301 new exoplanets in one fell swoop. Number to add to 4569 who Presence It has already been confirmed by astronomers. But how did researchers discover so many new things? worlds Once? The answer lies in Artificial intelligence. Specifically, in a new deep neural network called ExoMiner.
Neural networks are machine learning algorithms that are able to learn a task on their own when enough data is provided. And ExoMiner is exactly that: a deep neural network that harnesses the power of a supercomputer in NASAAnd chandelierIt can distinguish between true exoplanets and different types of rogues or “false positives”. Its design is inspired by many of the tests and characteristics that human experts use to confirm the existence of new exoplanets.
ExoMiner learns using worlds that have already been confirmed in the past and from cases that have been shown to be false positives. The system is very useful in supplementing astronomers and helping them analyze data to find out what a planet is and what it is not.
Specifically, the researchers applied the neural network to the massive amount of data collected over the years by the space mission. Kepler, one of the biggest “planet hunters” at NASA. The probe, which has thousands of different stars in its field of view, each with the ability to host one or more planets, produces a massive amount of data that takes a long time to be analyzed by astronomers. But not for ExoMiner, which can do it a thousand times faster.
“Unlike other machine learning programs for discovering exoplanets -explain John Jenkins, from NASA’s Ames Research Center- ExoMiner is not a black box; There is no mystery as to why he decided that something was a planet or not. We can easily explain the characteristics of the data that lead ExoMiner to reject or confirm a planet ».
Verified and confirmed
Before they can be completely sure that a new exoplanet has been discovered, they must go through two different states: verification and confirmation. The planet is “validated” using statistics, that is, how likely it is that an object captured by the instruments is or is not a planet, according to the available data. At this point one speaks of “candidate planets”. The filter does not move to a “certain” state until various observational techniques reveal properties that can only be explained by the presence of a planet.
In an article published in “Astronomical Journal“In this article, the Ames team shows how ExoMiner discovered the 301 planets using data from the candidate planet collection in the Kepler Archives. The 301 planets confirmed by the instruments were originally discovered at the Kepler Science Operations Center and have been promoted to planet candidate status by the Kepler Science Office. But so far no one has been able to verify its authenticity.
The article also shows how ExoMiner is more accurate and consistent in eliminating false positives and better able to detect true signatures of planets orbiting their parent stars. And all in a transparent way, where scientists can see in detail what led ExoMiner to its conclusions.
“When ExoMiner says something is a planet, you can be sure that it is a planet -Dice Hamid ValizadeganProject Leader. ExoMiner is highly accurate and in some ways more reliable than current automated classifiers and human experts, due to the biases that accompany human labeling ».
None of the newly confirmed planets are believed to be Earth-like or located in the habitable zone of their parent stars. But they share characteristics similar to the general population of confirmed exoplanets in the galaxy’s vicinity.
“These are 301 discoveries – says Jenkins- Help us better understand planets and solar systems beyond our constellation, and what makes our planet so unique ».
As the search for more exoplanets continues to expand with missions such as NASA’s Transiting Exoplanet Exploration Satellite, or he-goatAnd the next task PLAnetary transits and oscillations of starsAnd NS plateauFrom the European Space Agency, ExoMiner will have more opportunities to show that it is up to the task.
«Now that we have trained ExoMiner using Kepler data -includes Valizadegan-, With a little modification, we can transfer this learning to other tasks, including TESS, that we are currently working on ».
Source: Jose Manuel Neves/ABC,
“Beer enthusiast. Subtly charming alcohol junkie. Wannabe internet buff. Typical pop culture lover.”