Hot subdwarfs come into focus as AI scans thousands of stars in Gaia data

The research team at Vilnius University used machine learning to detect hot subdwarf stars in Gaia data. They trained a convolutional neural network on 2,500 stars with known classifications and used it to analyze 17,500 unknown stars. The AI learned to identify hot subdwarfs in the Gaia XP spectra, a method that was previously infeasible due to the rarity of the stars. The study found that hot subdwarfs are smaller, less luminous, and hotter than most stars, and they are the result of binary star systems where one star strips away the other's outer envelope.

• The research team used Gaia data to identify 60,000 hot subdwarf candidates.
• They examined one-third of them using XP spectra, a method that was previously infeasible due to the rarity of the stars.
• The AI was able to process the remaining 17,500 unknown stars in seconds, identifying them as hot subdwarfs.

The study highlights the importance of machine learning in detecting rare celestial objects and the value of interdisciplinary collaboration. The research team worked with colleagues from different disciplines to optimize the use of the Gaia XP spectra database and improve model interpretability. The project has led to the discovery of hot subdwarf stars in Lithuania and has provided opportunities for professional growth within the broader science community.

• The research team discovered 17,500 hot subdwarf stars using the AI.
• The AI was able to process the remaining 17,500 unknown stars in seconds.
• The study highlights the importance of machine learning in detecting rare celestial objects and the value of interdisciplinary collaboration.

AI-generated summary of the original article — see the source for the full story.