Virtual Seminar Series on Gaussian Processes, Spatiotemporal Modeling, and Decision-making Systems

2023-03-20

In this talk, I will present opportunities to use geospatial and temporal structures – for example those common to remotely sensed data – to enhance the effectiveness of machine learning systems, as well as several key challenges to making this happen. I’ll start by introducing techniques in machine learning with satellite imagery, focusing on MOSAIKS, a simple system designed to make machine learning with satellite imagery accessible to a wide variety of researchers with differing computational resources. After seeing specific examples with satellite imagery and machine learning, I will generalize our focus to spatio-temporal structures in general machine learning problems and present a general algorithm as a starting point and simple baseline for more complex spatio-temporal modeling. Lastly, I will highlight the importance of multifaceted evaluation of geospatial ML performance and provide a short overview of current evaluation techniques. Throughout, I will outline what I see as key challenges and concrete opportunities for enhancing progress in geospatial machine learning.

Esther Rolf is a postdoctoral fellow with the Harvard Data Science Initiative and the Center for Research on Computation and Society, and incoming assistant professor of Computer Science at the University of Colorado at Boulder (starting in Fall 2024). Esther’s current research interests span geospatial machine learning and spatial statistics (for example, with remotely sensed data), responsible and fair algorithm design and use, and the influence of data acquisition on the efficacy and applicability of machine learning systems. Esther earned her Ph.D. in Computer Science from UC Berkeley in May 2022.