 |
| Matteo Paz with Caltech President Thomas F. Rosenbaum. Credit: California Institute of Technology |
A local high school kid expanded the possibilities of a NASA mission, produced a single-author paper, and discovered 1.5 million previously undiscovered objects in space through his study at Caltech.
The Astronomical Journal article by Matteo (Matthew) Paz details a novel artificial intelligence algorithm he created that produced these findings and that other astronomers and astrophysicists can use for their own studies.
Since his mother took him to Caltech's public Stargazing Lectures while he was in elementary school, Paz has been interested in learning more about astronomy. Under the direction of Professor of Astronomy Andrew Howard, he arrived on campus in the summer of 2022 to study astronomy and related computer science at the Caltech Planet Finder Academy.
Paz was mentored by Davy Kirkpatrick, a senior scientist at IPAC and an astronomer.
When Paz says, "I'm so lucky to have met Davy," "I recall telling him on our first conversation that I was thinking of writing a paper to finish this, which is a far bigger objective than six weeks. I wasn't deterred by him. "All right," he said, "let's discuss that." He has permitted an unrestricted educational experience. I believe that's the reason I've developed so much as a scientist.
Growing up in a farming hamlet in Tennessee, Kirkpatrick's goal of becoming an astronomer was fulfilled with the assistance of Marilyn Morrison, his ninth-grade chemistry and physics instructor. She outlined the courses he needed take to get ready for college and assured him and his mother that he had promise.
"I wanted to pass on that same sort of mentoring to someone else and hopefully many someone elses," adds Kirkpatrick. "I want to make sure they are achieving their potential if I see it. I'll do everything in my power to assist them."
Additionally, Kirkpatrick sought additional information from NEOWISE (Near-Earth Object Wide-field Infrared Survey Explorer), a now-retired infrared observatory that spent more than a decade searching the whole sky for asteroids and other objects close to Earth.
In addition to observing asteroids, the NASA telescope picked up the fluctuating heat of other cosmic objects farther away that flashed brightly, pulsed, or diminished as they were eclipsed. Quasars, exploding stars, and paired stars eclipsing one another are examples of hard-to-catch occurrences that astronomers refer to as variable objects.
However, there was still no way to use the data on these variable objects. The resulting catalog might shed light on how the cosmic entities evolve over time if the NEOWISE team could find those objects and make them accessible to the astronomical community.
"At that point, we were creeping up towards 200 billion rows in the table of every single detection that we had made over the course of over a decade," says Kirkpatrick. "So, I thought we could look for some variable stars in a small section of the sky this summer. 'Here's some fresh material we uncovered by ourselves; just image what the potential is in the dataset,' we might then tell the astronomical community.
Paz didn't plan on carefully sorting through the data.
His education had equipped him to approach the problem from a fresh
perspective. During an elective that included formal mathematics, theoretical
computer science, and coding, he developed an interest in artificial
intelligence.
Paz was aware that large, well-organized datasets, such as the one Kirkpatrick
had provided, are ideal for AI training. Additionally, Paz has the
sophisticated mathematical understanding required to appreciate programming: He
was already enrolled in Pasadena Unified School District's Math Academy, where
students complete AP Calculus BC in the eighth grade, to pursue advanced undergraduate
mathematics.
Paz therefore started working on creating a machine-learning method to examine the complete dataset and identify any potentially variable things. He started drafting the AI model during those six weeks, and it started to show some promise. In order to study the pertinent astronomy and astrophysics, he conferred with Kirkpatrick while working.
 |
| The anomaly extraction pipeline. Credit: The Astronomical Journal (2024). DOI: 10.3847/1538-3881/ad7fe6 |
"Every meeting with Davy is 10% work and 90% us just chatting," Paz states. "It's been super cool just to have someone to talk to about science like that."
The Caltech astronomers Shoubaneh Hemmati, Daniel Masters, Ashish Mahabal, and Matthew Graham were also introduced to Paz by Kirkpatrick. They offered their knowledge of machine-learning methods for astronomy and the analysis of things that change over short and long durations. Paz and Kirkpatrick discovered that NEOWISE's unique rhythm of observations meant that it couldn't reliably identify and categorize a lot of things that either changed gradually over time or flashed once fast.
There was still a lot to do as the summer came to an end. Paz and Kirkpatrick worked together once more in 2024, this time with Paz serving as a mentor to other high school students.
In order to evaluate all of the raw data from
NEOWISE's observations, Paz has now improved the AI model and examined the
outcomes. The algorithms identified and categorized 1.5 million possible new
objects in the data, trained to pick up on subtle variations in the telescope's
infrared observations. Paz and Kirkpatrick intend to provide the full list of
objects in the NEOWISE data that had significant brightness variations in 2025.
"The model I implemented can be used for other
time domain studies in astronomy, and potentially anything else that comes in a
temporal format," Paz explains. The information also appears in a time
series, and periodic components can be crucial, so I could see some connection
to chart analysis of the stock market. You could also research how the seasons
and day-night cycles affect the atmosphere, such as in the case of pollution.
Paz now works for Caltech as he completes his high school education. He is
employed by Kirkpatrick in IPAC, which is responsible for the management,
processing, archiving, and analysis of data from NEOWISE as well as a number of
other space missions that are supported by NASA and NSF. Paz is working for the
first time.
.jpg)
0 Comments