What do galaxies look like? How do they behave in the long term? These are some of the questions that Dr. Sebastian Wolfschmidt and Christopher Straub, researchers at the University of Bayreuth, are trying to answer.
They use mathematical models of galaxies that incorporate Einstein’s general theory of relativity, which explains how gravity affects space and time.
However, these models are not easy to test, as astronomical observations are limited and numerical simulations are time-consuming.
That’s why Wolfschmidt and Straub have developed a novel approach that uses artificial intelligence (AI) to quickly predict the stability of galaxy models.
A deep neural network for galaxy models
The AI tool that they use is called a deep neural network, which is a type of machine learning that mimics the structure of the human brain.
A deep neural network can learn from large amounts of data and recognize complex patterns.
Wolfschmidt and Straub have trained their deep neural network with hundreds of galaxy models, and then used it to classify new models as stable or unstable.
This way, they can check whether a model is realistic or not, and whether it agrees with the existing astrophysical hypotheses.
“The neural network can make a prediction in a few seconds, while a numerical simulation would take hours or days,” says Wolfschmidt, who is a research associate at the Chair of Mathematics VI at the University of Bayreuth.
“This allows us to efficiently verify or falsify many hypotheses that have been proposed over the past decades.”
A breakthrough for galaxy research
The results of Wolfschmidt and Straub have been accepted for publication in the prestigious journal Classical and Quantum Gravity.
They are the first to apply a deep neural network to galaxy models based on general relativity, and they have shown that it works well and accurately.
“We have been working on this project since 2019, in collaboration with the Chair of Applied Computer Science II—Parallel and Distributed Systems at the University of Bayreuth,” says Straub, who is a doctoral student at the Chair of Mathematics VI.
“We realized that machine learning can be very useful for some of our problems, and we have plans to use similar methods for other applications in the future.”
The calculations of the Bayreuth mathematicians were performed by the supercomputer of the “Keylab HPC” at the University of Bayreuth, which provides high-performance computing resources for scientific research.
The Bayreuth scientists are investigating the structure and long-term behavior of galaxies using mathematical models based on Einstein’s theory of relativity.
Astronomical observations have limitations, so mathematical models are used to gain insights into the intricate workings of galaxies, especially since most galaxies contain a black hole at their center.
Neural networks, inspired by the human brain, are powerful computational models used to detect complex structures in large amounts of data. In this research, they help predict which models of galaxies can exist in reality and which cannot, significantly speeding up the verification process.
More information: Christopher Straub et al, EVStabilityNet: Predicting the stability of star clusters in general relativity, Classical and Quantum Gravity (2024). DOI: 10.1088/1361-6382/ad228a