Methods for large network analysis

Tutorial by

Vladimir Batagelj
(University of Ljubljana, FMF)

Monday 7 September (Room Libertas, Hotel Golf)

Abstract:

Networks appear in many data analytical tasks as a natural framework. Often we have to deal with networks with several hundred  thousands or millions of vertices/lines. For such networks special, subquadratic methods should be used to produce results in a reasonable time.In the tutorial an overview of methods for large network analysis will be presented:
– connectivity based methods: weak, strong and bi-components; condensation of networks; short-cycle connectivity.
– properties of vertices/lines: closeness, betweenness, hubs and authorities,Hummon-Doreian weights, 3-ring weights, . . . ; cuts and islands; cores and generalized cores.
– two-mode networks: multiplication of networks, two-mode cores, 4-ring weights.
– pattern search (motifs), triadic spectrum.
– clustering in large networks.
– visualization of large networks.
Most of presented methods are available in program Pajek for analysis and visualization of large networks http://pajek.imfm.si.
The tutorial is targeted towards the general research audience.