Russian Scientists Create a Neural Network for Hunting Exoplanets
A new next‑generation neural network, ExoMiner++, accelerates the search for distant worlds by hundreds of times and dramatically improves the accuracy of detecting planets around faraway stars
Researchers at Ural Federal University have introduced ExoMiner++, a new-generation neural network that reshapes the process of detecting exoplanets—planets beyond the Solar System.
The system analyzes vast streams of data from space telescopes with exceptional precision, identifying real planets and filtering out false signals.
TESS: The Transit Hunter
The neural network was created by researchers at the university’s Kourovka Astronomical Observatory in collaboration with NASA Ames scientists. ExoMiner++ is trained on data from the Kepler and TESS missions.
TESS is an orbital telescope that measures the brightness of thousands of stars every two minutes. It searches for moments when a planet crosses the face of its star—causing its brightness to drop by fractions of a percent. These tiny dips are called transits. If the transit repeats at regular intervals, it is likely caused by a planet.
But there are hundreds of thousands of such dips, and most of them are misleading artifacts.
A Highly Intelligent Filter
The uniqueness of ExoMiner++ lies in its ability to learn and consider dozens of factors that would be extremely difficult for a human to check. The algorithm analyzes not just brightness dips but their shape, periodicity, background behavior, and even the technical nuances of the spacecraft. This turns it into a highly intelligent filter rather than merely a program.
The neural network’s first results are already impressive. It automatically analyzed 147,000 suspicious events and identified 7,330 of the most likely exoplanet candidates. At the same time, it re-evaluated previous datasets and confirmed only 1,797 out of 2,506 previously selected candidates—allowing astronomers to focus on the most promising targets.
The Future of Space Missions
The team plans to continue using ExoMiner++ to process new TESS data and adapt it for future space missions. In the coming years, the scale of astronomical observations will only increase, and such intelligent systems will become key to groundbreaking discoveries—helping scientists find new worlds orbiting distant stars.
