2
1
2
21
1
2
1
2
1
2
12
1
2
1
2
1
2
1
21
2 1
2
1
2
12 1
2
1
2
12
1
2
12
1
2
1 2
12
Ada
Cora
Louise
Jean
Helen
Martha
Alice
Robin
Marion
Maxine
Lena
Hazel Hilda
Frances
Eva
RuthEdna
Adele
Jane
Anna
Mary
Betty
Ella
Ellen
Laura
Irene
girls’ school dormitory dining-table partners (Moreno, The sociometry reader, 1960)
first and second choices shown
Edge weights can have positive or negative values
One gene activates/inhibits another
One person trusting/distrusting another Research challenge:
How does one ‘propagate’ negative feelings in a social network? Is my enemy’s enemy my friend?
Transcription regulatory network in baker’s yeast
Adjacency matrices
Representing edges (who is adjacent to whom) as a matrix Aij = 1 if node i has an edge to node j
= 0 if node i does not have an edge to j
Aii = 0 unless the network has self-loops
Aij = Aji if the network is undirected,or if i and j share a reciprocated edge
ij
i
ij
1
2
3
4
Example:
5
0 0 0 0 0
0 0 1 1 0
0 1 0 1 0
0 0 0 0 1
1 1 0 0 0
A =
Adjacency lists
Edge list 2 3 2 4 3 2 3 4 4 5 5 2 5 1
Adjacency list is easier to work with if network is
large sparse
quickly retrieve all neighbors for a node 1: 2: 3 4 3: 2 4 4: 5 5: 1 2
1
2
3
45
Nodes
Node network properties from immediate connections
indegreehow many directed edges (arcs) are incident on a node
outdegreehow many directed edges (arcs) originate at a node
degree (in or out)number of edges incident on a node
from the entire graph centrality (betweenness, closeness)
outdegree=2
indegree=3
degree=5
Node degree from matrix values
Outdegree =0 0 0 0 0
0 0 1 1 0
0 1 0 1 0
0 0 0 0 1
1 1 0 0 0
A =
n
jijA
1
example: outdegree for node 3 is 2, which we obtain by summing the number of non-zero entries in the 3rd row
Indegree =0 0 0 0 0
0 0 1 1 0
0 1 0 1 0
0 0 0 0 1
1 1 0 0 0
A =
n
iijA
1
example: the indegree for node 3 is 1, which we obtain by summing the number of non-zero entries in the 3rd column
n
iiA
13
n
jjA
13
1
2
3
45
Other node attributes
Homophily: tendency of like individuals to associate with one another
take your pick… geographical location function musical tastes…
Network metrics: degree sequence and degree distribution
Degree sequence: An ordered list of the (in,out) degree of each node
In-degree sequence: [2, 2, 2, 1, 1, 1, 1, 0]
Out-degree sequence: [2, 2, 2, 2, 1, 1, 1, 0]
(undirected) degree sequence: [3, 3, 3, 2, 2, 1, 1, 1]
Degree distribution: A frequency count of the occurrence of each degree
In-degree distribution: [(2,3) (1,4) (0,1)]
Out-degree distribution: [(2,4) (1,3) (0,1)]
(undirected) distribution: [(3,3) (2,2) (1,3)]
0 1 20
1
2
3
4
5
indegree
fre
qu
en
cy
Network metrics: connected components
Strongly connected components Each node within the component can be reached from every other node
in the component by following directed links
Strongly connected components B C D E A G H F
Weakly connected components: every node can be reached from every other node by following links in either direction
A
B
C
DE
FG
H
A
B
C
DE
FG
H
Weakly connected components A B C D E G H F
In undirected networks one talks simply about ‘connected components’
Network metrics: shortest paths
Shortest path (also called a geodesic path) The shortest sequence of links connecting two nodes Not always unique
A and C are connected by 2 shortest paths
A – E – B - C A – E – D - C
Diameter: the largest geodesic distance in the graph
A
B
C
DE
The distance between A and C is the maximum for the graph: 3
Caution: some people use the term ‘diameter’ to be the average shortest path distance, in this class we will use it only to refer to the maximal distance
1
2
2
3
3
Giant components and the web graph
if the largest component encompasses a significant fraction of the graph, it is called the giant component
The bowtie model of the web
The Web is a directed graph: webpages link to other
webpages The connected components
tell us what set of pages can be reached from any other just by surfing (no ‘jumping’ around by typing in a URL or using a search engine)
Broder et al. 1999 – crawl of over 200 million pages and 1.5 billion links.
SCC – 27.5% IN and OUT – 21.5% Tendrils and tubes – 21.5% Disconnected – 8%
image: Mark Levene
bipartite (two-mode) networks
edges occur only between two groups of nodes, not within those groups
for example, we may have individuals and events directors and boards of directors customers and the items they purchase metabolites and the reactions they participate in
going from a bipartite to a one-mode graph
One mode projection two nodes from the first
group are connected if they link to the same node in the second group
some loss of information naturally high
occurrence of cliques
Two-mode networkgroup 1
group 2
Now in matrix notation
Bij = 1 if node i from the first group
links to node j from the second group = 0 otherwise
B is usually not a square matrix! for example: we have n customers and m products
i
j
1 0 0 0
1 0 0 0
1 1 0 0
1 1 1 1
0 0 0 1
B =
Collapsing to a one-mode network
i and k are linked if they both link to j Aik= j Bij Bkj
A= B BT
the transpose of a matrix swaps Bxy and Byx
if B is an nxm matrix, BT is an mxn matrix
i
j=1
k
j=2
B = BT =
1 0 0 0
1 0 0 0
1 1 0 0
1 1 1 1
0 0 0 1
1 1 1 1 0
0 0 1 1 0
0 0 0 1 0
0 0 0 1 1
Matrix multiplication
general formula for matrix multiplication Zij= k Xik Ykj
let Z = A, X = B, Y = BT
1 0 0 0
1 0 0 0
1 1 0 0
1 1 1 1
0 0 0 1
A =
1 1 1 1 0
0 0 1 1 0
0 0 0 1 0
0 0 0 1 1
=
1 1 1 1 0
1 1 1 1 0
1 1 2 2 0
1 1 2 4 1
0 0 0 1 1
1 1
1 2
11 1 1 1 1
1
0
0
= 1*1+1*1 + 1*0 + 1*0= 2
Collapsing a two-mode network to a one mode-network
Assume the nodes in group 1 are people and the nodes in group 2 are movies
The diagonal entries of A give the number of movies each person has seen
The off-diagonal elements of A give the number of movies that both people have seen
A is symmetric
A =
1 1 1 1 0
1 1 1 1 0
1 1 2 2 0
1 1 2 4 1
0 0 0 1 1
1 1
1 2
1
History: Graph theory
Euler’s Seven Bridges of Königsberg – one of the first problems in graph theory
Is there a route that crosses each bridge only once and returns to the starting point?
Eulerian paths
If starting point and end point are the same: only possible if no nodes have an odd degree
each path must visit and leave each shore
If don’t need to return to starting point can have 0 or 2 nodes with an odd degree
Eulerian path: traverse each
edge exactly once
Hamiltonian path: visit
each vertex exactly once
Bi-cliques (cliques in bipartite graphs)
Km,n is the complete bipartite graph with m and n vertices of the two different types
K3,3 maps to the utility graph Is there a way to connect three utilities, e.g. gas, water, electricity to
three houses without having any of the pipes cross?
K3,3
Utility graph
When graphs are not planar
Two graphs are homeomorphic if you can make one into the other by adding a vertex of degree 2
Cliques and complete graphs
Kn is the complete graph (clique) with K vertices each vertex is connected to every other vertex there are n*(n-1)/2 undirected edges
K5 K8K3
Edge contractions defined
A finite graph G is planar if and only if it has no subgraph that is homeomorphic or edge-contractible to the complete graph in five vertices (K5) or the complete bipartite graph K3, 3. (Kuratowski's Theorem)
graph density
Of the connections that may exist between n nodes directed graph
emax = n*(n-1)each of the n nodes can connect to (n-1) other nodes
undirected graphemax = n*(n-1)/2since edges are undirected, count each one only once
What fraction are present? density = e/ emax
For example, out of 12possible connections, this graphhas 7, giving it a density of 7/12 = 0.583
But it is more difficult for a larger networkto achieve the same density
measure not useful for comparing networks of different densities
#s of planar graphs of different sizes
1:1
2:2
3:4
4:11
Every planar graph
has a straight line
embedding
(homework exercise)
examples of trees
In nature trees river networks arteries (or veins, but not both)
Man made sewer system
Computer science binary search trees decision trees (AI)
Network analysis minimum spanning trees
from one node – how to reach all other nodes most quickly may not be unique, because shortest paths are not always unique depends on weight of edges
Using Pajek for exploratory social network analysis
Pajek – (pronounced in Slovenian as Pah-yek) means ‘spider’
website: vlado.fmf.uni-lj.si/pub/networks/pajek/ download application (free) tutorials lectures data sets
Windows only (works on Linux via Wine)
can be installed via NAL in the student lab (DIAD)
helpful book: ‘Exploratory Social Network Analysis with Pajek’ by Wouter de Nooy, Andrej Mrvar and Vladimir Batagelj first 2 chapters are required reading and on cTools
Pajek interface
Drop down list of networks opened or created with pajek. Active is displayed
Drop down list of network partitions by discrete variables, e.g. degree, mode, label
Drop down list of continuous node attributes, e.g. centrality, clustering coefficients
things we’ll use right away
things we’ll use later for clustering
opening a network fileclick on folder icon
to open a file
Save changes to your network, network partitions, etc., if you’d like to keep them
Working with network files in Pajek
The active network, partition, etc is shown on top of the drop down list
Draw the network
Pajek data format
*Vertices 26
1 "Ada" 0.1646 0.2144 0.5000
2 "Cora" 0.0481 0.3869 0.5000
3 "Louise" 0.3472 0.1913 0.5000
..
*Arcs 1 3 2 c Black
..
*Edges 1 2 1 c Black
..
2
1
1Ada
Cora
Louise
number of vertices vertex x,y,z coordinates (optional)
directed edges
undirected edges
from Ada(1) to Louise(3) as
choice “2” and color Black
between Ada(1) to Cora(2) as
choice “1” and color Black
Final project guidelines
Work individually or in groups (up to 4 people) Important dates
Feb. 13th Project proposals due (5%) 1 page abstract & 5 minute class presentation
March 20th Project status report due (5%) 3-6 pages of
result summaries (including figures and tables) plan of remaining work
April 17th in class student presentations of results (5%) April 24th final project reports due (25%)
6-12 pages of related work main results ‘future’ work/extensions
Final Project
Option 1: Analyze a network What it should be
More than just a measurement of the average shortest path, clustering coefficient, and degree distribution
An interpretation of measurement results If applicable:
discovery of community or other structure assortativity motifs weights, thresholds longitudinal data (how the network changes over time)
Visualizations of all or part of the network that point out a particular feature Qualitative comparison with other networks
What it should not be a literature review
The data can be artificially generated or a real-world dataset If you intend to work on data concerning human subjects, you may need
to start an IRB application ASAP
Final Project
Option 2: New network model What it should be
Method for generating a network e.g. preferential attachment optimization wrt. different criteria
Analysis of resulting network comparison with random graphs how do attributes change depending on model parameters
What it should not be an already thoroughly explored model
Final Project
Option 3: Novel algorithm What it should be
An algorithm to analyze the network e.g. clustering or community detection algorithm webpage ranking algorithm
OR a process that is influenced by the network gossip spreading games such as the prisoner’s dilemma
Analysis of algorithm on several different networks
What it should not be an exact replica of an existing algorithm applied to a network where
it has already been studied