Computers beat brainpower in counting stars
A team of University of Sydney astronomers has developed a new way to automatically classify huge numbers of astronomical objects, and to discover new, exotic ones almost as soon as they happen.
Massive torrents of raw data are now collected by telescopes on a daily basis creating an urgent need to massively accelerate the reliable classification of millions of stars and galaxies, and to quickly highlight objects that might be new discoveries or that have unusual properties.
“Next generation telescopes like the Square Kilometre Array will produce enough raw data to fill up 15 million iPods every day,” said Kitty Lo, lead author of the research published in The Astrophysical Journal.
“It will be too much for humans to sift through, and this is where computer classification comes in,” said Ms Lo.
Ms Lo completed the research as part of her PhD thesis at CAASTRO at the University of Sydney.
“We were particularly interested in finding objects that suddenly appear and then disappear,” said co-author Dr Tara Murphy, CAASTRO Associate Investigator from Sydney University’s School of Physics.
These transient sources can represent entirely new phenomena, but can easily get buried in the many thousands of gigabytes of data generated by modern observatories. Astronomers need to find them as soon as they happen, so they can be studied with other telescopes before they fade away.
“Using a catalogue of known objects with existing classifications, we trained our software to recognise the flaring and bursting patterns of more than 800 objects, and to sort them into classes such as black holes, binary stars and supernova explosions,” said Dr Murphy.
The researchers then put their software to the test, confronting it with 400 new objects that needed to be identified and classified.
“Computer classification tends to be much better than humans at finding subtle patterns in data sets, especially if there are many similar categories,” said Ms Lo.
“Our software performed extremely well and matched the manual analysis in almost all cases. And crucially, it was much faster than any human ever could be.”
Not only did their software catalogue stars and galaxies extremely accurately and efficiently, it also singled out a number of strange objects that could not be classified and which could hold the key to new discoveries.
Detailed study of unusual objects will still remain the task of the human researcher. However, automatic classification can now let astronomers find exciting events as they happen.
The ARC Centre of Excellence for All-sky Astrophysics (CAASTRO) is a collaboration between The University of Sydney, The Australian National University, The University of Melbourne, Swinburne University of Technology, The University of Queensland, The University of Western Australia and Curtin University, the latter two participating together as the International Centre for Radio Astronomy Research (ICRAR). CAASTRO is funded under the Australian Research Council (ARC) Centre of Excellence program, with additional funding from the seven participating universities and from the NSW State Government’s Science Leveraging Fund.
Kitty K. Lo, Sean Farrell, Tara Murphy, Bryan M. Gaensler “Automatic classification of time-variable X-ray sources“, published in the Astrophysical Journal 786
Ms Kitty Lo: +44 07918962091, firstname.lastname@example.org
Dr Tara Murphy: +61 2 9351 3041, email@example.com
Dr Wiebke Ebeling (CAASTRO media contact): +61 8 9266 9174 or 0423 933 444, firstname.lastname@example.org
Verity Leatherdale (University of Sydney media contact): +61 2 9351 4312 or 0403 067 342, email@example.com