ICTS:32948

Binary pulsar detector Network

(2025). Binary pulsar detector Network. SciVideos. https://scivideos.org/index.php/icts-tifr/32948

Binary pulsar detector Network. SciVideos, Oct. 12, 2025, https://scivideos.org/index.php/icts-tifr/32948 

          @misc{ scivideos_ICTS:32948,
            doi = {},
            url = {https://scivideos.org/index.php/icts-tifr/32948},
            author = {},
            keywords = {},
            language = {en},
            title = {Binary pulsar detector Network},
            publisher = {},
            year = {2025},
            month = {oct},
            note = {ICTS:32948 see, \url{https://scivideos.org/index.php/icts-tifr/32948}}
          }
Zhaocheng Gong
Talk numberICTS:32948
Source RepositoryICTS-TIFR

Abstract

Detecting pulsars in binary systems is crucial as they provide unique laboratories for testing gravitational theories. Traditional methods like Fourier Domain Acceleration Search and Fourier Domain `Jerk' Search, while effective, require significant computation resources and time to correct the effect of Doppler shifts, which limits real-time data processing and the efficiency of discovering binary pulsar systems.

LeoNet addresses this limitation by leveraging the fundamental principles of FDAS and FDJS and incorporating a filter template layer within a convolutional neural network. This layer contains minimal templates, allowing the neural network to optimize the filter parameters.  The filter template layer is divided into two parts, which process the real and imaginary parts of the spectrum, respectively. After convolution, the absolute value of the template layer result is calculated to generate the f_fdot plane for the classification task.

Several subsequent convolution layers and one fully connected layer act as a binary classifier to detect binary pulsar signals. The network continues to reduce the false negative rate while ensuring the false positive rate is 0 on the simulated dataset, reducing the number of binary pulsars that have not been detected. At the same time, it reduces the time required to search for binary pulsar signals. After using TensorRT, the speed of processing signals by the neural network has been improved.