Workgroup Functional Data Analysis

Due to recent advances in sensor technology and automated data processing, an increasing amount of data that can be continuously measured and therefore considered to be functional -- curves, trajectories, or even higher dimensional surfaces -- has become available in many scientific and commercial applications.

Our group works on methodology and software implementations that process, describe, visualize and model such data. As part of the Munich Center for Machine Learning, our research focuses on the analysis of functional data using generalized additive regression and on both supervised and unsupervised methods for functional data, for example for automated outlier detection or dimension reduction.

We are focused on evaluation, transparency and reproducibility. The development and documentation of suitable, user-friendly open-source software and data standards is as important to us as conducting benchmark experiments to  compare existing algorithms and new developments.

Functional data analysis has played an important role at the Institute of Statistics for a while, initiated by Prof. Sonja Greven's Emmy-Noether-Group "Statistical Methods for Longitudinal Functional Data" (2010-2016).