Newsletter July 2023

Research Unit

The German Research Foundation (DFG) has approved funding for the establishment of eight new research units in the field of artificial intelligence (AI) from 2023 to 2026, with a total funding of around 31.4 million euros. Our research unit, DeSBi, is one of the winners, and started in January 2023.

The DeSBi Research Unit on Fusing Deep Learning and Statistics towards Understanding Structured Biomedical Data is a unique collaboration platform between several institutions, including Humboldt-Universität zu Berlin, TU Berlin, TU Dortmund, University of Potsdam, Hasso-Plattner-Institute, Charité - Universitätsmedizin Berlin, Max Delbrück Center for Molecular Medicine and Fraunhofer Heinrich Hertz Institute.

Our team consists of senior experts and junior scientists with expertise in statistics, machine and deep learning, and their biomedical applications. Our scientific goal is to develop methods for the analysis of large structured data such as images or genetic sequences in the biomedical sciences. This requires the development of innovative methods that combine the strengths of deep learning (DL) and statistics. We aim to improve interpretability, uncertainty quantification and statistical inference for DL, and to improve modeling flexibility of statistical methods for structured data. In particular, we will develop methods that provide statistical inference for structured data by quantification of uncertainty, testing of hypotheses and adjustment for confounders, and that improve explanations of structured data through hybrid statistical and deep learning models, population- and distribution -level explanations, and robust sparse explanations. We create all methods in a feedback loop between method development and biomedical applications to achieve this goal.

We place a significant emphasis on scientific exchange with both national and international partners, which is facilitated also by the seminar series, international summer schools, and workshops we organize, as well as keynote speeches and tutorials by external experts during our annual retreats. Our research unit comprises seven projects, eight principal investigators, and sixteen scientific staff. We have a total budget of approximately 6 million Euros for the next four years.

DeSBi Annual Retreats

Our research unit holds an annual retreat to promote scientific exchange and benefit from the diverse expertise of our project teams. It's a crucial event for all members to discuss work progress, showcase achievements, and engage in informal exchange during social activities. The organization team in coordination with PIs schedules the retreat a year in advance to ensure all members can attend.

This year's annual retreat was held from June 19 to 23 at the Bildungszentrum Elstal, located close to Berlin to facilitate commuting for members with small children. Over the course of five days, we had the chance to connect with each other, delve deeper into the seven DeSBi projects, and attend insightful presentations, tutorials, and keynote speeches. It was a wonderful opportunity to learn and grow together. Our first retreat contained a range of tutorials by internal and external experts. We were privileged to have lead German professors specializing in AI as our guests, including Prof. Dr. Marius Kloft from TU Kaiserslautern, an expert in statistical machine learning and head of another DFG-AI research unit on the application of Deep Learning to sparse chemical data. He offered the DeSBi team an informative tutorial on statistical learning theory, which was greatly appreciated. Ralf Herbrich, Chair of AI and Sustainability at the Hasso-Plattner-Institute and an experienced professional who has worked for Microsoft, Amazon, and Zalando, provided a tutorial on probabilistic machine learning. Moreover, we were honored to have Prof. Dr. Lothar Wieler, the former Director of the Robert-Koch-Institute, as the keynote speaker for this year's retreat. During his presentation, he stressed the importance of Big Data and the use of AI methods in evidence-based policymaking.

We also had informative and engaging tutorials, where we greatly benefited from the diverse expertise of our project team, covering various topics such as Neuroimaging, Genomics, hybrid statistical and deep learning models, and explainable and probabilistic AI. We discussed important current research topics as well as gender and racism in academia to raise awareness within the research unit.

Our inaugural retreat in Elstal also offered many opportunities for informal exchange over meals and during social activities such as sports, outings and group games, which sparked additional exchange and collaboration ideas between the different DeSBi projects. The organizing team expresses gratitude to all participants and guests. The next retreat is scheduled for sometime between May and June 2024, and the date will be announced soon.

Seminar Series

The RU’s joint seminar series promotes scientific exchange and enhances the RU's visibility. We invite both internal and external speakers to present at the seminar, which is held every 3 to 4 weeks during the semester rotating between the RU’s member institutions. To allow students and researchers with diverse backgrounds to attend, we offer a comprehensive introduction to the topic before each seminar.

We are currently planning the seminar series for the winter term 2023/2024 and summer term 2024. These seminars will be widely advertised and accessible for virtual participation via live stream.

Our next seminar is scheduled for October 17th, 2023 at HU Berlin, where Prof. Marvin Wright, computational biostatistics professor at the University of Bremen, Head of the Emmy Noether junior research group: Beyond Prediction - Statistical Inference with Machine Learning, and bridge professor in the DFG-AI research unit Lifespan AI, will present on "From Explainable AI to Generative Modeling with Tree-Based Machine Learning." We are looking forward to seeing you at the seminar!