Machine Learning Methods for Coronal Hole Detection

Machine learning is a method of data analysis that has grown significantly in popularity over the past decades specifically in regards to dealing with large datasets. It is useful for variety of problems and can limit human interaction with the problem's solution. Coronal hole detection is an image segmentation problem. This means we are looking at how to partition an image into segments to locate objects and boundaries. This is useful for coronal hole detection as this is a sort of fuzzy problem with various definitions that are person dependent.

There are two main methods of machine learning: supervised and unsupervised. We have detection algorithms using both types of machine learning. For this problem, unsupervised learning provides more promising results as it directly works to counteract the issue of the fuzzy definition of a coronal hole. Additionally, unsupervised learning allows us to detect other features of interest on the solar disk such as active regions.