The creation of an artificial intelligence tool that allows highly accurate data collection on exoplanets, which contributed to the implementation of the mission of the European Space Agency, won the researcher Louis Simoins in the Ariel Machine Learning Data Challenge.
The award was presented after Louis Simoins improved his ability to detect exoplanets by capturing the light emitted from a star using a highly accurate algorithm.
The task was launched by the European Space Agency (ESA) to provide the necessary conditions for the study of exoplanets (planets that orbit a star but do not belong to the solar system) in the Ariel mission, which will be launched in 2029. …
For the Portuguese researcher who “occasionally competes in machine learning for space,” the competition had the advantage of developing his “ability to solve problems and apply different algorithms” and keep abreast of developments in artificial intelligence. “.
To a problem that he considered “excellent because of its relevance and complexity,” he answered by creating a solution with an average error of 0.00007.
“The specific task – the detection of exoplanets – is solved by different approaches, but the one that turned out to be more successful is associated with the study of light curves, therefore, the light that comes to us from the star, and when the planet passes forward in its orbit, the shape that takes this curve tells us a lot about the planet, ”explained Luis Simoes to Luce.
He admits that this method has already been used to “identify thousands of exoplanets,” “but this mission has a very ambitious goal,” which includes “understanding the chemical composition of the atmospheres of these planets.”
“It will only be possible with tools that are not there today, hence the interest, 10 years ago, to start learning and challenge the community,” added Luis Simoins.
The solution he came up with does not provide any new data, but Luis Simoins created “a model, almost error-free (…) high-precision model.”
This accuracy matters, because “the question of the chemical composition of the atmosphere is the subsequent processing of the data that the algorithm will give.”
“There he already enters into other areas of astrophysics, but the fact is that if an incorrect interpretation is made at this stage of the data interpretation chain, then the next steps will be misled as to what the real characteristics of the planet will be.”, Materialized.
For this, the algorithm is programmed “automatically based on the training data”.
To this end, the team leading the ESA mission “has created synthetic data using simulators that recreate with maximum accuracy at the moment what data the mission will collect in the future.”
This data was distorted “due to data distortions that may arise during the mission – due to thermal fluctuations, due to the whole problem of measuring hundreds of light years, how many photons are coming from different sources”, but knowing what the original data was , it was possible to assess the degree of accuracy of the presented solutions.
“We are several years away, this is not the final model, but it is already a significant step towards achieving the levels of accuracy that the mission wants to have,” the researcher said.
Luis Simoins began working on the use of artificial intelligence for solving space problems, creating “systems for Airbus and ESA to control the landing of ships on other planets.”
In 2008, due to the crisis, he moved to the Netherlands, where he began to cooperate with the ESA.
He returned to Portugal in 2018, where he founded ML Analytics with his wife.
