Theme: Geometrization of AI (TDAI Speaker Series)
Title: From Geographic Space to Learned Geometric Space
Speaker: Yao-Yi Chiang, University of Minnesota
Date & Time: Wednesday, March 11, 2026 · 3:00–4:00 PM (ET)
Location: Pomerene Hall Room 350 (Project Zone)
Food: Light refreshments provided
Host: Dena Asta, Subhadeep Paul (TDAI Geometrization of AI Theme Leads)
Event Overview
Many machine learning methods working with spatial data assume that similarity is determined by predefined spaces—such as geographic distance or lexical similarity between place names. In real-world settings, however, these assumptions often fail.
In this seminar, Dr. Yao-Yi Chiang will introduce a representation learning framework that replaces predefined spatial and lexical similarity with learned embedding spaces grounded in geographic and environmental context. He will present an autocorrelation-aware approach for constructing spatiotemporal embedding spaces that enable fine-scale spatial prediction from sparse observations, as well as a geospatially grounded language modeling framework that aligns structured geographic data and natural language within a unified embedding space.
Through this geometric perspective, the talk highlights how learning representation spaces directly from multimodal geographic data can improve prediction accuracy, interpretability, and generalization—advancing Spatial AI systems that are grounded, robust, and capable of reasoning about complex real-world environments.
About the Speaker
Yao-Yi Chiang is an Associate Professor in the Department of Computer Science & Engineering at the University of Minnesota. His research focuses on spatial artificial intelligence, representation learning, and multimodal geographic data integration.
His work develops machine learning systems that learn embedding spaces grounded in environmental and spatial context, integrating heterogeneous data sources such as geographic databases, remote sensing imagery, and natural language. Through a geometric lens, his research advances interpretable and robust Spatial AI systems that better model and reason about the physical world.