A group of Australian researchers has developed an artificial intelligence system capable of classifying tens of thousands of galaxies in seconds, according to The Next Web.
The team developed a convolutional neural network (CNN) that classifies galaxy morphologies under a three-class scheme (elliptical, lenticular, spiral) and a four-class scheme (+ irregular/varied). The overall accuracy of the algorithm was 83% and 81% respectively, the researchers said.
According to them, the method’s advantages are its speed and its ability to generalise.
“Although training a CNN may be a costly undertaking in terms of computational power, the speed of galaxy classification after training is several orders of magnitude higher than manual processing,” the study says.
The neural network could classify more than 100 million galaxies at various distances from Earth and in different environments, they added.
“But if you put a group of astronomers in a room and ask them to classify several images, there will almost certainly be disagreements. This inherent uncertainty is a limiting factor in any AI model trained on labeled data,” he says.
Cavana added that the method could deepen understanding of how galaxies transform over time. He said it could even shed light on the nature of the universe itself.
Earlier this year, researchers at Pennsylvania State University, using machine learning discovered hidden connections between the Milky Way and Andromeda.
In May, American scientists unveiled an AI supercomputer that will help build the largest-ever 3D map of the visible Universe for studying dark energy.
In early May, Carnegie Mellon University researchers developed an artificial intelligence that models the Universe on a single GPU 1000 times faster than existing methods.
Subscribe to ForkLog News on Telegram: ForkLog AI — all the news from the AI world