HAO Colloquium - Enrico Camporeale, CU CIRES / NOAA

Machine Learning and the grey-box approach in Space Weather forecasting

Artificial intelligence is changing the world more rapidly than any other technology in history.

Hence, it is becoming important to investigate if and how the process of scientific discovery can benefit from the methodologies developed in the fields of artificial intelligence and machine learning.

In space weather, we have the clear advantage of having access to an enormous amount of data. In fact, we have much more than we can possibly analyze with standard means and single case studies.

Machine learning can successfully be utilized for many purposes: to automatically collect catalogs of events, to discover patterns and nonlinear relationships in high dimensional data, to identify outliers in a dataset, possibly unveiling new physical mechanisms, and to better inform parametrizations that are routinely used within physics-based models.

However, the “black-box” nature of machine learning has so far represented a strong argument for skeptics and opponents and it has effectively posed a barrier to its systematic use in our community.

In this talk, I will give an overview of the multitude of possibilities that machine learning can offer in space weather, particularly emphasizing the “gray-box” approach, namely the combination of data-driven knowledge and physics laws. Such approach is designed to make the best possible use of our collective knowledge of a physical system. A specific example will be discussed about enhancing global MHD simulations with machine learning.

Date and time: 
Wednesday, October 30, 2019 - 2:00pm to 3:00pm
Building: 
CG-1
Room: 
2126