Machine Learning

CITEC with its interdisciplinary research in robotics, sports sciences, natural language acquisition and understanding, human memory and learning, neuronal control strategies, or the investigation of animal behavior, observes a rapidly increasing amount of electronic data in all areas of research which can often no longer directly be inspected or handled by humans due to its sheer size and complexity. Hence automatic techniques which help humans to extract relevant information from these data are required.

The research group Machine Learning (ML) was established at Bielefeld University on April 1st, 2010, as part of the CITEC center of excellence and the Faculty of Technology. Originally called the Theoretical Computer Science group, it was renamed to Machine Learning in 2016. Its main focus lies in the development and analysis of cognitively inspired machine learning techniques to automatically analyze digital data sets and to infer useful information from the data. As such, the research of the Machine Learning group centers around a key enabling technology for CITEC and its theoretical foundation.

Typical data analysis tasks include data clustering, data visualization, the inference of data models for classification, regression, or density estimation, relevance learning and feature extraction, etc. Due to improved sensor technologies, dedicated data formats and storage facilities, these classical objectives face severe challenges which make the development of novel data analysis technology necessary: data are becoming extremely heterogeneous and high dimensional, often, multiple modes and additional structural information are available. In addition, extremely large data sets have to be dealt with. These challenges can be faced using technologies such as metric learning, models for non-Euclidean data, hybrid symbolic-subsymbolic systems, and various approximations for very large or streaming data.

Within this line of research, different topics are currently investigated within the group:

  • relevance and matrix learning
  • relational data processing
  • neuro-symbolic integration
  • data visualization
  • dealing with very large or streaming data sets
  • applications in the biomedical domain such as the analysis of mass spectra
  • theoretical foundations e.g. by means of computational learning theory

Group Photo

A picture of the group members.