University of Hawaii researchers are harnessing the power of the world's largest solar telescope and advanced artificial intelligence to transform their understanding of the sun
Astronomers and computer scientists at the University of Hawaii's Institute for Astronomy are conducting groundbreaking research that could completely change our understanding of the Sun.
As part of the SPIn4D project, the team combines advanced solar astronomy with the latest developments in computer science to process data collected by the world's largest ground-based solar telescope located atop Maui's Mount Lakala.
The research is focused on developing deep learning models that can quickly analyze the large amount of data generated by the Inoue telescope. The purpose of this activity is to maximize the telescope's capabilities, and pave the way for significant progress in the speed, accuracy and depth of analyzing data from the sun.
"Large solar storms are responsible for amazing displays of the aurora borealis, but also pose a danger to satellites, radio communications and power grids. It is very important to understand the place of their formation, the atmosphere of the sun," said Kai Yang, the lead researcher. "We used state-of-the-art simulations to simulate what the rover would see. Combining this data with machine learning provides a valuable opportunity to study the 3D solar atmosphere in near real time."
The Inoue Solar Telescope, located atop Maui's three-kilometer Mount Lakala, is the world's most powerful solar telescope by far. The telescope's instruments are designed to measure the Sun's magnetic field using polarized light, and the SPIn4D project was specifically designed to use this data, which is only available from the Sun Telescope's instruments. .
deep neural networks
The team of scientists is using deep neural networks to estimate the physical properties of the solar photosphere from the high-resolution observations of Inoue. This method promises to significantly speed up the analysis of the large amounts of data produced by the solar telescope, which can reach tens of terabytes per day.
"Machine learning has a very good ability to quickly give approximations to expensive calculations. In this case, the model will allow astronomers to create a breast of the Sun's atmosphere in real time, instead of waiting hours to get the same accuracy," said co-author Peter Sadowski.
To train their AI models, the team created an extensive dataset of simulated solar observations. They used more than ten million CPU hours of the Shane supercomputer and created 120 terabytes of data that mimics Inoue's observations in very high resolution.
The researchers have already made 13 terabytes of their data publicly available, along with a detailed user guide. They intend to publish the deep learning models after they complete the training as a tool for the community to analyze the observations of the Inoue Space Telescope.
More of the topic in Hayadan:
