Concat Convolutional Neural Network for pulsar candidate selection

Abstract

Pulsar searching is essential for the scientific research in the field of physics and astrophysics. As the development of the radio telescope, the exploding volume and it growth speed of candidates growth have brought about several challenges. Therefore, there is an urgent demand for developing an automatic, accurate and efficient pulsar candidate selection method. To meet this need, this work designed a Concat Convolutional Neural Network (CCNN) to identify the candidates collected from the Five-hundred-meter Aperture Spherical Telescope (FAST) data. The CCNN extracts some "pulsar-like" patterns from the diagnostic subplots using Convolutional Neural Network (CNN) and combines these CNN features by a concatenate layer. Therefore, the CCNN is an end-to-end learning model without any need for any intermediate labels, which makes CCNN suitable for the online learning pipeline of pulsar candidate selection. Experimental results on FAST data show that the CCNN outperforms the available state-of-the-art models in similar scenario. It only misses 4 real pulsars out of 326 totally.

3D detection and characterization of ALMA sources through deep learning

“…The deconvolution process ideally provides the noiseless observation 𝐼 (𝑥, 𝑦) from the observed measurements and it is traditionally performed making assumptions about the structure of both the signal and of the forward operator. Many attempts have been made at solving this problem using Machine Learning (ML) based approaches (Bowles et al 2020;Zeng et al 2020;Schmidt, K. et al 2022;Rezaei et al 2021;Connor et al 2022). In this paper, we present a deep-learning-based pipeline for the detection and characterization of sources within ALMA "uncleaned" or "dirty" calibrated data cubes.…”

“…Moreover, they employed the produced attention maps (Hassanin et al 2022) to help the interpretation of the results. Zeng et al (2020) proposed a novel deep-learning pipeline, Concat COnvolutional Neural Network (CCNN), for selecting pulsar candidates within the Commensal Radio Astronomy FasT Survey (Li et al 2018). The main idea behind their pipeline is to use several specialized CNNs to extract the low-dimensional latent information from the pulse profile, the Dark Matter profile, the frequency versus phase plot, and the time versus phase plot.…”

Pulsar Candidate Sifting Using Multi-input Convolution Neural Networks

“…Recently, Guo et al (2019) used DCGAN to balance the pulsars and non-pulsars in the training set and combined it with SVM and enhance the accuracy of the automatic pulsar candidate identification. Zeng et al (2020) designed a Concat Convolutional Neural Network (CCNN) to identify the candidates collected from the FAST data using the four diagnosic plots as inputs and improved candidate sifting performance evidently. Therefore, SML methods have been successfully applied in pulsar candidate sifting.…”

“…A model based on CNNs often involves some basic types of layers such as convolutional layers, downsampling layers (pooling layers), and fully connected layers (dense layers). Some researchers have constructed models based on CNNs to sift the pulsar candidates and most of them achieved satisfactory performances in their specific surveys (Zhu et al 2014;Guo et al 2019;Wang et al 2019a,b;Zeng et al 2020).…”

DECORAS: detection and characterization of radio-astronomical sources using deep learning

“…Machine learning, and in particular, CNNs have already been widely used in the analysis of radio interferometric data. For example, they have been used to classify radio galaxies (Bowles et al 2021), to determine galaxy morphologies (Cheng et al 2020), and to select pulsar candidates (Zeng et al 2020). More specifically, CNNs have been employed to detect astronomical sources within the C S (Lukic et al 2019), D S (Vafaei Sadr et al 2019) and Point Proposal Network (PPN; Tilley et al 2020).…”