The reporter learned from the Chinese Academy of Sciences on October 18 that using artificial intelligence technology, researchers from the Shanghai Observatory of the Chinese Academy of Sciences and other units discovered five ultra-short-period planets with diameters smaller than the Earth. Four of them are the smallest planets discovered so far and are similar in size to their host stars. This is the first time astronomers have used artificial intelligence to complete the task of searching for suspected signals and identifying real signals at once. Relevant research results were published online in the Monthly Notices of the Royal Astronomical Society.
Ultra-short period planets refer to those planets with an orbital period of less than 1 day. They orbit their host stars at an extremely close distance. They are usually smaller in size, lighter in mass, and have extremely high surface temperatures. So far, astronomers have found a total of 145 ultra-short period planets, of which only 30 have radii smaller than the Earth's. "Understanding the relative abundance and properties of ultra-short-period planets is crucial to testing theoretical models. However, the sample size of known ultra-short-period planets is too small, making it difficult to accurately understand their statistical characteristics and occurrence rates." The paper said Ge Jian, corresponding author and researcher at the Shanghai Observatory of the Chinese Academy of Sciences.
This time, Ge Jian's team innovatively designed a deep learning algorithm that combines GPU phase folding and convolutional neural networks. Using this algorithm, the team successfully discovered five ultra-short period planets in the stellar photometry data of the Kepler Space Telescope.
Ge Jian said that this research work began in 2015, when the artificial intelligence "AlphaGo" made a major breakthrough and successfully defeated the professional masters in the Go world. In addition to being motivated and inspired by his colleagues, he decided to try to apply the deep learning technology of artificial intelligence to the star photometry data collected by the Kepler space telescope to look for weak transit signals that could not be detected by traditional methods.
After nearly 10 years of hard work, Ge Jian’s team finally had its first harvest. Ge Jian believes that if you want to use artificial intelligence to "dig" extremely rare new discoveries in massive astronomical data, you need to innovate artificial intelligence algorithms and use large-scale data sets generated based on the physical image characteristics of newly discovered phenomena for training. It can quickly, accurately and completely detect rare and weak signals that are difficult to find in traditional ways.
Josh Winn, a professor at Princeton University, commented that ultra-short-period planets have extremely extreme and unexpected properties that provide clues to people's understanding of how planetary orbits change over time. This technological achievement in finding new planets is impressive.
"The discovery of these ultra-short-period planets provides important clues for the early evolution of planetary systems, planet-planet interactions, and the dynamics of star-planet interactions, and is of great significance to theoretical research on planet formation." Ge Jian said, this The research provides a new research method for quickly and efficiently searching for transit signals in high-precision photometric observation data, and also fully demonstrates the broad application potential of artificial intelligence in exploring weak signals in massive astronomical data.