Star power

From a mountaintop in Chile, the Vera C. Rubin Observatory is preparing to capture the most detailed, decade-long movie of the night sky ever attempted. With the world's largest digital camera, the size of a small car, the telescope will capture billions of images of our changing universe, unmasking never-before-seen galaxies, exploding stars, asteroids, comets … and who knows what else hiding in the heavens?

Thousands of miles away in the Department of Physics and Astronomy on the University of Delaware campus, doctoral student Tatiana Acero-Cuellar will play a pivotal role in filtering and classifying these images for use by astronomers around the globe. She is developing an AI tool - a deep convolutional neural network inspired by how the brain recognizes patterns - to rapidly analyze the images and home in on objects of interest.

The neural network approach uses multiple or "deep" layers that learn to identify increasingly complex features in data. The "convolutional" component refers to the mathematical process that allows the system to first detect simple elements, such as edges, and then build up to recognizing shapes and complete objects.