Jin B, Peruzzi M, Dunson DB (2022) Bag of DAGs: Flexible Nonstationary Modeling of Spatiotemporal Dependence. arxiv.org/abs/2112.11870
Winds impact air quality, especially during forest fires, and lead to nonstationarities in the spatial covariance which characterizes air quality. In these settings, scalable GPs based on sparse DAGs may fail to capture nonstationarity. We introduce methods to retain scalability to large scale data while accounting for the effects of wind in modeling air quality. Unlike others, our methods can take advantage but do not necessarily depend on the availability of wind data.
- Best Student/Postdoc Contributed Paper Award, 2021 ISBA World Meeting
- Student Paper Award, 2022 ASA ENVR section
