Addressing research challenges in computer science specifically needed to make breakthroughs in data-driven research
The “Generic eScience initiative for the Netherlands eScience Center” project dealt with the data-explosion challenge, especially the increasing amounts of data and the fact that data becomes more complex and heterogeneous. In Astronomy, for the software for the LOFAR telescope we studied the computer systems and software that have to process large amounts of complex data as fast as possible. We developed a tool to get the most out of LOFAR’s supercomputer equipped with GPUs, technology originally developed for games.
In the forensic domain a similar problem is occurring. Since smartphones and computers in general become more ubiquitous, the Netherlands Forensic Institute receives an ever increasing amount of data from such devices acquired from suspects. We developed a software framework with which they can scale up their data analyses by using supercomputers equipped with GPUs. Because the software is generic, it can be applied in many other fields dealing with increasing amounts of data.
In addition, this project studied a more fundamental problem of how to extract useful information from data, in particular complex and linked data. This resulted in a generic framework to run machine learning experiments and a system for performing machine learning on linked data – a research field that has only started recently.