AI machines can help researchers analyze gravitational lensing images more quickly
(Mirror Daily, United States) – Gravitational lensing is a really useful phenomenon, as it helps astronomers analyze distant galaxies and cosmic objects. However, looking at an object distorted by it might take weeks or months of analysis, until researchers managed to find a solution. They trained an AI with thousands of simulated images, and it’s now able to get information from real images affected by gravitation in just a matter of seconds.
What is gravitational lensing?
Gravitational lensing occurs when the image of a small and distant cosmic object gets magnified by a bigger object placed in front of it. For instance, a faraway galaxy might be distorted by a massive cluster of galaxies coming between it and the observing researchers. In other words, the gravity of these bigger objects offered a better observation of smaller and distant bodies.
Gravitational lensing is extremely useful. So far, it has helped researchers find exoplanets, black holes, bright galaxies, understand how these evolve and, in the end, find evidence in support of Einstein’s theory of relativity.
AI helps researchers decode these images much more quickly
However, when researchers analyze images distorted by gravitational lensing, they need to compare them with simulated ones. This usually takes from several months to several weeks, so it’s quite a slow process. This is why they thought of using AI technology to ease their work.
Several neural networks received 500,000 images of simulated gravitational lensing that it had to analyze. During testing, the networks were able to extract more precise information than the classical methods, and over a shorter timespan. Also, they could learn on their own what they needed to look for in these images.
Researchers developed a study on the subject, which was published in the journal Nature. As technology and telescopes continue to evolve, they will be able to capture more instances of gravitational lensing. Therefore, training neural networks to extract information from them is a great way to get all the needed data more quickly than ever.
Image Source: Wikimedia Commons