Machine learning has been is used to discover 50 new alien planets in a first for astronomy.
A new “machine learning” algorithm has confirmed the existence of 50 new exoplanets in data collected by NASA’s ground-breaking Kepler mission.
Although algorithms have for long been used to comb through the huge amounts of data from telescopes in the hunt for signs of planets, this is the first time that astronomers have used an algorithm based on machine learning.
There are currently 4,201 confirmed exoplanets, though probably billions more to discover in the Milky Way alone. Some lurk in data collected years ago—as it the case with this new discovery.
With no input from humans, the algorithm created at the University of Warwick in the UK it was able to separate out real planets from fake ones in a large sample of thousands of “candidate planets.”
What is machine learning?
Machine learning is a form of artificial intelligence. It’s about automating repetitive tasks, essentially training a computer to recognise patterns and categorise data without any input from humans.
The example often given is that of photographs of cats and dogs. A computer program is given millions of images categorized as either cat or dog and the program then learns to identify them automatically; it creates a neural network.
How was machine learning used?
In this case, the algorithm was trained to recognise real planets using two large samples of confirmed planets and “false positives”—fake planets—from the now retired Kepler mission.
“The algorithm we have developed lets us take fifty candidates across the threshold for planet validation, upgrading them to real planets,” said lead author Dr. David Armstrong from the University of Warwick Department of Physics. “We hope to apply this technique to large samples of candidates from current and future missions like TESS and PLATO.”
The new machine learning algorithm in question was designed by researchers from University of Warwick’s Departments of Physics and Computer Science, as well as The Alan Turing Institute. Their results are published in the Monthly Notices of the Royal Astronomical Society.
“In terms of planet validation, no-one has used a machine learning technique before. Machine learning has been used for ranking planetary candidates but never in a probabilistic framework, which is what you need to truly validate a planet,” added Armstrong.
“Rather than saying which candidates are more likely to be planets, we can now say what the precise statistical likelihood is. Where there is less than a 1% chance of a candidate being a false positive, it is considered a validated planet.”
This is astronomy by numbers—statistical astronomy—which begs a big question:
The Transiting Exoplanet Survey Satellite (TESS) space telescope.
Does modern astronomy need humans?
Of course, it does, but our image of astronomers as brave and hardy academics that hang out on mountain tops operating big telescopes has to change. Modern astronomy is increasingly about big data, which naturally means that modern astronomers are just as likely to be data scientists designing powerful new algorithms and cutting-edge new machine learning techniques as they are experts on the night sky.
How to make discoveries from vast troves of big data is becoming a key concern of modern astronomy.
Why is machine learning important in astronomy?
It’s claimed that this new technique is faster than previous techniques, and can be both automated and improved with further training.
That’s important because exoplanet surveys produce huge amounts of data from sky-survey telescopes that contain signs of planets passing between the telescope and their star. That’s known as transiting—a kind of eclipse—that causes a very slight dip in light from the star.
Such a transit can only be detected if the telescope’s line of sight is spot-on—if it just happens to be looking at a particular star system side-on—but there are that many star system out there that this happens frequently.
However, that dip in the light curve of a star can also mean:
- there’s another star passing front of it.
- there’s some interference from another object.
- a technical problem with the telescope’s camera.
These are the false positives or “fake” planets that astronomers need to flush-out.
Now astronomers know that these 50 planets are real they can train telescopes one them in turn to tease-out more data.
Exoplanet surveys produce huge amounts of data from telescopes that contain signs of planets passing ... [+]
What are the 50 planets like?
Formerly viewed as a bit “iffy,” the 50 planets found using this new technique cover the entire gamut from smaller-than-Earth and likely rocky planets to Neptune-sized gas giants. Orbits range from as long as 200 days to as little as a single day.
“Almost 30% of the known planets to date have been validated using just one method, and that’s not ideal,” said Armstrong. “Developing new methods for validation is desirable for that reason alone. But machine learning also lets us do it very quickly and prioritise candidates much faster.”
Wishing you clear skies and wide eyes.
Read Again https://www.forbes.com/sites/jamiecartereurope/2020/08/25/does-astronomy-need-humans-an-algorithm-just-discovered-50-new-alien-planets-and-may-find-many-more/Bagikan Berita Ini
0 Response to "Does Astronomy Need Humans? An Algorithm Just Discovered 50 New Alien Planets And May Find Many More - Forbes"
Post a Comment