Prof. Dr. Michael Berthold
ContactPhone: +49 7531 88-2202
Room: Z 813Website
I am interested in everything related to making sense out of (large, heterogeneous) data sources. In particular, I am interested in methods from AI (rule learning, neural networks, fuzzy logic, general machine learning) but always try to keep the models interpretable. Preferred are problems where the user needs to be kept in the loop in order to find truly interesting stuff. I call all of this "Knowledge Discovery Support". More recently, David Krakauer coined the term "Intelligible Data Analysis" (IDA), which captures the essence of what we are after rather well, too. Either way, problems from the life science domain are often the underlying driver since they are enormously good at creating amazing data repositories but have huge problems trying to make sense out of them or even coming up with useful questions to ask. Ultimately, the methods we develop should help to "spark (new) ideas"!
More to the point, we work on topics such as:
- interactive clustering in parallel universes (HTS data)
- active learning (High Content Screening data)
- hierarchical (fuzzy) rule systems
- (molecular) graph mining (screening and synthesis data)
- distributed algorithms in data mining
and a bit of text, image and network analysis, but not in the sense of improving those methods but in order to use them to find relations in large, heterogeneous databases (see also the EU FP7 "Bison" project, which we coordinated from 2008-2011).
Dr. Matthias Rupp
Bioinformatics and Information Mining Group
Dr. Martin Horn