Real-time abnormal light curve detection based on a Gated Recurrent Unit network

Ruiqing Yan, Wei Liu, Meng Zhu, Yijing Wang, Cong Dai, Shuo Cao, Kang Wu, Yuchen Liang, Xianchuan Yu, Mengfei ZhangResearch 2020 · DOI

Abstract

Abstract Targeting the problem of high real-time requirements in astronomical data processing, this paper proposes a real-time early warning model for light curves based on a Gated Recurrent Unit (GRU) network. Using the memory function of the GRU network, a prediction model of the light curve is established, and the model is trained using the collected light curve data, so that the model can predict a star magnitude value for the next moment based on historical star magnitude data. In this paper,we calculate the difference between the model prediction value and the actual observation value and set a threshold. If the difference exceeds the set threshold, the observation value at the next moment is considered to be an abnormal value, and a warning is given. Astronomers can carry out further certification based on the early warning and in combinationwith other means of observation. Themethod proposed in this paper can be applied to real-time observations in time domain astronomy.
BibTeX
@inproceedings{yan2020real,
  author = {Ruiqing Yan and Wei Liu and Meng Zhu and Yijing Wang and Cong Dai and Shuo Cao and Kang Wu and Yuchen Liang and Xianchuan Yu and Mengfei Zhang},
  title = {{Real-time abnormal light curve detection based on a Gated Recurrent Unit network}},
  booktitle = {Research},
  year = {2020},
  doi = {10.1088/1674-4527/20/1/7},
}

Loading related content...

Figure