Apr 14, 2023 13:00:00

The image of the black hole photographed for the first time in history is reconstructed by machine learning and reborn into a more faithful one

In 2019,

a black hole was photographed for the first time ever . The images at that time were quite blurry, but by using a new machine learning technique called PRIMO (principal component interference modeling), we were able to clearly see the appearance of the black hole and create a more faithful image that is closer to reality. It was announced that a rebuild had taken place.

The Image of the M87 Black Hole Reconstructed with PRIMO - IOPscience
http://dx.doi.org/10.3847/2041-8213/acc32d

Iconic image of M87 black hole just got a machine-learning makeover | Ars Technica

https://arstechnica.com/science/2023/04/iconic-image-of-m87-black-hole-just-got-a-machine-learning-makeover/

1st-ever black hole image gets a sharp new AI makeover | Space

https://www.space.com/first-ever-black-hole-image-ai-makeover

In April 2019, the `` Event Horizon Telescope (EHT) '' project, which links radio telescopes around the world to form a virtual giant telescope on a global scale, succeeded in imaging a black hole for the first time in history. Although the appearance of the black hole is vague, it is highly valuable as an actual observation result, not the result of CG synthesis or simulation.

Succeeded in the first ever ``imaging of a black hole'', existing in the galaxy M87 55 million light years from the earth with a mass 6.5 billion times that of the sun - GIGAZINE

Lia Medeiros of the Institute for Advanced Study, Princeton and colleagues used a machine learning technique called PRIMO on the original data collected in 2017 that was used to create the published black hole image of the M87 galaxy. filled in the missing part of

This is the original image announced in 2019.

This is the result of applying PRIMO to the same dataset and reconstructing the image.

By matching the resolution of the EHT array after applying PRIMO, even in a slightly blurred state, it is this, so it is easier to see than the original state.

PRIMO is a dictionary learning that allows a computer to generate rules based on a large amount of training materials. To use it for black holes, 30,000 patterns of simulation images showing ``how black holes inhale gas'' are generated. I learned it.

The simulation image is something like this.


About the effort, Medeiros said, ``We're using physics and using machine learning to fill in missing data areas in a way that's never been done before. It may have important implications for ' interferometry ', which plays a role in a wide range of fields.'

By increasing the resolution of images using PRIMO, it is possible to more accurately estimate the characteristics of massive black holes, such as mass, size, and speed of swallowing matter.

“The 2019 imagery was just the beginning,” Medeiros said. It will continue to be an important tool for