The researchers at the Dutch university of Leiden have thought to train a neural network to detect possible threats in the solar System. They are, in fact, 11: asteroids with a diameter greater than 100 metres which will pass rather close to our Planet, a distance equal to ten times that which separates the Earth from the Moon. Forward and backward in time The method used is anything but simple and is the reverse of the ‘normal’ study of the prediction of the orbits. On the supercomputer of the University of Leiden, Alice, have done the ‘spin’ of the neuronal network that have called Hoi (which stands for " Hazardous object identifier, but in Dutch also means “hello”) and have pushed the button FFW, fast forward up to ten thousand years in the future. Then rewound the tape to the back, starting asteroids ‘virtual’ as if they had impacted the Earth. And have studied the distribution of the orbits of the date hereof.
In this way, “they have acquired a database of asteroids and hypothetical, the researchers knew that they would have hit the Earth,” reads the press release of the University of Leiden. And on this set of data the neural network is trained. We are still in the field of the virtual. To get to the real is enough to compare the database of the orbits of the virtual with those of real asteroids and it arrived the response. 11 asteroids, therefore, not been included in the 'black list' of Nasa and the Minor planet center, could be to keep an eye on.
Hoi has also been tested on the objects already present in the database of Nasa as “potentially hazardous”, riuscendone to identify the 90,99 percent. A good result, published in the journal Astronomy & Astrophysics, but you can do better: “we Now know that our method works, but obviously we'd like to investigate more thoroughly with help of a neural network with more data. The most complicated part is that even small perturbations to the calculations of the orbits may lead to large differences in conclusions.” explain the researchers.
Especially with a very long time, very high speeds, such as those of the heavenly bodies and great distances, a small perturbation of the orbit or a collision with another object can make the difference between a impact to touch me, and the book of revelation. The hope, conclude the authors of the study, is that in the future a neural network is well trained it can be used to detect objects that are potentially harmful. A method much faster than traditional ones, the spatial organization in use today. Notifying the arrival of an asteroid colliding with the Earth in advance, can think of a strategy to prevent the impact”.
"The Republic will fight always in defense of the freedom of information, to its readers and to all those who have at heart the principles of democracy and civil coexistence"Carlo Verdelli SUBSCRIBERS TO REPUBLIC © Reproduction reserved Today on “Aliens exterminated”. Massacre shakes Germany The bomber: “The secret service I spy”. The delirium of the bank that he wanted to wake up the whites That black monster that scares the Germany and Italy Instead of the one that sees Liliana Segre reason, the culture there aid