Project Details


Nove Nordisk Fonden - Data Science Collaborative Research Programme 2020.
Representations of data are the cornerstone of any data analysis. Current machine learning techniques build representations from data that are optimized for predictive tasks, but often these abstract representations are used for other purposes. For example, visualizations are used for phrasing scientific hypotheses about how the data was generated. As current machine learning techniques largely operate in a black-box fashion this process can easily result in misrepresentations of data leading to the phrasing of false or misleading hypotheses. In the Center for Basic Machine Learning Research in Life Science we gather some of the leading machine learning researchers in Denmark to conduct the methodological research needed to move this situation forward

The center is structured around four interlinked themes. The first asks the elementary questions of what constitutes a meaningful representation and develops computational and mathematical tools needed to realize the answer. The second uses geometric constructions to incorporate existing knowledge into the representations and ensure that the result is understandable by humans. In life science, data often appears in the form of trees, graphs, and sequences and the third theme explores how to meaningfully represent such data. Finally, the fourth theme acknowledges that real data is noisy and investigates how associated uncertainty is best encoded. Each theme advances different aspects of representation learning for life science, but are constructed to jointly also advance each other.

The center contributes to both teaching and research. It will foster a new generation of excellent methodology-oriented data scientists with enough knowledge in life science to advance practice. We will also organize international summer schools and workshops in collaboration with the elite academic European machine learning network ELLIS, and will offer at least two new MSc courses within the topics of the center
Effective start/end date01/01/202131/12/2025

Collaborative partners