Astronomy has entered the age of big data, in which astronomers are deluged with information thanks to the latest instruments and data exchange techniques. Facilities like the Vera Rubin Observatory (VRO) collect about 20 terabytes (TB) of data every day. Others, like the Thirty-Meter Telescope (TMT), are expected to collect up to 90 TB once operational. As a result, astronomers process 100 to 200 petabytes of data each year, and astronomy is expected to enter the “exabyte era” shortly.
In response, observatories have developed crowdsourcing solutions and made their data publicly available so that citizen scientists can assist in the time-consuming analysis process. In addition, astronomers are increasingly turning to machine learning algorithms to help them identify objects of interest (OI) in the Universe. In a recent study, a team led by the University of Georgia showed how artificial intelligence can simultaneously distinguish between false positives and exoplanet candidates, making the job of exoplanet hunters much easier.
The study was led by Jason Terry, a graduate student at the Center for Simulational Physics (CSP) at the University of Georgia (UGA) and a former researcher at Los Alamos National Laboratory (LANL). He was joined by researchers from the University of California San Francisco (UCSF), the Cardiovascular Research Institute (CRI) and the University of Alabama. The paper describing their research, Locating Hidden Exoplanets in ALMA Data Using Machine Learning, recently appeared in The Astrophysical Journal.
Remove all ads on Universe today
Join our Patreon for just $3!
Get the ad-free experience for life
Disk Substructures at High Angular Resolution Project. Credit: DSHARP
The first confirmed exoplanet was found in 1992, and the number has grown exponentially over the past fifteen years. To date, 5250 exoplanets have been confirmed in 3921 systems, while another 9208 candidates await confirmation. Still, the vast majority of these fall into one of three categories: Neptune-like (1,825), gas giants (1,630), and super-Earths (1,595). These planets are more massive and generally orbit farther from their stars than smaller, rocky (or “Earth-like”) planets, of which only 195 have been found.
Meanwhile, exoplanets that are in the formation phase are difficult to see for two main reasons: first, they are often hundreds of light-years from Earth (too far to see clearly), and second, the protoplanetary disks that make up they arise, very thick, with a diameter of up to 1 AU (the distance between Earth and Sun). From what astronomers have observed, planets tend to form in the centers of these disks, conveying a signature of the dust and gases that are kicked up in the process. But as Terry said in a recent AGU press release, research shows that artificial intelligence can help scientists overcome these difficulties:
“One of the novel things about this is analyzing environments where planets are still forming. Machine learning has rarely been applied to the kind of data we use before, especially to study systems that are still actively forming planets… We mostly analyze that data so you have tens, hundreds of images for a given disc and you just look through and ask, “Is that a wobble?” Then you run a dozen simulations to see if that’s a wiggle, and… it’s easy to miss them – they’re really tiny, and it depends on the cleaning, and so, first of all, this method is very fast , and secondly, their accuracy puts planets as far as humans would miss.”
For their study, the team developed a machine learning model based on computer vision (CV), a field of artificial intelligence that allows computers and systems to extract data from digital images and videos. The team trained their CV model with synthetic images it generated and then applied the model to real-world observations of protoplanetary disks made by the Atacama Large Millimeter-submillimeter Array (ALMA). In the end, they demonstrated that their machine learning method (based on CV) could correctly identify the presence of one or more planets in disks.
Atacama Large Millimeter/submillimeter Array (ALMA) located on the Chajnantor Plateau in the Chilean Andes. Photo credit: ESO
They further demonstrated that it could correctly constrain the position of the planets in these disks. Co-author Cassandra Hall, assistant professor of astrophysics and principal investigator in the Exoplanet and Planet Formation Research Group at UGA, explained:
“This is a very exciting proof of concept. The strength here is that we exclusively used synthetic telescope data generated through computer simulations to train this AI and then applied it to real telescope data. This has never been done before in our field and paves the way for a spate of discoveries when the James Webb Telescope data arrives.”
Several next-generation space-based and ground-based observatories will join the James Webb Space Telescope (JWST) in the coming years. These include the Nany Grace Roman Space Telescope (RST), the Extremely Large Telescope (ELT), the Giant Magellan Telescope (GMT), and the Thirty Meter Telescope (TMT). That and other telescopes will collect unprecedented amounts of data in multiple wavelengths, which will be used to search for exoplanets. In addition, the state-of-the-art instruments they will use will be able to characterize exoplanet atmospheres like never before. Terri said:
Beyond exoplanet research, these observatories will study cosmological mysteries such as dark matter and dark energy, and examine the earliest ages of the universe. Analyzing this high-quality data also requires next-generation analytical tools, allowing astronomers to spend more time interpreting the data and developing new explanatory theories. According to Terry, machine learning is already able to fill this need, saving time and money and efficiently managing scientific time, investment and new proposals:
“There remains a skepticism about machine learning and AI in science, and astronomy in general in particular, a valid criticism of it being that black box — where you have hundreds of millions of parameters and you kind of get an answer out of it. But we believe that in this work we have shown quite clearly that machine learning is up to the task. One can argue about the interpretation. But in this case we have very concrete results that demonstrate the power of this method.”
Further reading: UGA Today, The Astrophysical Journal
Like this:
Loading…
Comments are closed.