Cloud-Based Machine Learning Service for Astronomical Sub-Object Classification: Case Study On the First Byurakan Survey Spectra

Abstract

The classification of astronomical objects in the Digitized First Byurakan Survey (DFBS), comprising low-dispersion spectra for approximately twenty million objects, presents challenges regarding performance and computational resources. However, considering the distinct spectral characteristics within subgroups, sub-object classification becomes crucial for a more detailed understanding of the dataset. The article addresses these challenges by proposing a comprehensive cloud-based service for classifying objects into spectral classes and subtypes, with a focus on carbon stars, white dwarfs / subdwarfs, and Markarian (UV-excess) galaxies, which are the primary objects in DFBS. By leveraging the power of cloud computing, it effectively handles the computational requirements associated with analyzing the extensive DFBS dataset. The service employs advanced machine learning algorithms trained on labeled data to classify objects into their respective spectral types and subtypes. The service can be accessed and utilized through a user-friendly interface, making it accessible to a wide range of users in the astronomical community.

Keywords

• telescopes
• surveys
• catalogs
• byurakan plate archive
• databases
• big data
• data mining
• machine learning
• virtual observatories