Tulip -- where Social Network Analysis meets with Visual Analytics
Instructors: Bruno Pinaud (Université de Bordeaux, CNRS UMR 5800 LaBRI, France) & Guy Melançon(Université de Bordeaux, CNRS UMR 5800 LaBRI, France)
Tulip is part of the worldwide leading network visualization software (visit http://tulip.labri.fr). More than just a tool, Tulip offers a full framework that can be customized and extended to the needs of end-users. Tulip comes equipped with a default and basic GUI interface, but its design allows to embed the framework in a domain centered and fully customized GUI. Its plug-in architecture allows one to add any domain specific routine (metric computation, clustering algorithm, ...). Tulip also includes python scripting to easily experiment, and even animate networks.
The tutorial will go over Tulip's functionalities, describing its powerful and flexible data model allowing users to realize almost any manipulation they can imagine. This Tulip sightseeing tour will go over functionalities that are already implemented within the framework, most of them being available through menus of the default GUI:
- graph metrics such as node and edge centralities, k-core, clustering coefficient, etc.
- various layouts including most popular force-directed algorithms, planar graph algorithm, tree layout, etc.
- a set of clustering algorithm such as Newman and Girvan algorithm, etc.
Selected use cases will show the use of graph hierarchies, and multiple coordinated views (classical node-link views, together with scatter-plot matrices, parallel coordinates or self-organizing maps, for instance).
Examples dealing with specific datasets will show how the combination of various already implemented algorithms may be used to drive network analysis and obtain convincing visualizations. Participants will be encouraged to come with their own dataset and questions/tasks.
Depending on demands from the audience, the tutorial could also cover python scripting. Writing simple code scripts is the obvious answer to questions that cannot be handled through the basic GUI and available menus. Taking advantage of data specific attributes is a typical situation that can be handled through python scripting. Using python scripting, participants will be shown how computations may be performed on-the-fly to get immediate visual feedback, or animate networks.
We plan to run the whole workshop as an interactive session, where users should reproduce examples and results using their Tulip install on their own laptop computer (Windows, MacOs or Linux). All material (example data and code) will be made available prior to the workshop.
The tutorial requires attendees to use a laptop on which Tulip will be installed. Configuration issues should be checked prior to registration, as Tulip graphics capabilities expect to find proper graphics hardware (OpenGL required).