NASA wants to find out stars’ properties with machine learning

NASA astronomers have used machine learning to study the properties of thousands of stars in the Milky Way. The scientists hope to discover new patterns that people don’t notice.

The astronomers use machine learning computer systems to map certain basic properties of stars. Photographs of the universe are used for this. Normally, data about the spectrum of a star is needed to determine the properties of the celestial body using the different wavelengths, a time-consuming task for astronomers. However, NASA’s machine learning systems are content with photo collections because they can recognize patterns from which the properties of stars can be traced.

According to NASA, the machine learning system’s algorithms make it possible to analyze billions of stars in a relatively short time. Thus, the size of a star can be determined, as well as which metals are present. In addition to the speed gain, the software also saves costs.

The software algorithms first had to be ‘trained’ in recognizing stars. For this, NASA used a file containing 9,000 stars. By also feeding the algorithms with data files containing the light curves of the stars, the software was able to recognize patterns after a while. The NASA astronomers have documented their findings with the system so far in a research report.