Up front: Astronomers classify galaxies by shape to understand how they form and evolve. But this can be a time-consuming job. The researchers used convolutional neural network (CNN) architectures to hasten the task. Per the preprint paper: The team developed a CNN architecture that outperforms existing models in classifying the morphologies of galaxies in both 3-class (elliptical, lenticular, spiral) and 4-class (+irregular/miscellaneous) schema. Its overall classification accuracies were 83% and 81% respectively. They say it will be able to classify more than 100,000,000 galaxies at different distances from Earth and in different environments. Quick take: The main advantage of using AI to classify galaxies is speed. But lead study author Mitchell Cavanagh, a PhD student at the International Centre for Radio Astronomy Research (ICRAR), said the accuracy is also improving: Ultimately, the technique could deepen our understanding of how galaxies transform over time. Cavanagh says it could even shed light on the nature of the universe itself.
