Fellow-Klasse 2010/11 Prof. Dr. Fred Hamprecht

Arbeitsvorhaben am Marsilius-Kolleg

Networks: quantitative analysis in real-world settings

Some of the most interesting and relevant social, scientific and economic phenomena can most aptly be described in terms of networks. While many of these networks are not new in themselves, they now become observable for the first time due to a data accumulation that is without historical precedent. These new data allow and inspire new questions that in turn call for new automated and quantitative analysis techniques.

The aim of this project is to develop new algorithms that exploit network structure to solve exemplary concrete problems in the neurosciences and socioeconomic geography; and to work with the other Fellows to identify and discuss the possible impact of network analysis in their respective fields.

Details of exemplary problems and desired cooperations are given in the following:

Network Analysis in the Neurosciences * 

My group is working in image analysis and is taking part in a worldwide race for the reconstruction of the "connectome" (wiring diagram) of the human brain. The reconstruction itself exploits ideas from graphical models, and is hence based on networks; but even once a satisfactory reconstruction is obtained, a characterization and analysis of the network of billions of neurons at a higher level of abstraction will be necessary to make advances towards an understanding of neural computation and, ultimately, cognition. 

Data of this kind is currently acquired at the MPI in Heidelberg (W Denk). Recent developments by local microscopists and biologists (J Wittbrodt, E Stelzer, S Ryu) raise hopes to track the evolution of an entire neural system in statu nascendi. Such evolving networks are of great interest to Marsilius fellow G Reinelt.

Empirical social networks *

The partial shift of human communication to computer-based resources (VoIP, messaging services, email, social websites, blogs, . . . ) is now generating traces that are amenable to fully automated analysis. Current work [Leskovec 2008] exploits billions of messages between millions of users to establish patterns of human (digital) communication. Smaller networks attributed with more expressive features are currently established by Marsilius fellow J Glueckler. 

I am also looking forward to discuss with the other Fellows possible implications for their fields. Specifically, if there is interest from colleagues in law and ethics, I would like to organize a joint seminar for students from computer science & physics, law and philosophy to discuss the opportunities and possible abuses of (wireless) sensor and surveillance networks.