Learning Machines Make Sense of the Stars

Astronomers are aiming to sort and archive all the stars in our galaxy - a task they hope will help them better understand the nature of stars, their distribution, and how galaxies are formed. However, it has been estimated that the Milky Way boasts between 200 and 400 billion stars. So who can help astronomers with this incredible workload? Learning machines, of course!

Machine learning isn't exactly a new concept, and has become a powerful tool for scientists looking to process incredibly massive sets of data. Like in the case of many other big-data sets, information on the thousands upon thousands of stars recorded over decades of sky surveys is far too complex even for the most powerful of supercomputers. So instead, experts make programs that can "learn" how to categorize star images based on patterns and properties.

This is not unlike the internet word and image program LEVAN - which stands for "Learn EVerything about ANything." You can learn more about this stunningly intelligent and somewhat unsettling definitions program here.

Experts at NASA's Jet Propulsion Laboratory (JPL) in Pasadena, Calif. are now attempting to apply similar technology in tackling the ever-growing star image archives of the world, and they reportedly are seeing some success.

"This is an exciting time to be applying advanced algorithms to astronomy," JPL scientist Adam Miller said in a statement. "Machine learning allows us to mine for rare and obscure gems within the deep data sets that astronomers are only now beginning to acquire."

According to a study recently published in the Astrophysical Journal, Miller and his colleagues managed to craft a program that not only follows a complex algorithm to categorize stars, but also learns more about these stars as it goes, making new rules to more precisely archive them.

"It's like video-streaming services not only predicting what you would like to watch in the future, but also your current age, based on your viewing preferences," Miller explained.

He added that after a brief "training session" with a mere 9,000 stars, the program was already able to make associations between star properties and their light curves - an important and time-saving trait.

Now the program and its associated machines have been tapped into a great many more unsorted star image archives, launching what will be an impressive crusade to make more sense of the sky.

For more great nature science stories and general news, please visit our sister site, Headlines and Global News (HNGN).