The Digitized First Byurakan Survey (DFBS) is the largest and the first systematic objective-prism survey of the extragalactic sky. The detection, extraction, and classification of about 40 million spectra of about 20 million astronomical objects available in the survey require distinguishing the pixels containing photons from the source and the noise pixels per object. This project aims at developing a service to classify the spectra of UV-excess galaxies, quasars, compact galaxies, and other objects in the survey. Supervised and unsupervised convolutional neural network deep learning algorithms will be studied and developed
The usage of the large astronomical surveys is critical to study the universe. The scientific discoveries of such astronomical datasets rely on efficient and robust techniques and methodologies to recognize, discover and search astronomical objects. The project outcome are the machine learning and deep learning methods providing powerful tools to address challenges in astronomical data analysis, focusing on DFBS datasets.