Our solar system is enormous. Eight planets, hundreds of moons, asteroid belts. Gas and ice giants. Planet Earth. Ex-planet Pluto.
The Milky Way, which contains the solar system, is even bigger — containing an estimated 100 billion planets.
PhD student Matthew Chan has just developed a groundbreaking AI which is capable of identifying objects which are even bigger still. It searches ultra high-resolution telescope images for images of ‘cluster galaxies’ — collections of up to 100 galaxies, which are often millions of light years in length.
Why? Finding them can teach us much more about deep space — the purpose of X-ray gases, the use of dark energy and dark matter.
Maybe it can even tell us what else is out there.
Science Daily relays the story of pioneering astronomer George Abell, who spent years in the 1950s analysing thousands of photographic plates by eye — using a magnifying glass to identify cluster galaxies. His seminal work is known as the Abell catalogue of galaxy clusters in the northern hemisphere.
In the present day, the newest high-power telescopes are set to take huge numbers of images of the sky in the Southern hemisphere. From 2021 the Large Synoptic Survey telescope will generate an estimated 15TB of data every night — equivalent to downloading three millions songs every single day.
That’s far too much data to go through by eye.
Chan’s solution, created as part of a doctorate at the University of Lancaster, is an AI that can identify galaxy clusters from colour images.
It’s called Deep-CEE (Deep Learning for Galaxy Cluster Extraction and Evaluation), and is based on neural networks — replicating the way a human brain activates specific neurons when recognising objects.
Chan trained the machine-learning programme until it could associate objects on its own, then tested the algorithm to see whether it could identify and classify galaxy clusters in images that contain other astronomical objects.
To date, it’s been successful. The hope is that the deep learning will allow scientists to study enormous data outputs and find thousands of clusters “never seen before by science.”
“We have successfully applied Deep-CEE to the Sloan Digital Sky Survey” Chan says. “Ultimately, we will run our model on revolutionary surveys such as the Large Synoptic Survey telescope (LSST) that will probe wider and deeper into regions of the Universe never before explored.”
Nichole Onome Yembra