Hans Riess

Abstract

Spatiotemporal data, which tracks changes across space and time, can be notoriously complex due to potential irregularity in both the spatial and temporal domains. This talk presents a fresh approach by merging path signatures and graph neural networks (GNNs). Path signatures transform multivariate time series into robust and interpretable algebraic features, capturing temporal patterns. GNNs, on the other hand, shine at modeling spatial dependencies, such as those in sensor networks, by linking nodes in a graph and applying local graph convolution or attention operators. Together, they form a combo with great potential: path signatures handle the "when," while GNNs tackle the "where." We’ll showcase this in action with a real-world case study: analyzing slow slip events (SSEs) with GPS sensor network data. Finally, we'll discuss alternative approaches and present preliminary numerical experiments probing the representational ability of our approach---SignatureGNN---for synthetic spatiotemporal tasks.

Our speaker

Dr. Hans Riess is an applied mathematician and engineer who currently serves as a Research Scientist in the School of Electrical and Computer Engineering at the Georgia Institute of Technology. Hans has established an independent research program which aims to leverage category theory and algebraic topology to drive innovations in multi-agent systems, machine learning, and optimization. At Georgia Tech, Hans also works closely with the Control, Optimization, and Robotics Engineering (CORE) Lab, directed by Dr. Matthew Hale, which engages in research activities ranging from hands-on robotics to developing sophisticated mathematical techniques. In 2016, Hans earned a B.S. in Pure Mathematics from Duke University, where he completed the entire Ph.D.-level topology sequence. He later completed his Ph.D. in Electrical and Systems Engineering (ESE) at the University of Pennsylvania in 2022, working under the supervision of Professor Robert Ghrist to make advances in cellular sheaf theory. Following his doctorate, in 2022, he joined the Autonomous Systems Lab at Duke University, directed by Dr. Michael M. Zavlanos, and, in 2025, he joined the research faculty at the Georgia Institute of Technology.

To become a member of the Rough Path Interest Group, register here for free.