Learning distributions over high-dimensional data is a fundamental problem in machine learning. Recent advances in generative AI have identified models that learn complex distributions over images from finite data, making it possible to generate surprisingly realistic images from captions. In this work we apply score-based diffusion models to problems in weather and climate modeling and demonstrate the value of these methods for forecasting and statistical downscaling.
Peter Sadowski is an Associate Professor of Computer Science at the University of Hawaii Manoa. He completed his Ph.D. at University of California Irvine, and his undergraduate studies at Caltech. His research focuses on machine learning and applications to the physical sciences, with funding from NASA, DOE, and a CAREER award from NSF.
See upcoming and previous presentations at CS Seminar Series.