Context Discovery in Information Networks

The growing popularity of social networks as well as the increasing amount of heterogeneous data, particularly in the area of the life sciences, has caused the analysis of large information networks to attract ballooning attention in recent years. The flexible structure of these networks enables heterogeneous information from all kinds of data sources to be represented in a uniform manner. This uniform representation allows higher-level information i.e. groups of objects with common properties or hierarchical structures for example, to be extracted from the integrated data, thus enabling information networks to be analyzed at a more abstract level. However, most existing analysis methods only take the most basic, structural properties of the networks into account and neglect this valuable information. The aim of this project is therefore to advance the development of methods to use higher-level information in order to make discoveries at a more abstract level. These methods will, among others, enable previously unknown relationships to be discovered among objects, indicating properties of objects that have been hitherto unperceived. In addition, they will facilitate the detection of missing concepts, which, in turn, will reveal previously undiscovered groups of objects that share common properties. Thanks to the uniform representation of heterogeneous data and the extraction of more valuable and previously unknown information, the methods to be researched in the context of this project will further the comprehension of known objects and relations as well as the discovery of previously unknown objects and relations and subsequently promote the development of new ideas.



  • T. Kötter, S. Günnemann, C. Faloutsos, and M. Berthold, "Fault-tolerant Concept Detection in Information Networks," in PAKDD, 2014. [BIBTEX]
      author = {Tobias K{\"o}tter and Stephan G{\"u}nnemann and Christos Faloutsos and Michael Berthold},
      title = {Fault-tolerant Concept Detection in Information Networks},
      booktitle = {PAKDD},
      year = {2014},