Topical Discussion Meeting - Machine Learning Systems for Space Weather Prediction

Thomas Berger (University of Colorado Boulder):Ryan Mcgranahan(NASA Jet Propulsion Laboratory):Enrico Camporeale(Centrum Wiskunde & Informatica, Amsterdam)
Friday 9/11, 15:15-16:30
MTC 01.03

Machine learning is a rapidly advancing field of computer science that applies statistical inference techniques to the task of complex data mining, prediction, and decision making. These techniques have revolutionized our daily lives and are poised to have a similar impact across the sciences as well. In the field of space weather, machine learning techniques may be applicable to some of the most intractable prediction problems such as solar eruption triggering, geomagnetic storm intensity, and solar energetic particle events. Pioneering applications of machine learning techniques have shed light on some of the promises and obstacles facing these “statistical observing systems”.

This discussion will seek to gather current practices, anecdotal experiences, dataset knowledge, and lessons learned from machine learning developments in space weather as well as ideas for moving forward into the realm of “deep learning” using very large neural networks with massive datasets such as the full SDO dataset and outputs from full-physics models of the Sun-Earth system.