Researchers hope to use algorithms to improve multimessenger astronomy, and more precisely simulate evolutionary cosmic phenomena by automating certain discovery phases, and filtering massive datasets with optimal speed and efficiency.
For example, signals of gravitational waves collected by the Laser Interferometer Gravitational-Wave Observatory must be matched by supercomputers against hundreds of thousands of templates of potential wave signatures.
Scientists at the University of Illinois at Urbana-Champaign's National Center for Supercomputing Applications used convolutional neural networks (CNNs) for real-time detection/decryption of gravitational-wave signals.
The team then scaled up the initiative with supercomputer-trained deep learning algorithms, which search through a larger series of parameters to identify overlooked signals.
Meanwhile, Harvard University researchers developed a CNN to analyze x-ray images of galaxy clusters, and applied a technique allowing the network's observations to be visualized to give users a better idea of how the CNN was operating.
From Scientific American
View Full Article
Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA