Deep-CEE: The AI Model Helping Astronomers Find Galaxy Clusters

Galaxy clusters, made up of several galaxies bound together by gravity and dark matter, are giants of the universe.

To put things in perspective, our own Milky Way galaxy is estimated to be home to about 250 billion stars. 

The problem is that, despite being millions of lightyears across, they also tend to be millions of lightyears away from us, making them hard for astronomers to spot.

Enter Deep-CEE (Deep Learning for Galaxy Cluster Extraction and Evaluation), a deep learning technique developed by researchers at Lancaster University. The AI was built to find galaxy clusters much faster than any human would be capable of.

RELATED: AN AI UNIVERSE SIMULATOR SO ACCURATE THAT ITS CREATORS DON'T FULLY UNDERSTAND IT

Understanding dark matter

Scientists have found the main factor binding galaxy clusters is dark matter. As such, learning more about these extreme environments can help us better understand the mysterious properties of dark matter and dark energy.

In the 1950s, astronomer George Abell found the 'Abell catalog' of galaxy clusters after analyzing roughly 2,00 photographic plates of the observable universe.

Deep-CEE, built by Matthew Chan, a PhD student at Lancaster University, builds on Abell's approach but replaces the astronomer with an AI model trained to search through color images in order to identify galaxy clusters.

The AI was trained by being shown examples of tagged, known objects until it was able to associate objects by itself. Pilot tests then demonstrated Deep-CEE's ability to be trained onto galaxy clusters.

Huge amounts of data

"We have successfully applied Deep-CEE to the Sloan Digital Sky Survey" Chan said in a press statement. "Ultimately, we will run our model on revolutionary surveys such as the Large Synoptic Survey telescope (LSST) that will probe wider and deeper into regions of the Universe never before explored.

Huge amounts of data are generated by telescopes on a daily basis. The upcoming LSST sky survey (due in 2021), for example, will generate an estimated 15 TB of data every night in order to image the entire skies of the southern hemisphere.

"Data mining techniques such as deep learning will help us to analyze the enormous outputs of modern telescopes," says Dr. John Stott (Chan's Ph.D. supervisor). "We expect our method to find thousands of clusters never seen before by science".

Chan is set to present his AI model and the findings of his paper, "Fishing for galaxy clusters with "Deep-CEE" neural nets" on 4 July at 3:45 pm in the 'Machine Learning in Astrophysics' session.