Deep Learning Structure is a research group dedicated to uncovering the theoretical foundations of modern neural networks. We investigate how geometry, architecture, and representation interact to shape generalization, robustness, and learning dynamics in high-dimensional models, including transformers and large language models. Our work combines mathematical insight with empirical rigor to build a structural understanding of deep learning behavior.

Junior Research Group Leader

Dr. Linara Adilova

TU Dortmund University
Room 315
Otto-Hahn-Straße 14
44227 Dortmund
Germany

Scroll To Top