Machine Learning

»Do more with data!« Machine learning, data mining and knowledge discovery form the foundation of our motto. To make sense of big data, to detect and uncover, and to turn the rear-view mirror of historical data into predictive and prescriptive insights, powerful algorithms are needed. Where one-size-fits-all solutions can no longer cope with the challenges of big data, we develop smarter and faster machine learning approaches for complex data and complex questions. At the same time, we address issues of privacy, understandability and accountability in data mining.

Technologies and methods

Machine learning for smart systems

Machine learning algorithms fuel future smart systems, make them more adaptable to new situations and more intuitive to use. Whether in factories or smartphones, business processes or the web, machine learning needs to be deeply engrained into complex systems, work with heterogeneous and quickly changing data situations, and deliver results that can be processed seamlessly. 

Examples

  • Detecting patterns, with applications in fraud detection or decision support
  • Detecting complex events in times series, with applications in sensor data and factories of the future
  • Mining in big data architectures including stream mining

 

Mining complex data

In modeling real-world applications, often various dependencies need to be taken into account: friends linking to friends in social networks, the flows of goods in logistic networks along streets, tracks or flight lanes, or biological processes in the human body. These are just a few examples of data that can be represented adequately only in the form of graphs. Because of their complexity compared to tabular data, mining graphs and other highly structured objects is significantly more challenging than other forms of data mining. We investigate efficient methods to adequately extract knowledge out of graphs and network data.