Jonas Fischer (Research Leader)

I am heading the Explainable Machine Learning group in the Department for Computer Vision and Machine Learning of the Max Planck Institute for Informatics. My research is driven by the questions of what information is encoded in complex Machine Learning models, how they use this information to arrive at predictions, and how we can build inherently interpretable models from scratch. Connecting ideas from Data Mining and Neural Network pruning with XAI, the goal is to generate global and human-interpretable explanations of the encoded information that leads to a better understanding and design of ML models with respect to robustness as well as human alignment of the decision-making process.

Before my time at MPI, I was a postdoctoral fellow at the Department of Biostatistics at Harvard University, which largely influenced my research towards Explainability of complex models. There, I worked on statistical as well as machine learning models to understand gene regulatory systems, in particular in cancer, for which interpretability and transparency is absolutely essential to the domain experts.