300 new exoplanets discovered by an AI in Kepler’s data

The Kepler Space Telescope was NASA’s first telescope to be entirely dedicated to the search for exoplanets, that is, planets outside our solar system. During its mission, which ended in 2018, the spacecraft observed hundreds of thousands of stars to locate potential habitable planets. Today, the data collected continues to be analyzed even though the telescope is no longer in service. But to do this work, scientists are no longer alone since they are helped by Artificial Intelligence.

ExoMiner is an algorithm that can copy the procedure followed by researchers to find exoplanets, but in addition, it is faster and more efficient. In this context, ExoMiner has just discovered more than 300 new planets, so far gone unnoticed, in the data collected by Kepler.

To detect exoplanets, Kepler recorded decreases in star luminosity. These fluctuations can indeed be caused by the passage of a planet between its star and the telescope. Not all brightness reductions are caused by a planet, however, and the procedures developed by scientists were intended to distinguish false positives.

More efficient than humans

According to reports, ExoMiner is a neural network. It is an algorithm that can learn and improve using a large amount of data. As far as Kepler is concerned, the telescope has generated a fairly substantial amount of data. The NASA instrument has discovered thousands of candidates of which nearly 3,000 have been confirmed to be exoplanets.

For each candidate for the “title” of an exoplanet, scientists must analyze the light curves and calculate how much area of ​​the star the object covers. They then calculate how long the object crosses the star’s disk. This is the same process that the algorithm follows, and according to the researchers, AI is far more efficient than humans. Hamed Valizadegan, head of the ExoMiner project, explains that when the algorithm indicates that it is a planet, we can be sure that it is indeed a planet.

A capacity to exploit

With these very encouraging results, scientists are currently thinking of using the algorithm to review existing databases and those of future exoplanet search missions. There is, for example, NASA’s TESS or Transiting Exoplanet Survey Satellite mission, or ESA’s PLATO or Planetary Transits and Oscillations of Stars mission, which will be launched in 2026.

With this new tool, the search for new exoplanets will now accelerate. The more planets we discover outside our solar system, the more the chances of finding a habitable world will increase.

Leave a Comment