Professor Casey DeRoo and graduate student, Dustin Swarm and undergraduate student, Samantha Watkins featured in the Iowa Magazine
SEARCHING FOR BLACK HOLES AND OTHER ASTRONOMICAL ANOMALIES
PHOTO: JUSTIN TORNER/UI OFFICE OF STRATEGIC COMMUNICATIONA UI team is teaching computers to categorize celestial objects on their own, then flag for further examination the phenomena that don't fit into neat categories.
Astronomy has a storied history of serendipitous discovery, notes Casey DeRoo (16PhD), UI assistant professor of physics. From Galileo training his telescope on Jupiter's moons for the first time in 1610 to Jocelyn Bell Burnell's discovery of pulsars while searching for quasars in 1967, sometimes the biggest breakthroughs happen by chance.
In that spirit, a UI research team is working to train AI to assist in making similar serendipitous discoveries in the cosmos. Partnering with Liu and under DeRoo's direction, physics and astronomy graduate student Dustin Swarm and undergraduate Samantha Watkins are training computers to comb through massive datasets to identify anomalies like new black holes and neutron stars that haven't previously been scrutinized by scientists.
"AI tools are going to have to be a part of every astronomer's toolkit. If you want to get something out of astronomical data, you need to be able to speak the language of artificial intelligence."— CASEY DEROO
The research team is currently analyzing the Chandra Source Catalog, a collection of 300,000-plus space objects documented by the satellite Chandra X-ray Observatory. "There are not enough professional X-ray astronomers in the world to sit down and look at 300,000 sources one by one, let alone perform a detailed analysis of them," DeRoo says, "so we have to make choices about what we're going to spend time analyzing. That's where we look to AI."
Essentially, says DeRoo, the UI team is teaching computers to categorize celestial objects on their own, then flag for further examination the phenomena that don't fit into neat categories. That includes supermassive black holes, pulsars, and other mysterious objects in our galaxy and beyond. With the AI's help, the UI team catalogued 119 outliers in the Chandra data set that are worthy of in-depth study and could potentially yield new discoveries.
The problem of how to analyze the enormous amounts of data being generated by observatories is a pressing one in astronomy. The Chandra X-ray dataset examined by the UI team, for example, amounts to about 50 gigabytes of information gathered over the past two decades. But DeRoo says that in the coming years, new observatories will come online that will generate that much data each day.
"This is really the direction the field is going," DeRoo says. "AI tools are going to have to be a part of every astronomer's toolkit. If you want to get something out of astronomical data, you need to be able to speak the language of artificial intelligence. These cross-disciplinary studies at the IIAI are helping make that happen."