Jakarta, CNBC Indonesia \u2014 Artificial intelligence (AI) is beginning\nto help astronomers search for hidden Earth-like planets.<\/p>\n
A machine-learning model developed by researchers at the University\nof Bern and Switzerland\u2019s National Centre of Competence in Research\n(NCCR) Planet identified at least 44 planetary systems that could\ncontain undiscovered Earth-like planets.<\/p>\n
The study, published in the journal Astronomy & Astrophysics,\nfound that the model achieved precision of up to 99% when tested on\ncomputer-simulated planetary systems.<\/p>\n
However, the predicted planets around actual stars have not yet been\nconfirmed by telescope observations. The method\u2019s success still depends\non further observations.<\/p>\n
The AI model uses the arrangements of known planets to look for clues\nthat other, unseen planets may exist. A planet\u2019s mass and orbit can\npreserve traces of how an entire planetary system formed and evolved,\nincluding evidence of planets too small or faint for telescopes to\ndetect.<\/p>\n
Planets form together in discs of gas and solid material surrounding\nyoung stars. As they grow, migrate and interact with one another, these\nprocesses leave patterns in their masses and orbital distances.<\/p>\n
The researchers tested whether those patterns could help distinguish\nsystems containing small, temperate planets from those without them.\nSome systems have planets of similar sizes and at similar distances, a\npattern often described as \u201cpeas in a pod\u201d, while others have more\nvaried arrangements.<\/p>\n
Jeanne Davoult developed the machine-learning algorithm during her\ndoctoral research at the University of Bern, working with Romain\nEltschinger and Yann Alibert.<\/p>\n
The researchers found that the strongest clues came from the\nobservable architecture of a system, as well as the mass and orbital\nperiod of its innermost detected planet.<\/p>\n
The team faced a challenge in training the model on observed\nplanetary systems: astronomers rarely know all the planets orbiting a\nstar. Smaller planets with longer orbits are also harder to find,\nleaving the picture obtained from observations incomplete.<\/p>\n
The team therefore used the Bern Model of Planet Formation and\nEvolution to generate tens of thousands of synthetic planetary systems.\nThe simulations covered stars with masses equivalent to that of the Sun,\nhalf the Sun\u2019s mass and one-fifth of the Sun\u2019s mass.<\/p>\n
In these simulations, every planet was known, including those that\nobservers might not be able to detect. The calculations began with\nplanetary embryos inside discs of gas and small solid particles.<\/p>\n
The researchers then simulated growth, migration and gravitational\ninteractions over 20 million years. Subsequent calculations tracked\ncooling, contraction, atmospheric loss and tidal migration for up to 10\nbillion years.<\/p>\n
Before training the model, the team concealed planets whose\ngravitational effect on their star fell below a set detection threshold.\nThis was intended to reproduce the limitations of the radial-velocity\nmethod, which measures a star\u2019s movement caused by an orbiting\nplanet.<\/p>\n
The model used a random-forest method comprising 500 decision trees.\nEach tree cast a vote on whether a system contained an Earth-like\nplanet.<\/p>\n
The researchers used 80% of the synthetic data to train the model and\nthe remaining 20% to test it.<\/p>\n
They prioritised precision: the proportion of systems that actually\ncontain an Earth-like planet among all those the model classified as\nhaving one.<\/p>\n
When the voting threshold was set above 90%, precision reached 99%\nfor models trained on the two populations of higher-mass stars. The\nmodel for the lowest-mass stars achieved precision of 94%.<\/p>", "url": "https:\/\/jawawa.id\/newsitem\/ai-identifies-44-previously-hidden-earth-like-planet-candidates-1792759842", "image": "" }, "sponsor": "Okusi Associates", "sponsor_url": "https:\/\/okusiassociates.com" }