Where and when

speaker: Sabrina Guastavino
affiliation: DIMA | Department of Mathematics, Università di Genova
date: May 15^th, 2023
venue: on-line seminar (Google Meet) organized by "Gruppo outreach and media public relations" of SWICo

Highlights on artificial intelligence in space weather for solar flare forecasting

abstract: In recent decades, artificial intelligence has been obtaining an increasing interest in the field of space weather, such as in solar flare forecasting, due to both the availability of a plethora of data acquired by many telescopes and the numerous recent developments in machine and deep learning. Flare forecasting may rely on features extracted from magnetogram images of Active Regions (ARs), e.g. the ones recorded by the Helioseismic and Magnetic Imager (HMI) on-board the Solar Dynamics Observatory (SDO). Such features can be physical parameters, previously extracted from HMI images as the ones provided by HMI/Space-Weather HMI-Active Region Patch (SHARP) data files, or they can be artificial features, extracted directly by some suitable deep learning methods as Convolutional Neural Networks applied directly to HMI images. In this talk we focus on some key points for the effectiveness of both feature-based machine or image-based-deep learning techniques in forecasting problems as (1) the splitting-strategy for the separation of the training, validation and test sets to effectively validate the robustness of the machine or deep learning technique, (2) the choice of the loss function for the training process, and (3) the choice of the evaluation metrics to assess the goodness of predictions.