A new computer program developed and trained by astronomers has the potential to classify tens of thousands of galaxies within a few seconds. It is a task that generally takes months to execute.
Different shapes of galaxies, left to right: elliptical, lenticular, spiral, and irregular/miscellaneous. Image Credit: NASA/Hubble (elliptical galaxy M87), ESA/Hubble & NASA (lenticular galaxy NGC 6861 and the colliding Antennae galaxies), and David Dayag (the Andromeda spiral galaxy).
In a new study, astrophysicists from Australia employed machine learning to accelerate a process that is usually performed manually by citizen scientists and astronomers across the globe.
Galaxies come in different shapes and sizes. Classifying the shapes of galaxies is an important step in understanding their formation and evolution, and can even shed light on the nature of the Universe itself.
Mitchell Cavanagh, PhD Candidate and Study Lead Author, The University of Western Australia Node, International Centre for Radio Astronomy Research
With wider surveys of the sky occurring all the time, astronomers are gathering too many galaxies to analyze and classify, added Mr Cavanagh.
“We’re talking several million galaxies over the next few years. Sometimes citizen scientists are recruited to help classify galaxy shapes in projects like Galaxy Zoo, but this still takes time.”
Convolutional neural networks (CNNs) play a crucial role exactly here. In the present high-tech world, computer programs of this type are all over the place, finding use in everything right from stock markets, medical imaging, and data analytics, to how Netflix produces suggestions depending on the users’ viewing history.
In the past years, CNNs have started to find major use in astronomy. A majority of the CNNs utilized by astronomers are binary, wherein there is a doubt if it is a spiral galaxy or not. However, this new CNN makes use of multiclass classification, giving rise to a question of whether it is a lenticular, elliptical, irregular, or spiral galaxy, with more precision compared to current binary networks.
According to Mr Cavanagh, machine learning has found extensive use in the field of astronomy.
The massive advantage of neural networks is speed. Survey images that would otherwise have taken months to be classified by humans can instead be classified in mere minutes. Using a standard graphics card, we can classify 14,000 galaxies in less than 3 seconds.
Mitchell Cavanagh, PhD Candidate and Study Lead Author, The University of Western Australia Node, International Centre for Radio Astronomy Research
Cavanagh continued, “These neural networks are not necessarily going to be better than people because they’re trained by people, but they’re getting close with more than 80% accuracy, and up to 97% when classifying between ellipticals and spirals.”
“If you place a group of astronomers into a room and ask them to classify a bunch of images, there will almost certainly be disagreements. This inherent uncertainty is the limiting factor in any AI model trained on labeled data,” added Cavanagh.
The latest AI holds one great benefit with which the scientists will be able to categorize over 100,000,000 galaxies at various distances (or redshifts) from Earth and in different surroundings (clusters, groups, etc). Hence, this would help the researchers gain better insights into how galaxies are being transformed with respect to time, and why this might occur in specific surroundings.
The CNNs designed by Mr Cavanagh are not only for astronomy. They could be repurposed and used in various other fields, as long as they have a highly large dataset to train with.
CNNs will play an increasingly important role in the future of data processing, especially as fields like astronomy grapple with the challenges of big data.
Mitchell Cavanagh, PhD Candidate and Study Lead Author, The University of Western Australia Node, International Centre for Radio Astronomy Research
Journal Reference:
Cavanagh, M. K., et al. (2021) Morphological classification of galaxies with deep learning: comparing 3-way and 4-way CNNs. Monthly Notices of the Royal Astronomical Society. doi.org/10.1093/mnras/stab1552.