Despite the spread of AI -based research lately, researchers have sometimes needed a human eye to make real discoveries. This collaboration was in a recent work by Dr. Veselin Kostov, a research scientist at the SETI institute and the Goddard Space Flight Center of NASA, a team of almost 1,800 to check a data record from the transit exoplanet -Survey satellites (Tess).
A binary system in the shade is a star system in which two stars circle, one in front of the other in our view. The way to find them is similar to that of exoplanet – look at a star and look for dips in brightness. If the dip is large enough, there is probably another star instead of an exoplanet (albeit a weak one), which circles this star.
Tess data is great for this work because they cover around 98% of the sky that observe exactly this type of transits. However, this does not mean that the researchers only presented Tess data directly to a number of volunteers. The data has pre -processed several steps before they were handed over to the public.
A look at how a binary (and even an exoplanet) can influence the light curves of Star. Credit – NASA GSFC
First, they limited the data record to only stars with a strength that is brighter than 15 years old. After narrowing the sheer number of stars to look at them, they used a tool developed in Python, the Eleanor Pipeline developed to create a massive data set of millions of light curves. These light curves were then artificially padded and scaled to a uniform number of data points to ensure that regular changes in Tess's observation instrument were not confused with solar eclipses.
After all this preliminary processing, the researchers fed their millions of updated light curves to (they probably guessed it) to a AI. This is a relatively simple folding network that has been trained to find the shape of a solar eclipse and not a certain periodic signal, which it better finds in the solar eclipse regardless of its periodicity. It was trained on some data with manually marked results and then subjected to the tess and even Kepler data on solar eclipse on the catalog. Around 85% of the well -known dark nine files were successfully found in Tess's data, and around 56% of them from Kepler's data records. Interestingly, around 32% of the exoplanet candidates in Tess's data record was also possible.
Even after all this AI processing, the data record was not quite ready for the help of volunteers. The research team as well as a selected group of trained amateurs used a platform called Exogram to identify 10,000 initial goals, which were then published on zoonive, a crowd-based research platform. Between September 2024 and March 2025, 1,800 volunteers carried out 320,000 classifications of Eclipsing bineme systems, at the same time their period checked and the quality of the data that was used for identification was assessed.
https://www.youtube.com/watch?v=YB7IMAUEZCO
Fraser discusses the discoveries of Tess
The result of all work led to the identification of 10.001 in the shade binary systems. 7,936 of them are new to science, while the other 2,065 were known, but the study provided updated, more precise parameters for their periods, since Tess' data record provided better insights. There were also some particularly interesting systems that were able to record new discoveries, including several that had a variable solar eclipse time, and a lot that could have a third star, and some that show a significant dynamic between the surrounding star and the circuit.
All of these systems expect further research, but there is another unspoken factor in this data – exoplanets. Tess was originally designed as an exoplanet hunter, and this type of KI/Human cooperation from Lightcurve -Analysis is exactly the kind of work that could possibly create even more precise exoplanet catalogs, as some of the work already done in this article. That seems to be the next step for this data record. Dr. Kostov tells an interviewer: “I can't wait to look for exoplanets!” In view of the data, the data has already been collected and the team has already been put together, it is very likely that he will soon get his chance.
Learn more:
NASA – NASA citizens find new umbrellas
V Kostov et al.-The Tess Tenusklichen catalog: 10,001 uniformly nicer and validated in the umbrella in image data in the umbrella in the umbrella in image data after machine learning and analyzed by civil scientists
Ut -astronomy -jargon 101: darkness
Ut – rare stars in the shards provide sophisticated measurements in the universe
Comments are closed.