AIM AND SCOPE

Big Data science has the ability to identify unexpected relations in results, because the mechanisms and phenomena involved, as well as their combined workings, may be different than foreseen by experts. Consequently, two extreme approaches utilising machine learning emerge: data-driven discovery and data-driven modelling. The first one completely avoids models based on previously established insights into the systems at hand, and instead describes the processes entirely based on acquired data. The second approach employs machine learning algorithms in order to identify new relations in data, and only use these new relations to re-assess, reformulate and enhance previously established understanding. Since the second approach has the potential to give a better understanding of reality, and is based on the understanding already present in a research domain, DRIVEN employs this second approach to

• harvest new insights in specific research fields, and
• equip the domain scientists with a solid data science background,

enabling them to generate unprecedented insights in their areas of interest.

GLOBAL OBJECTIVES

DRIVEN aims to empower researchers to

• gather an overview and understanding of the different classes of data-driven and machine learning approaches,
• become familiar with the most promising ones for each research field, and
• develop one or more approaches in the most useful and computationally efficient manner for each specific field.

Several research fields are of interest in DRIVEN and hence the DTU does not only focus on the development of new/improved data-driven discovery approaches themselves, but also on the actual discoveries that will be synergistically facilitated in each research area during the course of the research and training activities.