Machine Learning Studies of Solar Active Regions

Active regions (ARs) on solar surface are areas of concentrated magnetic field which host explosive events such as flares. Active regions emerge from the solar interior in the form of bipolar flux tubes, grow as they rotate across solar disk and ultimately submerge on the surface. From Helioseismic and Magnetic Imager, on-board Solar Dynamics Observatory, high resolution, high cadence vector magnetograms are available for study of magnetic field in ARs. I shall discuss our work on application machine learning for understanding statistical precursors associated with M- and X-class flares within AR magnetic fields. In particular, we show that ARs exist in flaring states, characterized by total magnetic field, for days before and after a flare. We also detect systematic increase and decrease of free energy, in form of current along magnetic field lines, during flares. Study of newly emerging ARs show that, these ARs experience increasing pressure from coronal magnetic field leading to first flares. I shall conclude by discussing application of deep convolutional neural networks for understanding flares and active region emergence.