Researchers have taught AI to classify tens of thousands of galaxies

Astronomers have applied artificial intelligence to images of the distant universe with an ultra-wide field of view taken by the Subaru telescope. They have been able to achieve very high accuracy for finding and classifying spiral galaxies in these images. This method, combined with citizen science, is expected to lead to further discoveries in the future.

The research team, made up mostly of scientists from the National Astronomical Observatory of Japan (NAOJ), has applied deep learning, a type of AI, to classify galaxies in a large dataset of Subaru telescope images. Due to the high sensitivity, 560,000 galaxies were detected in the images. It would be extremely difficult to visually process this large number of galaxies one by one with human eyes for morphological classification. AI allowed the team to perform processing without human intervention.

Automated processing methods for extracting and evaluating features using deep learning algorithms have evolved rapidly since 2012. They are now generally superior to humans in accuracy and are used for autonomous vehicles, security cameras, and many other applications. Dr. Ken-ichi Tadaki, associate professor of the project at NAOJ, suggested the idea that if AI can classify images of cats and dogs, it should be able to distinguish “galaxies with spiral patterns” from “galaxies without spiral patterns.” Indeed, using training data prepared by humans, the AI ​​has successfully classified the morphology of galaxies with 97.5% accuracy. Then, by applying the trained AI to the full dataset, he identified spirals in about 80,000 galaxies.

Now that this technique has proven to be effective, it can be extended to classify galaxies into more detailed classes by training the AI ​​based on the significant number of them classified by humans. The NAOJ is currently implementing the civil science project Galaxy Cruise, in which citizens are studying images taken with a Subaru telescope in search of features that indicate that a galaxy is colliding or merging with another galaxy.

Galaxy Cruise project consultant Assistant Professor Masayuki Tanaka has high hopes for the study of galaxies using artificial intelligence.