Intelligent algorithms for analysing complex data

This group is jointly sponsored by the Department of Molecular Biology and by the Department of Empirical Inference at the Max Planck Institute for Biological Cybernetics.

Karsten's research area is the development of intelligent algorithms for analyzing complex data in biology. It is located at the intersection of machine learning, data mining and bioinformatics, and contributes to these three fields. His research reaches deep into statistics, algorithmics and scientific computing.

In machine learning, Karsten has been working on kernels for structured data, in particular graph kernels, and on a large family of algorithms based on kernel means in feature space, reaching from feature selection, via clustering, dataset shift correction and dimensionality reduction to network inference.

In data mining, his research has dealt with many different aspects of graph mining: efficient graph comparison, feature selection on graphs, graph pattern mining, mining dynamic networks and sampling from graphs.

For bioinformatics, Karsten has presented methods for comparing protein structures and biological networks, and statistical tests for analysing microarray data. A current focus is to develop algorithmic and statistical machinery for genome-wide association studies.

See also his personal page.

Personnel

Dr. Karsten Borgwardt karsten.borgwardt@tue.mpg.de

Group leader

Christoph Lippert

Ph.D. student

Nino Shervashidze

Ph.D. student

Key publications

• Christoph Lippert, Oliver Stegle, Zoubin Ghahramani, Karsten Borgwardt: A kernel method for unsupervised structured network inference, Accepted at AISTATS 2009 (full oral presentation).

• Nino Shervashidze, SVN Vishwanathan, Tobias Petri, Kurt Mehlhorn, Karsten Borgwardt: Efficient Graphlet Kernels for Large Graph Comparison, Accepted at AISTATS 2009.

• Oliver Stegle, Katherine Denby, David L. Wild, Zoubin Ghahramani, Karsten Borgwardt:
  A robust Bayesian two-sample test for detecting intervals of differential gene expression in microarray time series, Accepted at RECOMB 2009. (PDF)