Data science methods
The data science methods group
The group focuses on supporting the research groups “Clinical AI” and “Bioinformatics and biostatistics” with methodological development of machine learning algorithms for data correlated in time and space, data fusion, and data synthesis.
Machine learning methods for data correlated in time and space
We work on adapting ideas from generalized linear mixed models to machine learning algorithms to account for health data correlated in time and space in order to increase the effectiveness of prediction models and unsupervised clustering algorithms.
Algorithmic and theoretical properties of data fusion
Different data modalities may have different optimal machine learning methods for predictive purposes. We work on the algorithmic and theoretical properties of fusing different machine learning algorithms.
Methodological development of data synthesis methods
We develop Bayesian methods to evaluate the ability of data synthesis methods to protect data privacy. The purpose of this is to build a bridge between abstract mathematics and legislation to ensure that the methods live up to the General Data Protection Regulation.