Lade Daten...

Important: Registration

In order to participate in the seminar you have to register for the course in the studiDB. The registration here in the iLearn course is mandatory as well (for the exchange of material and communication), but if the capacity exceeded, only the registration in the studiDB will count! The registration is done on a first come - first served basis.

Next Meeting

The next meeting will take place on Monday, November 23rd 14:00 in this Zoom-room. By November 16th you need to have formed groups and given your top 3 topic preferences.

Kick-Off Slides

… can be found here

Brief Introduction to Heterogeneous Information Networks

This Survey provides broader overview over HINs. We really recommend that you at least skim over it before starting to read too much else.

Heterogeneous Information Networks are a large-scale, semi-structured, very flexible data model that make a recent and hot topic in data science. Formally, an HIN is a graph where all nodes and edges are assigned types. For example, a node type could be ‘person’, ‘web-page’ or any other object type, and an edge type models a relationship between nodes, such as ‘has viewed’ for connecting persons and web-pages.

Another commonly used example is a publication graph as can be seen in this figure:

HIN example figure

Such an HIN could be mined in order to find patterns in the publication habits of authors, rank authors and papers, or recommend papers for citation and authors for future collaboration.

Other application domains of HINs include medical science, material science, marine science and archeology.

HINs allow for semantically rich modelling. This semantic richness can provide deeper insights and knowledge of the underlying data if leveraged correctly. However, with said semantic richness also comes increased complexity of all models and algorithms.

Consequently, HINs make an excellent field of study in data science, since they provide both great opportunities and challenges.