Jung M2K

The Java Universal Network/Graph Framework (JUNG): A Brief Tour
Joshua O!Madadhain (UCI)

Danyel Fisher (MSR), Tom Nelson (RABA Technologies)
Scott White (UCI), Yan-Biao Boey (UCI) http://jung.sourceforge.net
What!s it for?
phenomenae pertaining to relationships/links collaborations, citations/hyperlinks, email, biological systems, friendships, telecom networks, ecological and biological systems, Markov processes, Bayes nets...
! !
SUNBELT 2000 270 People & Structural Holes Database to UCINET to Pajek to MAGE
! !
http://jung.sourceforge.net
Programmatic Solution
!
Programmable
Make it easy to do the same thing again
Extensible
If you need a new routine, you can put it in yourself
Library
Can be used in a server, a client, or as part of a larger application
Open Source
Take it apart and put it together some other way Free!
What is JUNG?
Framework for the modeling, analysis, and visualization of graphs in Java
"
"
"
"
"
"
supports most types of graphs/networks separate, exible visualization framework rich attribute structure (metadata) network event handling predicates (subsets, constraints, ltering, ...) extensive (and growing!) library of algorithms
Dependencies
!
Java (1.4.2+) Commons-Collections API (3.1)

http://jakarta.apache.org/commons/collections/ Predicate, set operations, wrappers, data types (BidiMap)
CERN Colt API (1.2.0)

http://dsd.lbl.gov/~hoschek/colt/ matrix operations, statistics
Xerces (2.6.2)
http://xml.apache.org/xerces2-j/index.html XML parsing http://jung.sourceforge.net
Overview
!
Creating a graph Annotating with data Filters Graph drawing Algorithms Demo
Creating a graph
" Load from a le
- Pajek - GraphML (XML)
" Build from scratch

- Create vertices individually - Specify edges
" Generate from an algorithm/model

- Small-world graphs - Power-law graphs
Graph Types
ArchetypeGraph
Hypergraph
AbstractArchetypeGraph
Graph
AbstractSparseGraph
HypergraphBPG UndirectedGraph DirectedGraph
SparseGraph KPartiteGraph
KEY interface abstract class instantiable class A B
B implements A A B
B extends A
UndirectedSparseGraph
DirectedSparseGraph
KPartiteSparseGraph
Vertex Types
ArchetypeVertex
AbstractHyperUnitBPG
Hypervertex
Vertex
HypervertexBPG
AbstractSparseVertex
KEY interface abstract class instantiable class A B
LeanSparseVertex (mixed, parallel)
SimpleSparseVertex (mixed)
SimpleDirectedSparseVertex (directed)
SimpleUndirectedSparseVertex (undirected)
B implements A A B SparseVertex (mixed, parallel) DirectedSparseVertex (directed, parallel) UndirectedSparseVertex (undirected, parallel)
B extends A
Edge Types
ArchetypeEdge
AbstractHyperUnitBPG
Hyperedge
Edge
HyperedgeBPG DirectedEdge KEY interface abstract class instantiable class A B
AbstractSparseEdge UndirectedEdge
B implements A A B DirectedSparseEdge UndirectedSparseEdge
B extends A
Creating a Graph
Graph g = new DirectedSparseGraph(); Vertex v1 = new SparseVertex(); Vertex v2 = new SparseVertex(); g.addVertex( v1 ); g.addVertex( v2 ); Edge e = new Edge( v1, v2 ); g.addEdge( e );
Adding user-dened data

"
"
key-value pairs associated with graphs, edges, vertices easy to write algorithms that operate on relational data
String name = vertex.getUserDatum(NAME); edge.setUserDatum(DATE, new Date());
Copying and Equality

!
copies of vertices and edges are equal to the originals

equals(), hashCode() subgraphs allows experimentation with mutations to graph, without affecting original v2 = v1.getEqualVertex(g2)
Predicates
!
logical expressions, e.g.

v.degree() > 4 e instanceof DirectedEdge
can be used to dene subsets and constraints

k-partite graphs multigraphs, simple graphs, trees, mixedmode graphs, hypergraphs, ...
Using graph lters

!
Extract subgraphs of vertices and edges

All edges with user-data weight > 0.2 All nodes whose location is California All nodes within distance 3.5 of v
Chain lters together Automate complex graph transformations

Visualizing graphs
"
Basic layout algorithms:

Fruchterman-Rheingold Kamada-Kawaii Eades Self-Organizing Maps
"
Plug and play architecture

- Drop in your favorite renderer, layout
"
Provided rendering engine is Swing-based

- (but layouts are generic)
Simple Graph Display

to display an existing graph:
l = new FRLayout(g); r = new PluggableRenderer(); vv = new VisualizationViewer(l,r); jpanel.add(vv);
now, start customizing...
Pluggable Renderer
Customizing
Layout algorithms specify vertex positions advancePositions() Renderers draw vertices and edges.
paintEdge( g, Edge, x1, y1, x2, y2 ) paintVertex( g, Vertex, x, y )
Algorithms
clustering
k-neighborhood connected components k-means clustering
centrality/importance
betweenness, markov, PageRank, hubs-and-authorities, voltage, ...
connectivity maximum flow network distances structural equivalence
ranking
R Almond K Mosurski B Levidow E Hunt D Golinelli G Consonni D Donnell R Kronmal M Levitz A Raftery M Perlman J Gavrin C Volinsky J Hoeting S Hanks C Triggs S Andersson
T Richardson D Madigan D Sornette P Yiou M Latif A Robertson F Lott G Ridgeway T Hastie J Iv
K Laakso M Salmenkivi I Verkamo J Boulicaut A Verkamo M Holsheimer E Nikunen H Toivonen P Orponen M Kersten G Linden H Ahonen M Klemettinen P Kilpelainen O Heinonen M Jaeger H Olander A Korhola E Weydert G Gottlob T Eiter P Ronkainen D Roth H Mannila G Das R Khardon B Bollobas
E Ukkonen
M Friedl B Draper P Utgoff J Fortes N Kapadia T Lane
P Courtier C Lai K Chang M Ghil D Pregibon C Glymour J Kivinen
D Gunopulos
J Mao B Dom
C Brodley D McNeill C Matheus P Chan
K Ide P Prado D Peel
D Pavlov
D Dobkin D Chudova G McLachlan V Tsaoussidis G Piatetsky-Shapiro C Reina J Roden H Badr S Chaudhuiri MS Tartagni I Cadez Gaffney D Mitchell C Fowlkes K Pentikousis S Casadei S Mitter P Kube G Graefe P Bradley C McLaren L Cordero A Gray D Psaltis L Guibas D Hart X Ge J Bouguet R Manduchi G Steckman M Fernandez L Ugarte M Welling O Mangasarian W Einhauser M Munich J Matousek U Fayyad M Henzinger P Stolorz F Yao E Bernardo E Ursella J Erickson M Paterson PBurl Smyth M P Agarwal K Bennett L Goncalves E Friedman D Haussler C Meek D Eppstein Y Song E Demaine E Di Bernardo M Demaine N Amenta P Perona E Keogh E Kuo D Geiger J Rice M Bern G Barequet D Hirschberg W Pratt M Pazzani J Gilbert S Teng M Dickerson E Knill R Fikes A Farquhar M Dillencourt T Leung D Chickering H King G Picci R Beigel R Giancarlo R Frezza B Thiesson R Engelmore G Miller M Weber R Goodman G Cooper G Italiano P Cheeseman J Frank D Wolf I Sim Z Galil M Hearst C Strauss J Stutz S Goodbody S Belongie D Heckerman P Vetter H Greenspan D Wolpert B Kanefsky S Soatto M Sahami R Kohavi E Horvitz W Buntine R Kraft C Stauss K Wheeler W Taylor Z Wang S Payandeh K Tumer M Kawato A Smith S Dumais A Mayer W Macready J Crowcroft A Newton B MacLennan B Tuttle C Sayers J Adams R Shachter A Mamdani A Lapedes L Su K Rommelse J Haas R Paul J Breese A Weigend M Wellman P Sabes Z Ghahramani T Pressburger G Overton J Aaronson B Fischer D Thomas L Saul J Malik C Bishop M Jordan J Kosecka S Singh N Lawrence D Cohn M Meila Q Morris A Barto S Russell R Blasi D Rumelhart R Jacobs T Jaakkola J Weber A Ng P Debevec J seck L Xu P Hrensen E Alpaydin M Atkinson N de Freitas L Daynes K Murphy T PrintezisY Weiss S Spence M Vanter C Taylor D Young S MacDonell P Sallis R Kilgour
clustering
A Jones M Faghihi S Jung I Dryden J Ostrowski D Kriegman D Cooper K Nafasi A Lanitis A Hill S Solloway J Graham J Waterton A Brett O Miglino J Haslam G Edwards T Cootes C Hutchinson E Dimauro P Sozou M Mitchell C Jackson G Page D Cunliffe K Walker
Event Handling
!
addition/removal of vertices/edges
graphs generate events
modication of user data repository

(JUNG) objects generate events
mouse events
can be parsed as selection, panning, zooming, pop-ups, property changes, functions, ...
objects can register as listeners

Statistical Analysis
!
Graph measures:
Degree distributions Average shortest path length, diameter Clustering coefcients
can use existing Java statistical packages

!
CERN Colt Scientic Library Visual Numeric!s JMSL for Java

Conclusion
! !
Provides a common language for graphs Complements rather than replaces other graph packages/network tools A powerful framework for working with graphs with rich attribute structure Ideally suited for building tools/ applications related to network exploration and data mining
Adoption
Users are writing

social network analyses, games, animated graph views, transportation network algorithms, Google Cartography, ...
Email visualization
Newsgroup Networks
Temporal Egocentric Communication Networks
Open-Source Collaboration
LiveJournal networks
Integrated into an application

Freeman!s Southern Women

Shortest Path
Betweenness Centrality
Small Multiples
[Sun, E1]
Thanks
http://jung.sourceforge.net Feel free to just play with it Or join us! Active collaboration is welcome. Supported in part by NSF.

Jung M2K

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Jung M2K

Uploaded by

Copyright:

Available Formats

The Java Universal Network/Graph Framework (JUNG): A Brief Tour

Joshua O!Madadhain (UCI)

Java (1.4.2+) Commons-Collections API (3.1)

CERN Colt API (1.2.0)

" Build from scratch

" Generate from an algorithm/model

HypergraphBPG UndirectedGraph DirectedGraph

KEY interface abstract class instantiable class A B

KEY interface abstract class instantiable class A B

LeanSparseVertex (mixed, parallel)

HyperedgeBPG DirectedEdge KEY interface abstract class instantiable class A B

B implements A A B DirectedSparseEdge UndirectedSparseEdge

Adding user-dened data

String name = vertex.getUserDatum(NAME); edge.setUserDatum(DATE, new Date());

Copying and Equality

copies of vertices and edges are equal to the originals

logical expressions, e.g.

can be used to dene subsets and constraints

Using graph lters

Extract subgraphs of vertices and edges

Chain lters together Automate complex graph transformations

Basic layout algorithms:

Plug and play architecture

Provided rendering engine is Swing-based

Simple Graph Display

now, start customizing...

connectivity maximum flow network distances structural equivalence

T Richardson D Madigan D Sornette P Yiou M Latif A Robertson F Lott G Ridgeway T Hastie J Iv

M Friedl B Draper P Utgoff J Fortes N Kapadia T Lane

P Courtier C Lai K Chang M Ghil D Pregibon C Glymour J Kivinen

C Brodley D McNeill C Matheus P Chan

K Ide P Prado D Peel

modication of user data repository

objects can register as listeners

can use existing Java statistical packages

CERN Colt Scientic Library Visual Numeric!s JMSL for Java

Users are writing

Temporal Egocentric Communication Networks

Integrated into an application

Freeman!s Southern Women

You might also like