[ad_1]
Perennial optimists from Search for Extraterrestrial Intelligence, or SETI, have joined the rest of the world to deploy the AI to manage huge data sets – and their efforts have almost instantly paid off. Seventy-two new "fast radio bursts" of a mysteriously noisy galaxy at 3 billion kilometers were discovered in previously analyzed data using a personalized machine learning model.
To be clear, this is not Morse code or encrypted instructions to create a teleporter, to the Contact, or at least not to our knowledge. But these fast radio bursts, or FRB, are poorly understood and may very well represent, at the very least, a previously unnoticed cosmic phenomenon. FRB 121102 is the only stellar object known to emit signals regularly, and is therefore the target of continuous observation.
The data comes from the Green Bank telescope in West Virginia (above), which was directed to this source of speed and brilliance (hence the name) bursts for five hours in August 2017. Believe it or no, this five-hour session gave 400 terabytes transmission data.
The initial "standard" algorithms identified 21 FRBs, all of which intervened in the hour of the observations. But Gerry Zhang, a graduate student at UC Berkeley and a participant in the Breakthrough Listen project, has created a convolutional neural network system that can more effectively navigate the data set. Of course, the model of machine learning has chosen 72 more FRB in the same period.
A Berkeley GIF visualizing data from a series of bursts.
This is quite an improvement, although it is interesting to note that without manual and traditional methods to find an initial set of interesting data, we would have little to train such neural networks. They are complementary tools. one does not necessarily follow the other.
The article on discoveries, co-written by Cal Postdoc Vishal Gajjar, is expected to be published in the Astrophysical Journal. Breakthrough Listen is one of the initiatives funded by billionaires Yuri and Julia Milner, of mail.ru and the fame of DST. The organization has issued its own press release for the work.
The new data suggests that signals are not received in any type of model that we can determine, at least no pattern of more than 10 milliseconds. It may seem daunting, but it is just as important to exclude things as to find something new.
"Gerry's work is exciting not only because it helps us understand more in detail the dynamic behavior of FRBs, but also to promise to use machine learning to detect signals missed by them." classical algorithms, "says Andrew Siemion of Berkeley. On-site SETI Research Center and Principal Investigator for Breakthrough Listen.
And if we are imaginative, there is no reason for hyper-advanced civilizations to be able to introduce interesting information into such bursts, or to use a motif that we have not yet understood. We do not know what we do not know, after all.
Anyway, SETI and Breakthrough will continue to keep their antennas attached to FRB 121102. Even if they are not foreign SOS signals, it's a good solid science. You can follow the work of the Berkeley SETI Center here.
Source link
