RSSNet: A Novel Network Model for FRBs Search

Fast Radio Brusts (FRBs) are often subjected to a large number of false-positive samples and radio-frequency interference (RFI) during the search process, making it difficult to achieve an efficient search. In order to optimize the search results and improve the search efficiency, this paper proposes a FRBs search network model which named RSSNet (Radio Source Search Network), and adopts the gradient-weighted class-activation mapping (Grad-CAM) method to provide an interpretable analysis for this model. At the same time, the influence of the signal-to-noise ratio (SNR) of the dynamic spectral on the search is also studied, and the search results of dynamic spectra with high SNR (SNR ⩾ 11) and those with low SNR (SNR < 11) in the test set are compared and analyzed respectively. The experimental results show that RSSNet is the best in low SNR dynamic spectrum search compared with other networks, with an accuracy of 99.37%. Based on the above work, an automatic FRBs search system is designed to realize the automatic data reading and dynamic spectrum based FRBs recognition.