Skip to main content
Have a request for an upcoming news/science story? Submit a Request

Machine learning tool for astronomical data deluge developed on Brown community cluster

  • Science Highlights

A new telescope in Chile will soon be surveying the night sky more comprehensively than any before it, imaging the entire sky every three days, and looking deep into the universe thanks to its eight meter mirror. While it’s an incredible resource for the astronomy community, figuring out how to best make use of the near constant stream of data coming from the telescope, will be a challenge – one that a Purdue team is tackling with the help of ITaP Research Computing resources.

Dan Milisavljevic, assistant professor of physics and astronomy and his postdoctoral associate Niharika Sravan worked with Geoffrey Lentner, data scientist for Research Computing, to develop a tool, the Recommender Engine For Intelligent Transient Tracking (REFITT) that will absorb the telescope’s alerts about interesting astronomical events, and figure out which ones are worth pursuing.

Milisavljevic, who studies supernovae, or massive star explosions, likens the data stream that will come from the Vera Rubin Observatory’s Legacy Survey of Space and Time (LSST) to opening a fire hydrant. While current telescopes find supernovae at the rate of a couple thousand per year, the Rubin telescope is expected to find that many every day.

REFITT ingests all the data coming in from the telescope and uses a machine learning algorithm and known aspects of supernova behavior to make probabilistic inferences about the events, which gives the user insight into which events should be studied further and how.

The team relied heavily on the Brown community cluster, which is especially well suited to machine learning applications, to develop and train their algorithm, and backs up their data to the Fortress archive.

“I was excited about joining Purdue, knowing we have so many resources dedicated to getting the science done,” says Sravan, who has a background in data science and computational astronomy.

The REFITT project will also involve citizen scientists, in what Milisavljevic describes as “crowdsourcing Earth to watch the universe.” The goal is to give the many amateur astronomers out there guidance on the best place for them to point their telescopes, eliminating redundancy and making sure that as many interesting celestial events as possible have someone observing them.

To learn more about Purdue’s Community Cluster Program, contact Preston Smith, executive director of ITaP Research Computing, or 49-49729.

Originally posted: