PhD Student for Data Science
at Bielefeld University
About Me
I am a PhD student in computer science working at the Semantic Computing Group at Bielefeld University, Germany. I am working towards improving Clinical Decision Support Systems within the KINBIOTICS project funded by the German Federal Ministry of Health. As part of my individual research, I am investigating applications of Federated Learning, particularly within the healthcare domain. Moreover, I am curious to find determinants for the success and failure of Federated Learning in certain applications. In a broader sense, I seek to utilize methods from data science and machine learning to solve real-world problems, mostly related to the healthcare domain.
Originally from a small town of North Rhine-Westphalia, Germany, I received my B.Sc. in Business Information Systems at FHDW Paderborn, Germany and my Master's degree (M.Sc. Business Information Systems, focusing on Data Science) at Paderborn University. During that time, I worked at Bertelsmann SE & Co. KGaA for four years and joined the Social Computing Group of Paderborn University as Student Assistant.
My research interests include:
We share some research interests? Feel free to contact me to exchange some ideas!
We introduce GATE-SR, a social recommender system that uses graph neural networks to model deeper social influences and improve recommendation accuracy, especially in cold-start scenarios.
We overcome the issue of limited availability of data many hopsitals face during the generation of high-quality counterfactual explanationsusing federated learning. Moreover, we showcase the integration into a clinical decision support system for sepsis therapy.
In this paper, we explore local data sampling strategies to improve federated learning performance under data imbalance. We benchmark existing methods and propose a novel sampling approach, showing significant gains in performance and convergence, and training time.
This paper introduces a client selection method for federated learning that addresses data imbalance by aligning client label distributions with one of two target distribution. The proposed method improves performance, with the optimal alignment strategy depending on the type of imbalance.
I presented our paper titled "Leveraging Local Data Sampling Strategies to Improve Federated Learning" at the International Journal of Data Science and Analytics Journal Track
Courses:
Furthermore I actively participate in maintaining academic operations in the following capacities:
