by Stars are the basic building blocks of our universe. Most stars organize planets as our sun houses our solar system, and when they look more general, groups of stars form huge structures such as clusters and galaxies. Before astrophysicists can try to understand these large -scale structures, we first have to understand basic properties of stars such as their mass, their radius and their temperature.
However, measuring these basic properties has proven to be extremely difficult. This is because stars are literally in astronomical distances. If our sun were a basketball on the east coast of the USA, the closest star, proxima, would be an orange in Hawaii. Even the world’s largest telescopes cannot solve an orange in Hawaii. Measuring radii and masses of stars does not seem to be within reach of the scientists.
Enter binary stars. Binary files are systems of two stars that revolve around a mutual mass center. Your application is subject to the harmonious law of Kepler, which connects three important quantities: the sizes of each orbit, the time you need for the orbits that are referred to as an orbital period and the total mass of the system.
I am astronomer and my research team has worked on driving our theoretical understanding and modeling approaches for binary stars and several star systems. In the past two decades, we have also worked the use of artificial intelligence in interpreting observations of these cornerstones of heavenly objects.
Measurement of star masses
Astronomers can measure the orbital size and period of a binary system easily enough from observations so that you can calculate the total mass of the system with these two parts. Kepler’s harmonious law acts as a scale for the weight of sky bodies.
Think of a playground. If the two children weigh about the same way, they have to sit roughly from the distance from the center. However, if a child is larger, it has to sit closer and the smaller child further away from the center.
It is the same with stars: the more massive the star in a binary couple, the closer to the middle and the slower it turns around the center. When astronomers measure the speeds where the stars move, you can also see how big the orbits of the stars are and what you have to weigh.
Measurement of star radii
Unfortunately, Kepler’s harmonious law tells the astronomers about the radii of the stars. For them, astronomers rely on another random characteristic of Mother Nature.
Binary stars are randomly aligned. Sometimes it happens that the line of sight of a telescope with the level circles a binary star system. This random alignment means that the stars put each other in the shade when they turn around the center. The shapes of these darkness make it possible to find out the radii of the stars with a simple geometry. These systems are known as binary stars.
More than half of all sun -like stars are located in binary files, and about 1% to 2% of all stars are eliminated in the sloping blades. That may sound low, but the universe is huge, so there are many, many solar substances – hundreds of millions in our galaxy alone.
By observing binary files in the shade, astronomers can not only measure the masses and radii of stars, but also how hot and how bright they are.
Complex problems require complex computing
Even with the binary files, it is not an easy task to measure the properties of stars. Stars are deformed when they rotate in a binary system and pull each other together. They interact, they irradiate each other, they can have stains and magnetic fields and they can be inclined on these or so.
To examine them, astronomers use complex models that have many buttons and switches. As input, the models take parameters – for example the shape and size of a star, its orbital properties or how much light it emits – to predict how an observer would see such an Eirting binary system.
Computer models need time. The computer model forecast typically takes a few minutes. To ensure that we can trust you, we have to try many parameter combinations – usually tens of millions.
These many combinations require hundreds of millions of minutes of calculation time to determine the basic properties of stars. This is over 200 years of computer time.
Computers that are linked in a cluster can calculate faster, but even with a computer cluster it takes three or more weeks to “determine or determine all parameters for a single binary date”. This challenge explains why there are only about 300 stars, for which astronomers have precise measurements of their basic parameters.
The models used to solve these systems have already been strongly optimized and cannot become much faster than already. Researchers therefore need a completely new approach to shorten the computer time.
Use Deep Learning
A solution that my research team has researched includes profound neural networks. The basic idea is simple: we wanted to replace a calculating physical model with a much faster AI-based model.
First of all, we calculated a huge database with predictions about a hypothetical binary star – using the characteristics that astronomers can easily observe – where we varied the properties of the hypothetical binary star. We speak hundreds of millions of parameter combinations. Then we compared these results with the actual observations to see which best match. AI and neural networks are ideally suited for this task.
In short, neural networks are assignments. They map a certain known input of a specific edition. In our case, they map the properties of binary files on the expected predictions. Neuronal networks emulate the model of a binary without taking into account the complexity of the physical model.
We train the neural network by showing every prediction from our database from the properties used for the generation from our database. As soon as the neural network is fully trained, it can predict exactly what astronomers should observe from the specified properties of a binary system.
Compared to a few minutes term for the physical model, a neural network uses artificial intelligence to achieve the same result within a tiny fraction of a second.
Use the advantages
A tiny fraction of a second corresponds to a millions of run -time reduction. This brings the time from weeks to a supercomputer on a single laptop for just a few minutes. This also means that in a few weeks we can analyze hundreds of thousands of binary systems on a computer cluster.
This reduction means that we can preserve basic properties – star masses, radii, temperatures and luminsities – for anyone in the broken binary star that has ever been observed within a month or two. The big challenge is to show that AI results really provide the same results as the physical model.
This task is the core of my team’s new paper. In it we showed that the AI-controlled model actually delivers the same results as the physical model in over 99% of the parameter combinations. This result means that the performance of the AI is robust. Our next step? Set the AI for all observed sun umbrella files.
The best of everything? While we applied this methodology on binary files, the basic principle applies to every complex physical model. Similar AI models already accelerate many real applications, from the weather forecast to stock market analysis.
This article will be released from the conversation, a non -profit, independent news organization that brings you facts and trustworthy analyzes to help you understand our complex world. It was written by: Andrej Prša, Villanova University
Read more:
Andrej Prša receives funds from the National Aeronautics and Space Administration.
