SSKA is expected to generate several times the data traffic of the global Internet, and archive approximately 600PB of data every year. Such a massive amount of astronomical data far exceeds the limits of existing computing power. Therefore, astronomical data is accumulated and is idly stored in equipment rooms. Therefore, real-time observation and analysis cannot be performed over a large area of the sky, or for a specific star, which means that precious opportunities are missed for observing and interpreting astronomical phenomena. To resolve this conundrum, researchers will need to depend on powerful algorithms and hardware infrastructure with powerful computing, buttressed by artificial intelligence for crucial assistance.
Huawei has worked closely with the Shanghai Astronomical Observatory to explore new applications for AI in astronomical research. In order to address longstanding challenges related to star classification, Huawei converts radio signals into images based on the observation data of SKA precursor facilities, before converting the observed astronomical data into visible image data through image segmentation and pre-processing. A deep learning method is then used to train the AI model to recognize different star types. This process has helped identify the type and position of more than 200,000 stars in the database.
While astronomers currently need 169 days to identify and locate objects in the southern hemisphere, the Atlas 900, equipped with the Ascend 910 AI chipset, only requires 10 seconds to complete the task, demonstrating the HUAWEI CLOUD Ascend cluster service’s powerful computing capabilities. In addition, the Shanghai Observatory utilized the image recognition and classification technology capabilities of HUAWEI CLOUD ModelArts to detect pulsars. By preprocessing the data set, which contained 1196 pulsar positive samples and 89996 non pulsar negative samples, analyzing the pulse signal matrix, and converting it into a gray-scale image, the ModelArts proved capable of determining a pulsar with a staggering accuracy of 99%!
In addition to the basic hardware, the Atlas 900 comes equipped with HUAWEI CLOUD Ascend cluster services, utilizing full-stack optimization, from the underlying hardware to the developmental architecture, for cutting-edge computing capabilities. Drawing from the standout AI algorithm and computing power, the Shanghai Astronomical Observatory and HUAWEI CLOUD aim to jumpstart a new revolution in astronomical research, promote China’s participation in the groundbreaking SKA project, and inspire new revelations about the nature of our universe.