Geometric Deep Learning
SE(3) equivariant graph neural networksMy passion for Geometric Deep Learning can be unmistakenly traced back to my physics background. Doing a PhD in high energy physics was confronting me with the standard model of particle physics on a daily basis. Under the hood, the standard model is a gauge quantum field theory which contains the internal symmetries of the unitary product group $SU(3) \times SU(2) \times U(1)$. The theory is commonly viewed as describing the fundamental set of particles – leptons, quarks, gauge bosons and the Higgs boson – as well as their interactions.
Unsurprisingly, I have been deeply attracted to the field of Geometric Deep Learning which has groups, gauges, and graphs at its core. I have contributed to the fields of graph neural networks, equivariant architectures, and neural PDE solvers. Furthermore, I have lead efforts to introduce Lie Point Symmetries, and, most recently, Clifford (Geometric) Algebras into the Deep Learning community.
For my group “AI for data-driven simulations”, Geometric Deep Learning techniques will play an important role when working on grids, graphs, meshes, or with complex geometries in general. Recent trends on implicit parametrization, generative modeling, and latent space dynamics are very appealing as well.