Speaker
Description
We present an ongoing effort dedicated to the development of reproducible and operationally viable artificial intelligence models for solar flare forecasting. Our approach leverages the extensive archive of multi-wavelength solar images captured by the Solar Dynamics Observatory (SDO) using the Atmospheric Imaging Assembly (AIA). Specifically, we employ a self-supervised learning strategy, utilizing a masked autoencoder task to pre-train a transformer-based architecture. This pre-training phase enables the model to learn robust and meaningful representations of diverse solar activity patterns from unlabeled data. Subsequently, we transfer the knowledge acquired during pre-training to a downstream task focused on segmenting sunspots in images of the photosphere obtained from the SDO's Helioseismic and Magnetic Imager (HMI). This work aims to bridge the gap between fundamental research and practical deployment by systematically addressing key challenges related to the generalization of the models to unseen solar conditions , ensuring their scalability to handle large volumes of solar data and optimizing their performance for deployment in resource-limited operational environments. The ultimate goal of this research is to provide the way for future real-time AI applications in space weather operations, which will give important advance notice for reducing the effects of solar flares on critical infrastructure and human activities.
| Do you plan to attend in-person or online? | In-person |
|---|
