Modeling The Spatiotemporal Dynamics Of Active Regions On The Sun Using Deep Neural Networks

Solar active regions are areas on the Sun\u27s surface that have especially strong magnetic fields. Several phenomena that can have significant negative effects on technology and subsequently on human life, such as solar flares and coronal mass ejections (CMEs), are often associated with active regions.Since the physical phenomena underlying the evolution of active regions are still poorly understood, the accurate prediction of solar flares and coronal mass ejections remains an open problem. Extracting insights from the available datasets of solar activity that can lead to a better understanding of solar active regions has been an important research goal at the intersection of artificial intelligence and solar physics. With the advancement in artificial intelligence, some machine learning models have been applied to predict solar flares from a 6 hour to 48-hour window. Support Vector Machine (SVM) [6, 42], K-Nearest-Neighbor (KNN) [29], Extremely Randomized Trees (ERT) [37], and deep neural network [36] are some of the machine learning models that have been used in predicting solar flare but results are not good. This can be attributed to the fact that the models are trained using a selection of Active regions parameters and an imbalance data (few positive flare examples). As a result, there is a need to understand space weather and the basis by which these events occur. In this study, we applied a deep learning architecture originally designed for video prediction to predict the changes happening on the Sun in continuous time by using time series Helioseismic and Magnetic Imager data captured by Solar Dynamics Observatory (SDO) and compared it against a no-change baseline and a regression baseline. These three approaches were compared against one another based on their mean squared error (MSE) and structural similarity index measure (SSIM) and it was found out that the regression model outperforms the others in MSE whilst the deep learning model outperforms the rest in SSIM. From this, we seek to continue our work by adapting deep learning models used in solving image-to-image translation problems to produce high-quality synthetic data to solve the class imbalance data problem and incorporate other time-series data of the Sun to improve upon the predictions of the spatiotemporal changes of active regions on the Sun