A new artificial intelligence algorithm has found more than 300 previously unknown exoplanets in data collected by the defunct Exoplanet Hunting Telescope.
NS Kepler space telescope, NASA’s first dedicated exoplanet hunter, has explored hundreds of thousands of stars in search of potentially habitable worlds outside of our own. Solar system. The collection of potential planets you have amassed continues to make new discoveries even after the telescope dies. Human experts analyze data for signs of exoplanets. But a new algorithm called ExoMiner can now simulate this process and clean up the catalog faster and more efficiently.
The telescope, which had ceased operations in November 2018, was looking for a temporary decrease in stellar brightness that might be caused by a planet passing in front of the star disk from Kepler’s perspective. But not all of this darkness is the reason outer planet, and scientists must follow detailed procedures to distinguish false positives from real things A NASA statement.
ExoMiner is what is called a neural network, a type of artificial intelligence algorithm that can learn and improve its skills if it receives enough data. And Kepler produced a wealth of data: in less than 10 years, telescopes have found thousands of planetary candidates, nearly 3,000 of which have been confirmed. That’s a big part of the total 4,569 exoplanets currently known.
Scientists looking at the Kepler data will look at the light curve for each exoplanet candidate and the fraction size of. to calculate Stern It looks like the planet is closed. They will also analyze how long it would take a potential planet to traverse the star disk. In some cases, the observed changes in brightness are unlikely to be explained by an exoplanet orbiting the solar system. The ExoMiner algorithm follows exactly the same process, but is more efficient and allows researchers to simultaneously add a previously unknown group of 301 exoplanets to Kepler’s catalog of planets.
“When ExoMiner says something is a planet, you can be sure that it is a planet,” said Hamid Valizadegan, ExoMiner project leader and director of machine learning at the University Space Research Consortium at NASA Ames Research Center, in the statement. “ExoMiner is extremely accurate and in some ways more reliable than current machine classifiers and human experts that it should emulate due to the distortions inherent in the human label.”
Now that ExoMiner has proven its expertise, scientists want to use it to examine data from current and future exoplanet search missions such as current NASA missions. Satellite transit to survey the outer planets (TESS) or the Transiting Planets and Star Oscillation (PLATO) mission of the European Space Agency, which will start in 2026.
Unfortunately, there are no newly confirmed exoplanets that are likely to harbor life as they are outside of their parent star’s habitable zone.
In a statement, NASA said it had accepted the paper for publication in the Astrophysical Journal. A Concept paper Available for reading in prepress arXiv.org.