AI Case Study
Researchers at UC Berkeley find 72 previously undetected bursts in data gathered on FRB 121102 using machine learning
Scientists from UC Berkeley leveraged machine learning to restudy the data gathered in 2017 on FRB 121102, the only fast radio burst (FRB) to date known to emit repeated bursts. While in the initial study 21 repeated bursts were detected using the Green Bank Telescope (GBT), a convolutional neural network enabled scientists to detect 72 further, previously undetected, bursts.
Public And Social Sector
Education And Academia
"Now, UC Berkeley PhD student Gerry Zhang and collaborators have developed a new, powerful machine learning algorithm, and reanalyzed the 2017 GBT dataset, finding an additional 72 bursts that were not detected originally. Zhang's team used some of the same techniques that internet technology companies use to optimize search results and classify images. They trained an algorithm known as a convolutional neural network to recognize bursts found with the classical search method used by Gajjar and collaborators, and then set it loose on the 400 TB dataset to find bursts that the classical approach missed."
R And D
Core Research And Development
"Fast radio bursts, or FRBs, are bright pulses of radio emission, just milliseconds in duration, thought to originate from distant galaxies. Most FRBs have been witnessed during just a single outburst. In contrast, FRB 121102 is the only one to date known to emit repeated bursts, including 21 detected during Breakthrough Listen observations made in 2017 with the Green Bank Telescope (GBT) in West Virginia.
The source and mechanism of FRBs are still mysterious. Previous studies have shown that the bursts from 121102 are emanating from a galaxy 3 billion light years from Earth, but the nature of the object emitting them is still unknown. Theories range from highly magnetized neutron stars, blasted by gas streams near to a supermassive black hole, to suggestions that the burst properties are consistent with signatures of technology developed by an advanced civilization.
In search of a deeper understanding of this intriguing object, the Listen science team at the University of California, Berkeley SETI Research Center2 observed FRB 121102 for five hours on August 26, 2017, using the Breakthrough Listen digital instrumentation at the GBT. Combing through 400 TB of data, they reported (in a paper led by Berkeley SETI postdoctoral researcher Vishal Gajjar, recently accepted for publication in the Astrophysical Journal3) a total of 21 bursts. All were seen within one hour, suggesting that the source alternates between periods of quiescence and frenzied activity."
"The results have helped put new constraints on the periodicity of the pulses from FRB 121102, suggesting that the pulses are not received with a regular pattern (at least if the period of that pattern is longer than about 10 milliseconds). Just as the patterns of pulses from pulsars have helped astronomers constrain computer models of the extreme physical conditions in such objects, the new measurements of FRBs will help figure out what powers these enigmatic sources."
"They trained an algorithm known as a convolutional neural network to recognize bursts found with the classical search method used by Gajjar and collaborators, and then set it loose on the 400 TB dataset to find bursts that the classical approach missed."
400 TB of GBT data gathered in 2017