Feature Matching Conditional GAN for Fast Radio Burst Localization with Cluster-fed Telescope

Feature Matching Conditional GAN for Fast Radio Burst Localization with Cluster-fed Telescope
复制标题

特征匹配条件 GAN,通过集群馈送望远镜进行快速无线电爆发定位

DOI:
10.3847/2041-8213/ab595e
复制
发表时间:
2019-12
期刊:
The Astrophysical Journal Letter
影响因子:
--
通讯作者:
Shizhong Yang
Shizhong Yang
中科院分区:
其他
文献类型:
--
作者:
Decheng Wu;Hailin Cao;Nanjie Lv;Jin Fan;Xiaoheng Tan;Shizhong Yang

文献摘要

参考文献

相似文献

对于实时瞬态定位,由于簇馈望远镜中的欠采样焦平面,快速射电爆发(FRB)位置存在很大的不确定性。提出了一种基于条件生成对抗网络 F3M-cGAN 的新型焦点场特征匹配算法,用于 FRB 定位。通过利用 FRB 的多频率特性,开发了一种改进的三维卷积自动编码器来联合合成焦场分布(FFD)特征。同时,将接收到的信号作为条件输入到鉴别器以加强映射关系。此外,引入基于深度神经网络的估计器来估计快速射电暴的位置,并设计位置编码器以根据能量分布来提高位置精度。对于图案重建,纹理损失用作生成器的反馈。基于F3M-cGAN框架的交替训练过程,可以实现有效的重建和位置估计。利用500米口径球面射电望远镜的缩比模型验证了该方法的有效性。仿真结果表明,该方法能够在有限馈源下准确重建FFD特征,并实现比射电望远镜角分辨率更高的实时定位精度。此外,所提出的方法可以增加瞬态的搜索区域。
For the real-time transient localization, there are large uncertainties of the fast radio burst (FRB) positions due to the undersampled focal plane in cluster-fed telescopes. A novel focal-field feature matching algorithm based on a conditional generative adversarial network, F3M-cGAN, is proposed for FRB localization. By exploiting the multiple frequency characteristics of FRBs, an improved three-dimensional convolution autoencoder is developed to jointly synthesize focal–field distribution (FFD) features. Meanwhile, the received signals are input as a condition to the discriminator to strengthen the mapping relation. Moreover, an estimator based on a deep neural network is introduced to estimate the position of the FRBs, and a position encoder is designed to enhance the position precision according to the energy distribution. For pattern reconstruction, texture loss is used as feedback for the generator. Based on the alternate training process of the F3M-cGAN framework, effective reconstruction and position estimation could be achieved. A scale model of the Five-hundred-meter Aperture Spherical radio Telescope is used to verify the effectiveness of the proposed method. The simulation results demonstrate that this approach can accurately reconstruct FFD features with limited feeds, and realize a higher real-time localizing precision than the angular resolution of a radio telescope. Furthermore, the proposed method can increase the search area for transients.
DOI: 10.1109/apusncursinrsm.2017.8073111
发表时间: 2017-07
期刊: 2017 IEEE International Symposium on Antennas and Propagation & USNC/URSI National Radio Science Meeting
影响因子: --
作者:
S. Smith;A. Dunning;K. Smart;R. Shaw;S. Mackay;M. Bowen;D. Hayman
通讯作者: S. Smith;A. Dunning;K. Smart;R. Shaw;S. Mackay;M. Bowen;D. Hayman
DOI: 10.1109/tap.2017.2763174
发表时间: 2018-02
影响因子: 5.7
作者:
O. Iupikov;M. Ivashina;N. Skou;C. Cappellin;K. Pontoppidan;C. V. '. van ’t Klooster
通讯作者: O. Iupikov;M. Ivashina;N. Skou;C. Cappellin;K. Pontoppidan;C. V. '. van ’t Klooster
DOI: 10.3847/2041-8213/aa71ff
发表时间: 2017-05
期刊: The Astrophysical Journal Letters
影响因子: --
作者:
K. Bannister;R. Shannon;J. Macquart;C. Flynn;P. Edwards;M. O’Neill;S. Osłowski;M. Bailes;B. Z
通讯作者: K. Bannister;R. Shannon;J. Macquart;C. Flynn;P. Edwards;M. O’Neill;S. Osłowski;M. Bailes;B. Z
DOI: 10.1017/9781108539258
发表时间: 2018-07
期刊: --
影响因子: --
作者:
K. Warnick;R. Maaskant;M. Ivashina;D. Davidson;B. Jeffs
通讯作者: K. Warnick;R. Maaskant;M. Ivashina;D. Davidson;B. Jeffs
DOI: 10.1126/science.aaf6807
发表时间: 2016-11
期刊: Science
影响因子: 56.9
作者:
V. Ravi;R. Shannon;M. Bailes;K. Bannister;S. Bhandari;N. Bhat;S. Burke-Spolaor;M. Caleb;
通讯作者: V. Ravi;R. Shannon;M. Bailes;K. Bannister;S. Bhandari;N. Bhat;S. Burke-Spolaor;M. Caleb;