Thomas Gebhart

Ph.D. Candidate at the University of Minnesota.

I am a computer scientist, applied mathematician, and a Ph.D. candidate at the University of Minnesota, Department of Computer Science. My research is focused on the application of ideas from topology and category theory to a range of multi-disciplinary problems. I am interested in the general class of problems regarding the characterization of computations that exist on or are parameterized by networks. Of particular interest are neural networks, knowledge networks, and other complex systems. My hope is that by applying mathematical tools with the proper level of abstraction, we can better probe representations embedded in complex networks, allowing for more structured data science, explainable AI, and further insight into the structure of knowledge.

My CV is available here.

Dec 12, 2022 Graph Convolutional Networks from the Perspective of Sheaves and the Neural Tangent Kernel, accepted as part of the TAGML Workshop at ICML 2022, was published in PMLR.
Jan 26, 2021 A Unified Paths Perspective for Pruning at Initialization, work with Udit Saxena, is now on the arXiv.
Sep 28, 2020 Sheaf Neural Networks, work with Jakob Hansen, was accepted as a Spotlight presentation in the NeuRIPS 2020 Workshop on TDA-in-ML.
Sep 28, 2020 Our work on the topological structure of scientific knowledge networks is on the arXiv.
Aug 12, 2020 I gave a talk on PCA, Dimensionality, and NSD at the NSD mini conference. Slides here.