Research Profile

The Machine Learning Group at TU Berlin is chaired by Prof. Dr. Klaus-Robert Müller and includes more than 50 doctoral and postdoctoral researchers. We work on a diverse set of areas in methodology and application, including the following:

For most of these topics, we have focus areas coordinated by senior research leads. Furthermore, our group works on additional research areas not explicitly captured by these focus areas, e.g., representational alignment and federated learning. We recommend browsing our team page to gain a more granular impression of our research activities.

Research Focus Areas

Explainable AI

Senior Research Lead: Grégoire Montavon

Work conducted in the context of this focus area aims at advancing the foundations and algorithms of explainable AI (XAI) in the context of deep neural networks. One particular focus is on closing the gap between existing XAI methods and practical desiderata. Examples include using XAI to build more trustworthy and autonomous machine learning models and using XAI to model the behavior of complex real-world systems so that the latter become meaningfully actionable. Future research will treat the questions (1) how to use XAI to assess on which data a deep neural network model can be trusted to perform autonomously or requires human intervention, and (2) how to use XAI in combination with a deep neural network to model complex real-world systems and identify actionable components.

Probabilistic Modeling, Bayesian Inference, Uncertainty Estimation

Senior Research Lead: Shinichi Nakajima

The purpose of this focus area is to develop novel probabilistic models and inference methods for multimodal, heterogeneous, and complex structured data analysis. In particular machine learning tools that can incorporate multiple aspects of data samples observed under different circumstances, in efficient and theoretically grounded ways, are investigated. This includes:

  • Developing novel probabilistic models with efficient inference methods
  • Exploring novel applications of probabilistic models
  • Establishing uncertainty estimation methods for deep probabilistic models

Many-Body Dynamics, Physics-Informed Models, Numerical Methods

Senior Research Lead: Stefan Chmiela

Within the scope of this focus area, we are developing machine learning methods for molecular simulations, with a special emphasis on many-body problems in quantum chemistry. Modeling many-body problems is computationally intensive due to the rapidly growing number of non-local interactions with system size. In quantum chemistry, even the smallest practical problems already involve enough interacting electrons to render analytical solutions impossible. This combinatorial complexity carries over to the simplified atomistic picture adopted by most empirical models, where a lower number of particles interact. To address this challenge, the group develops methods that combine fundamental principles from computational physics with statistical modeling approaches to foster a better understanding of quantum phenomena in complex systems. This data-driven angle allows questions to be asked in new ways and can give rise to new perspectives on established problems.

Biomedical Engineering & Multimodal Data Analysis for Wearable Neurotechnology

Senior Research Lead: Alexander von Lühmann

The Intelligent Biomedical Sensing (IBS) group, led by Dr. Alexander von Lühmann concentrates on Machine Learning and Instruments for Comprehensive Brain-Body Monitoring. The IBS lab develops miniaturized wearable neurotechnology and body-worn sensors for unobtrusive monitoring of the embodied brain in the everyday world. It uses machine learning on the multimodal sensor data, together with environmental context information, to contribute to a paradigm shift in individualized Comprehensive understanding of physical and mental health: Toward intelligent assessment and treatment of physical and mental states and risk factors. The expertise of the group encompasses

Biomedical Electrical Engineering: Development of novel wearable sensing technology for brain and body that is non-invasive/non-hazardous, unobtrusive, multimodal and robust. Current focus in instrumentation development: functional Near Infrared Spectroscopy (fNIRS), diffuse optical tomography (DOT) and Oximetry, Electroencephalography (EEG), Electro -myo-, -oculo-, -cardiograpy (ExG). Multimodal Machine Learning: Exploration of innovative methods for the extraction of biomarkers from complex multivariate bio signals derived from diffuse optics and electrophysiology. Physiological modelling, latent component analysis, and physiological transfer functions considering non-stationary and non-instantaneous relationships, context sensitivity, and automatic data annotation.