Classifying Galaxies with Artificial Intelligence…

Namaskaram!

Today we have an application of AI in Astronomy. The Universe is vast. So mind-boggling is its vastness that astronomers need the use of technology to find and classify astronomical objects.

The Subaru Telescope, located in Hawaii, USA, had taken numerous images of galaxies from Earth's vantage point. Thanks to its high sensitivity, as many as 560,000 galaxies have been 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 (shape) classification. The AI technique enabled astronomers of the NAOJ (National Astronomical Observatory of Japan) to process these galaxies without human intervention.

Deep Learning based image classification techniques have been used to classify images based on their pixel-data. You might have heard of the "dog" and "cat" classifier. Well, it turns out Deep Learning is also good at distinguishing galaxies "with spiral patterns" and galaxies "without spiral patterns". This falls under a larger class of machine learning called "Supervised Learning" where labelled data can be used to train models which can be used to make predictions (classification probabilities) on unseen data. 

The AI model successfully classified the galaxies with an accuracy of 97.5%. Then applying the trained AI to the full data set, it identified spirals in about 80,000 galaxies. Now that the technique has been proven to be effective it can be used to train and identify galaxies into more detailed classes. This technique, in combination with citizen science, is expected to yield further discoveries in the future.

Credit: National Astronomical Observatory of Japan

The research paper that was published at arXiv can be found here.

Good day!