January 30, 2020, 14:00 - 14:30 @ Y25 (Room: Y25-H-38)
Towards Geographically-Aware Machine Learning
Konstantin Klemmer, PhD student at University of Warwick, UK
Summary: Machine learning methods have shown great promise for modelling complex, high-dimensional data environments. However, they still struggle with inherently non-iid data such as geographical data. On the other hand, the academic fields of geographic information science and spatial statistics have long known this issue and developed approaches to identify and embed spatial dependencies. This opens up the opportunity to combine approaches from both areas to enable geographically-aware machine learning with high-dimensional, non-linear data. This talk will highlight why these methods are needed, looking at real-world examples. Further, we will explore some useful spatial metrics and how they can be applied in generative and predictive machine learning models.
Bio: Konstantin Klemmer is a third year PhD student at the University of Warwick. His research focuses on methods in spatial machine learning and how it can be applied to decision- and policy-making in urban environments. Konstantin has spent time at the Alan Turing Institute, the UKs national institute for Data Science and AI. He is currently a visiting student at the Machine Learning for Good (ML4G) laboratory at New York University. Konstantin holds a Bachelor’s degree in Economics from the University of Freiburg (Germany) and a Master’s in Transportation from Imperial College London and UCL.