This research group develops theory and algorithms for geometric learning, focusing on artificial intelligence methods that exploit the geometric structure of complex data such as shapes or graphs. By leveraging intrinsic structure both in the data and in the learning methods themselves, we aim to improve robustness, interpretability, and generalization in modern AI systems. Our work builds on and advances concepts from geometric deep learning, non-Euclidean statistics, and applied geometry, with the goal of bridging advances in mathematics with practical machine learning methods.
Applications include shape analysis in medicine, archaeology, and biology, as well as emerging problems in data science and artificial intelligence. Current research topics range from modeling disease-related shape variation and morphological classification to the analysis of single-cell sequencing data and geometric approaches to AI model adaptation and model merging.
| Software | ||
 | The algorithms developed within the group are released as parts of the open source Morphomatics library. It contains Python-based implementations of the intrinsic manifold-based methods and is freely available on GitHub. |