Pellizzoni, Paolo (2026): Understanding and improving the expressivity and scalability of geometric deep learning. Dissertation, LMU München: Faculty of Mathematics, Computer Science and Statistics