When the National Science Foundation announced a competition to fund institutes in astronomy and AI in the summer of 2023, physics and astronomy Prof. Vicky Kalogera brought together a team of researchers that included both astronomy and AI experts. Kalogera is the director of the Center for Interdisciplinary Exploration and Research in Astrophysics.
Gathering experts from institutions including Northwestern, the University of Chicago and the University of Illinois Urbana-Champaign, the team entered the competition. As one of two teams that won the competition, the team received a $20 million grant, allowing it to establish the NSF-Simons AI Institute for the Sky, or SkAI, in October 2024, with Kalogera as director.
The institute consists of more than 20 partner institutions and hopes to open up possibilities for the discovery of revolutionary sky surveys in order to help solve emerging problems related to astrophysics and AI. Around 60 faculty make up the team, with 22 from NU.
Physics and astronomy Prof. Tjitske Starkenburg, a research assistant professor in CIERA, and McCormick Prof. Emma Alexander are both members of the sky research team and researchers for the institute. They are currently working on developing new computer vision machine learning networks in order to reconstruct images of galaxies.
Starkenburg and Alexander are currently working on this project together, and they emphasized the diverse array of projects the institute is working on.
“They range across astronomy, from looking at stars, galaxies, the universe as a whole, so it’s across multiple scales on the astronomy side,” Starkenburg said. “And then on the AI side, we focus on uncertainty quantification, which is very challenging as well as generative models to accelerate astrophysical simulations but also help with interpretation of data and astrophysics.”
The institute is currently open to graduate students in engineering and the sciences who are pursuing their Ph.D. research through the institute. In the summer of 2026, undergraduates and high school students will have the opportunity to get involved as well.
Kalogera said astronomy is a data-driven field, and much of the advancement in understanding the cosmos is driven by data collected by telescopes.
“We’re going through a major revolution in data collection in the sense of building telescopes that can take a tremendous amount of data at very high speed and at very high volumes,” Kalogera said.”
Working with the Rubin Observatory in Chile, Kalogera said data is being collected at an unprecedented rate. According to Kalogera, for the next 10 years, SkAI will collect 15 to 20 terabytes of data in sky imaging from the observatory each night.
Kalogera said that by adding and subtracting images of the same location in the sky from each other, the photos can be combined to see very faint objects far away from Earth and to observe what’s changed between the images.
The AI part of the project comes into play due to the large amount of data being collected. Over 10 years, the volume of collected data will be equivalent to 3000 times the amount of data in the Library of Congress, Kalogera said.
Given the 40 billion objects of interest, Kalogera said having a person analyze the images from a screen is ineffective.
“We need automated algorithms, and we need algorithms that can actually not just make discoveries, but learn from the data, and not just discover things we already know and have seen before, but discover things that we don’t know and we have never seen before,” Kalogera said.
McCormick Prof. Aggelos Katsaggelos, deputy director of SkAI, has worked with Kalogera for over 10 years on astronomy research. Katsaggelos said he does research and teaching machine learning, deep learning and reinforcement learning.
Katsaggelos is involved in multiple projects within SkAI, including one on galaxy distortion and one on latent space. He emphasized how important stimulating stellar evolution is to society.
“It allows us to understand how a single star and two stars, binary stars, how they interact, by exchanging mass and adjusting their orbital period, and by stimulating large populations from 10 million to 10 billion stars,” Katsaggelos said. “It allows us to have realistic galaxy-scale systems and allows us to estimate important astronomical parameters.”
Kalogera emphasized how the data collected from the institute will deepen understanding of sciences from stars and galaxies, all the way back to the beginning of the universe with the Big Bang.
Email: [email protected]
Related Stories:
— Northwestern to develop $20 million AI Institute for astrophysics research
— Harvard professor talks asteroids, supernovas and dark matter at CIERA event
