Classification of celestial objects is a long-standing problem. The sources of celestial objects can sometimes be at unimaginable distances, it becomes difficult for researchers to differentiate between objects such as galaxies, stars, quasars, or supernovae.
A team of two researchers jointly took up the challenge to solve the classical problem by creating SHEEP – a machine learning algorithm that determines the nature of astronomical sources.
Meanwhile, the problem to classify celestial objects is very challenging in terms of complexity of universe and numbers, and AI is a very promising tool for this type of task.
The work is born as a side project of the MSc thesis of the first author of the article. The work combines the lessons learned at that time into a unique project.
Function-wise, SHEEP is a supervised machine learning pipeline that approximates photometric redshifts and uses the information when succeeding classifying the sources as a quasar, galaxy, or star. The photometric information is easiest to receive and thus is very important to provide a first study about the nature of observed sources.
Importantly, in the pipeline, a novel step is that prior to performing classification, SHEEP first studies photometric redshifts. This is then placed into the data set as an additional feature for classification model training.
Incidentally, the team discovered that inclusion of redshift and coordinates of the objects allowed AI to comprehend them within a 3D map of the universe, and the team used this together with color information for better examination of source properties. For instance, AI found that there is a higher chance of finding starts closer to plane of the Milky Way at the galactic poles.
