Thomas Gebhart

Researcher at the University of Minnesota.

I am a computer scientist, applied mathematician, and postdoc at the University of Minnesota. My research is focused on the application of ideas from machine learning, network science, and algebraic topology, to a range of multi-disciplinary problems. My work is broadly motivated by a desire to understand how knowledge is acquired, structured, and employed by machines, humans, or their larger collective structures. I am particularly interested in systems where knowledge is parameterized or transformed by networks, as is the case with deep learning, knowledge graphs, and numerous problems within the computational social sciences. Modeling these network-structured knowledge domains with the proper level of mathematical abstraction allows for better control and more efficient usage of the domain’s accumulated knowledge, resulting in, for example, more faithful representations of embedded information, more explainable AI, and better insight into the drivers of scientific and technological change.

I received a Ph.D. in Computer Science from the University of Minnesota in 2023. I also hold B.S. degrees in Mathematics and Economics from the University of Minnesota.

My CV is available here.

May 10, 2023 I will be giving a talk about cellular sheaf theory in AI at the Categories for AI seminar on May 29th.
Apr 15, 2023 I will be giving a talk about scientific disruption and citation centrality at Sunbelt 2023 this year (June 27-30).
Apr 14, 2023 I will be presenting a poster at ICSSI this year (June 26-28) regarding spectral modeling of scientific novelty and disruption.
Apr 12, 2023 I will be presenting our Knowledge Sheaves paper at AISTATS in Valencia April 24-29.
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.