Pulsar Candidates Research Articles

Pulsar Candidate Classification Using a Computer Vision Method from a Combination of Convolution and Attention

Artificial intelligence methods are indispensable to identifying pulsars from large amounts of candidates. We develop a new pulsar identification system that utilizes the CoAtNet to score two-dimensional features of candidates, implements a multilayer perceptron to score one-dimensional features, and relies on logistic regression to judge the corresponding scores. In the data preprocessing stage, we perform two feature fusions separately, one for one-dimensional features and the other for two-dimensional features, which are used as inputs for the multilayer perceptron and the CoAtNet respectively. The newly developed system achieves 98.77% recall, 1.07% false positive rate (FPR) and 98.85% accuracy in our GPPS test set.

Asymmetric gradient penalty based on power exponential function for imbalanced data classification

Model bias is a tricky problem in imbalanced data classification. An asymmetric gradient penalty method is proposed based on the power exponential function to alleviate this. The methodology integrates a power exponential function as a moderator into the cross-entropy loss of the negative samples, driving the model to focus on hesitant samples while ignoring easy and singular samples. The rationality of the algorithm is explored from the gradient point of view, and it is demonstrated that the approach improves focal loss and asymmetric focal loss. Then, the imbalanced data classification experiments were deployed on MNIST, CIFAR10, CIFAR100, and Caltech101, respectively. For binary classification, datasets with several imbalance ratios constituted by varying the sample size of the majority class and minority class are included in the experiments. In the multi-category classification experiments, we discuss imbalanced datasets with only a single majority category and those with several majority categories and examine step-imbalance datasets and linear-imbalance datasets. The results reveal that the proposed method exhibits competitiveness on various imbalanced datasets and better robustness on high imbalance ratio datasets. Finally, the approach is deployed on the pulsar candidate dataset HTRU, and the state-of-the-art results are yielded. Our code is available at https://github.com/gzmtzly/GPPE.

Advances in Pulsar Candidate Selection: A Neural Network Perspective

Pulsars have played an important role in comprehending the universe. They play a key role in understanding various phenomena like general relativity, gravitational waves, properties of matter, collision of black holes and the evolution of stars and nebulae. Thus, identifying them is a crucial task. The increasing number of surveys has created a large volume of candidate samples, in the range of several million. Hence, it is impossible to select pulsars from these samples using human-driven methods. Automatic Pulsar Candidate Identification (APCI) was introduced for this purpose. In recent years, various deep-learning techniques and models have been implemented for this purpose. Specific deep neural network models and hybrid models were designed to select pulsar candidates from various surveys consisting of radio and X-ray samples. In this study, a series of models implementing ANN, CNN and GNN are discussed capable of selecting pulsar candidates. These models were trained using a wide range of surveys.

HAWC Study of the Very-high-energy γ-Ray Spectrum of HAWC J1844−034

Recently, the region surrounding eHWC J1842−035 has been studied extensively by γ-ray observatories due to its extended emission reaching up to a few hundred TeV and potential as a hadronic accelerator. In this work, we use 1910 days of cumulative data from the High Altitude Water Cherenkov (HAWC) observatory to carry out a dedicated systematic source search of the eHWC J1842−035 region. During the search, we found three sources in the region, namely, HAWC J1844−034, HAWC J1843−032, and HAWC J1846−025. We have identified HAWC J1844−034 as the extended source that emits photons with energies up to 175 TeV. We compute the spectrum for HAWC J1844−034, and by comparing with the observational results from other experiments, we have identified HESS J1843−033, LHAASO J1843−0338, and TASG J1844−038 as very-high-energy γ-ray sources with a matching origin. Also, we present and use the multiwavelength data to fit the hadronic and leptonic particle spectra. We have identified four pulsar candidates in the nearby region in which PSR J1844−0346 is found to be the most likely candidate due to its proximity to HAWC J1844−034 and the computed energy budget. We have also found SNR G28.6−0.1 as a potential counterpart source of HAWC J1844−034 for which both leptonic and hadronic scenarios are feasible.

MFPIM: A Deep Learning Model Based on Multimodal Fusion Technology for Pulsar Identification

With the development of radio telescope technology, the quantity and types of acquired pulsar candidate data have increased dramatically. However, it is difficult to accurately identify pulsar candidates. Therefore, we propose to use multimodal fusion technology, called the multimodal fusion-based pulsar identification model (MFPIM), to build a deep learning model to improve the efficiency and accuracy of pulsar candidate identification. MFPIM treats each diagnostic plot of pulsar candidates as a modality and uses multiple convolutional neural networks to extract effective features from the diagnostic plots. After fusing the features, the commonality of different modalities in high-dimensional space is obtained to ensure that the model can take full advantage of the complementarity between diagnostic plots and thus identify pulsar candidates, achieved better classification performance than other current supervised learning algorithms. In addition, a channel attention mechanism is used in the model to enable it to learn the importance of different channel features so that the model focuses more on the channel information in the input data that is more meaningful for classification, reducing the model size while extracting pulsar diagnostic map features more accurately. We conducted experiments on the Five-hundred-meter Aperture Spherical radio Telescope (FAST) data set, and the results show that MFPIM can effectively identify the pulsars in the FAST data set with an identification accuracy of over 98%. To further verify the robustness of the model, we applied the MFPIM to the High Time Resolution Universe data set using transfer learning, with the test accuracy and F1 score reaching over 99%.

A new pulsar candidate in 47 Tucanae discovered with MeerKAT imaging

ABSTRACT MeerKAT imaging of the globular cluster 47 Tucanae (47 Tuc) reveals 1.28 GHz continuum emission at the locations of 20 known millisecond pulsars (MSPs). We use time series and spectral imaging to investigate the image-domain characteristics of the MSPs, and search for previously unknown sources of interest. The MSPs exhibit a range of differences in their temporal and spectral properties compared the general background radio source population. Temporal variability differs strongly from pulsar to pulsar, some appearing to vary randomly on 15 min time-scales, others varying coherently by factors of >10 on time-scales of hours. The error in the typical power law fit to the spectrum emerges as a powerful parameter for indentifying the MSPs. This behaviour is likely due to differing diffractive scintillation conditions along the sightlines to the MSPs. One MSP exhibits tentative periodic variations that are consistent with modulation due the orbit of an eclipsing binary system. One radio source has spectro-temporal properites closely resembling those of the MSP population in the cluster, and we report its position as a candidate new MSP, or alternatively an interferometric localization of one of six MSPs which do not yet have an accurate position from the timing solutions.

A new redback pulsar candidate 4FGL J2054.2+6904

ABSTRACT The Fermi catalogue contains about 2000 unassociated γ-ray sources. Some of them were recently identified as pulsars, including so-called redbacks and black widows, which are millisecond pulsars in tight binary systems with non- and partially-degenerate low-mass stellar companions irradiated by the pulsar wind. We study a likely optical and X-ray counterpart of the Fermi source 4FGL J2054.2+6904 proposed earlier as a pulsar candidate. We use archival optical data as well as Swift/XRT and SRG/eROSITA X-ray data to clarify its nature. Using Zwicky Transient Facility data in g and r bands spanning over 4.7 yr, we find a period of ≈7.5 h. The folded light curve has a smooth sinusoidal shape with the peak-to-peak amplitude of ≈0.4 mag. The spectral fit to the optical spectral energy distribution of the counterpart candidate gives the star radius of 0.5 ± 0.1 R⊙ and temperature of 5500 ± 300 K implying a G2–G9-type star. Its X-ray spectrum is well fitted by an absorbed power law with the photon index of 1.0 ± 0.3 and unabsorbed flux of ≈2 × 10−13 erg s−1 cm−2. All the properties of 4FGL J2054.2+6904 and its presumed counterpart suggest that it is a member of the redback family.

No MSP Counterparts Detected in GBT Searches of Spider Candidates 4FGL J0935.3+0901, 4FGL J1627.7+3219, and 4FGL J2212.4+0708

We performed radio searches for the “spider” millisecond pulsar (MSP) candidates 4FGL J0935.3+0901, 4FGL J1627.7+3219, and 4FGL J2212.4+0708 using the Green Bank Telescope in an attempt to detect the proposed radio counterpart of the multi-wavelength variability seen in each system. We observed using the VEGAS spectrometer, centered predominantly at 2165 MHz; however, we were also granted observations at 820 MHz for 4FGL J1627.7+3219. We performed acceleration searches on each data set using PRESTO as well as additional jerk searches of select observations. We see no evidence of a radio counterpart in any of the observations for each of the three systems at this time. Additional observations, perhaps at different orbital phases (e.g., inferior conjunction), may yield detections of an MSP in the future. Therefore, we urge continued monitoring of these systems to fully characterize the radio nature, however faint or variable, of each system.

Discovery of 37 new pulsars through GPU-accelerated reprocessing of archival data of the Parkes multibeam pulsar survey

ABSTRACT We present the discovery of 37 pulsars from ∼ 20 yr old archival data of the Parkes Multibeam Pulsar Survey using a new FFT-based search pipeline optimized for discovering narrow-duty cycle pulsars. When developing our pulsar search pipeline, we noticed that the signal-to-noise ratios of folded and optimized pulsars often exceeded that achieved in the spectral domain by a factor of two or greater, in particular for narrow duty cycle ones. Based on simulations, we verified that this is a feature of search codes that sum harmonics incoherently and found that many promising pulsar candidates are revealed when hundreds of candidates per beam even with modest spectral signal-to-noise ratios of S/N∼5–6 in higher-harmonic folds (up to 32 harmonics) are folded. Of these candidates, 37 were confirmed as new pulsars and a further 37 would have been new discoveries if our search strategies had been used at the time of their initial analysis. While 19 of these newly discovered pulsars have also been independently discovered in more recent pulsar surveys, 18 are exclusive to only the Parkes Multibeam Pulsar Survey data. Some of the notable discoveries include: PSRs J1635−47 and J1739−31, which show pronounced high-frequency emission; PSRs J1655−40 and J1843−08 belong to the nulling/intermittent class of pulsars; and PSR J1636−51 is an interesting binary system in a ∼0.75 d orbit and shows hints of eclipsing behaviour – unusual given the 340 ms rotation period of the pulsar. Our results highlight the importance of reprocessing archival pulsar surveys and using refined search techniques to increase the normal pulsar population.

The TRAPUM L-band survey for pulsars in Fermi-LAT gamma-ray sources

ABSTRACT More than 100 millisecond pulsars (MSPs) have been discovered in radio observations of gamma-ray sources detected by the Fermi Large Area Telescope (LAT), but hundreds of pulsar-like sources remain unidentified. Here, we present the first results from the targeted survey of Fermi-LAT sources being performed by the Transients and Pulsars with MeerKAT (TRAPUM) Large Survey Project. We observed 79 sources identified as possible gamma-ray pulsar candidates by a Random Forest classification of unassociated sources from the 4FGL catalogue. Each source was observed for 10 min on two separate epochs using MeerKAT’s L-band receiver (856–1712 MHz), with typical pulsed flux density sensitivities of $\sim 100\, \mu$Jy. Nine new MSPs were discovered, eight of which are in binary systems, including two eclipsing redbacks and one system, PSR J1526−2744, that appears to have a white dwarf companion in an unusually compact 5 h orbit. We obtained phase-connected timing solutions for two of these MSPs, enabling the detection of gamma-ray pulsations in the Fermi-LAT data. A follow-up search for continuous gravitational waves from PSR J1526−2744 in Advanced LIGO data using the resulting Fermi-LAT timing ephemeris yielded no detection, but sets an upper limit on the neutron star ellipticity of 2.45 × 10−8. We also detected X-ray emission from the redback PSR J1803−6707 in data from the first eROSITA all-sky survey, likely due to emission from an intrabinary shock.

Image-based searches for pulsar candidates using MWA VCS data

Abstract Pulsars have been studied extensively over the last few decades and have proven instrumental in exploring a wide variety of physics. Discovering more pulsars emitting at low radio frequencies is crucial to further our understanding of spectral properties and emission mechanisms. The Murchison Widefield Array Voltage Capture System (MWA VCS) has been routinely used to study pulsars at low frequencies and discover new pulsars. The MWA VCS offers the unique opportunity of recording complex voltages from all individual antennas (tiles), which can be off-line beamformed or correlated/imaged at millisecond time resolution. Devising imaged-based methods for finding pulsar candidates, which can be verified in beamformed data, can accelerate the complete process and lead to more pulsar detections. Image-based searches for pulsar candidates can reduce the number of tied-array beams required, increasing compute resource efficiency. Despite a factor of $\sim$ 4 loss in sensitivity, searching for pulsar candidates in images from the MWA VCS, we can explore a larger parameter space, potentially leading to discoveries of pulsars missed by high-frequency surveys such as steep spectrum pulsars, exotic binary systems, or pulsars obscured in high-time resolution time series data by propagation effects. Image-based searches are also essential to probing parts of parameter space inaccessible to traditional beamformed searches with the MWA (e.g. at high dispersion measures). In this paper we describe the innovative approach and capability of dual-processing MWA VCS data, that is forming 1-s visibilities and sky images, finding pulsar candidates in these images, and verifying by forming tied-array beam. We developed and tested image-based methods of finding pulsar candidates, which are based on pulsar properties such as steep spectral index, polarisation and variability. The efficiency of these methodologies has been verified on known pulsars, and the main limitations explained in terms of sensitivity and low-frequency spectral turnover of some pulsars. No candidates were confirmed to be a new pulsar, but this new capability will now be applied to a larger subset of observations to accelerate pulsar discoveries with the MWA and potentially speed up future searches with the SKA-Low.

The Southern-sky MWA Rapid Two-metre (SMART) pulsar survey—I. Survey design and processing pipeline

Abstract We present an overview of the Southern-sky MWA Rapid Two-metre (SMART) pulsar survey that exploits the Murchison Widefield Array’s large field of view and voltage-capture system to survey the sky south of 30 $^{\circ}$ in declination for pulsars and fast transients in the 140–170 MHz band. The survey is enabled by the advent of the Phase II MWA’s compact configuration, which offers an enormous efficiency in beam-forming and processing costs, thereby making an all-sky survey of this magnitude tractable with the MWA. Even with the long dwell times employed for the survey (4800 s), data collection can be completed in $<$ 100 h of telescope time, while still retaining the ability to reach a limiting sensitivity of $\sim$ 2–3 mJy (at 150 MHz, near zenith), which is effectively 3–5 times deeper than the previous-generation low-frequency southern-sky pulsar survey, completed in the 1990s. Each observation is processed to generate $\sim$ 5000–8000 tied-array beams that tessellate the full $\sim 610\, {\textrm{deg}^{2}}$ field of view (at 155 MHz), which are then processed to search for pulsars. The voltage-capture recording of the survey also allows a multitude of post hoc processing options including the reprocessing of data for higher time resolution and even exploring image-based techniques for pulsar candidate identification. Due to the substantial computational cost in pulsar searches at low frequencies, the survey data processing is undertaken in multiple passes: in the first pass, a shallow survey is performed, where 10 min of each observation is processed, reaching about one-third of the full-search sensitivity. Here we present the system overview including details of ongoing processing and initial results. Further details including first pulsar discoveries and a census of low-frequency detections are presented in a companion paper. Future plans include deeper searches to reach the full sensitivity and acceleration searches to target binary and millisecond pulsars. Our simulation analysis forecasts $\sim$ 300 new pulsars upon the completion of full processing. The SMART survey will also generate a complete digital record of the low-frequency sky, which will serve as a valuable reference for future pulsar searches planned with the low-frequency Square Kilometre Array.

Pulsar-candidate Selection Using a Generative Adversarial Network and ResNeXt

Pulsar research has been a hot topic in the area of astronomy since they were first discovered. Pulsar discovery is fundamental for pulsar research. While pulsars are now visible across the electromagnetic spectrum, pulsar searches with modern radio telescopes are most promising. As the performance of astronomical instruments improves, the number of pulsar candidates detected by modern radio telescopes grows at an exponential rate. The application of artificial intelligence to the field of pulsar-candidate identification can automatically and efficiently address the identification problem with enormous amounts of data. However, there are still significant challenges in enhancing the accuracy of deep-learning-based pulsar-candidate identification. These problems result primarily from the fact that real pulsar data is scarce: the number of candidates that can be successfully identified as real pulsars (positive samples) is much smaller than those candidates that turn out to not be pulsars but instead radio-frequency interference or noise (negative samples). This makes it difficult to train a machine-learning model that can accurately select those candidates that are real pulsars. Therefore a novel pulsar-candidate identification framework is proposed that combines a deep convolutional generative adversarial neural network (DCGAN) and a deep aggregation residual network (ResNeXt). To overcome sample imbalance, the DCGAN is utilized to generate images that approximate real pulsars, while observed and generated candidates are employed together to train the pulsar-candidate identification model ResNeXt. Experiments on the HTRU Medlat data set back up the framework’s performance. The precision, recall, and F1-score of the framework are 100%.

Searching for Pulsars with Phase Characteristics

We present a method by using the phase characteristics of radio observation data for pulsar search and candidate identification. The phase characteristics are relations between the pulsar signal and the phase correction in the frequency-domain, and we regard it as a new search diagnostic characteristic. Based on the phase characteristics, a search method is presented: calculating dispersion measure (DM)—frequency data to select candidate frequencies, and then confirming of candidates by using the broadband characteristics of pulsar signals. Based on this method, we performed a search test on short observation data of M15 and M71, which were observed by Five-hundred-meter Aperture Spherical radio Telescope, and some of the Galactic Plane Pulsar Snapshot survey data. Results show that it can get similar search results to PRESTO (PulsaR Exploration and Search TOolkit) while having a faster processing speed.

Pushchino multibeam pulsar search – I. Targeted search of weak pulsars

ABSTRACT The search for pulsars in a sample of pulsar candidates found based on a multi-year survey conducted with low (six channels; sampling 0.1s) time-frequency resolution on declinations −9° < δ < +42° was carried out with the Large Phased Array of the Lebedev Physical Institute (LPA LPI). LPA is a transit telescope operating at 111 MHz with a bandwidth of 2.5 MHz. Search, analysis and evidence of pulsar detection were carried out using a visualization programme of summed up power spectra obtained from the survey data with high (32 channels; sampling 12.5 ms) time-frequency resolution. 11 new pulsars with periods P0 = 0.41–3.75 s and dispersion measure DM = 15–154 pc cm−3 have been discovered. In total, in the survey with a low-time-frequency resolution for the period 2016–2021 in a blind search 203 pulsars were found, among them 42 new and 161 known pulsars. It is shown that in the search on the data with high-time-frequency resolution accumulated over a time interval of seven years, pulsars with a flux density of 0.1–0.2 mJy at the frequency of 111 MHz can be detected. When searching for pulsars with regular (periodic) emission at declinations +21° < δ < +42o, all pulsars located outside the galactic plane having P0 ≥ 0.5 s, DM ≤ 100 pc cm−3, and the flux density S ≥ 0.5 mJy can be detected.

Probabilistic learning for pulsar classification

In this work, we explore the possibility of using probabilistic learning to identify pulsar candidates. We make use of Deep Gaussian Process (DGP) and Deep Kernel Learning (DKL). Trained on a balanced training set in order to avoid the effect of class imbalance, the performance of the models, achieving relatively high probability of differentiating the positive class from the negative one (roc-auc ∼ 0.98), is very promising overall. We estimate the predictive entropy of each model predictions and find that DKL is more confident than DGP in its predictions and provides better uncertainty calibration. Upon investigating the effect of training with imbalanced dataset on the models, results show that each model performance decreases with an increasing number of the majority class in the training set. Interestingly, with a number of negative class 10× that of positive class, the models still provide reasonably well calibrated uncertainty, i.e. an expected Uncertainty Calibration Error (UCE) less than 6%. We also show in this study how, in the case of relatively small amount of training dataset, a convolutional neural network based classifier trained via Bayesian Active Learning by Disagreement (BALD) performs. We find that, with an optimized number of training examples, the model — being the most confident in its predictions — generalizes relatively well and produces the best uncertainty calibration which corresponds to UCE = 3.118%.

Pulsar candidate selection with residual convolutional autoencoder

ABSTRACT The pulsar detection survey has contributed to the study of celestial evolution by providing scientists with a large amount of observational data. In addition, the amount of data collected by the survey has grown exponentially, and there is a large class imbalance in the corresponding data. In this paper, we design a residual convolutional autoencoder (RCAE) based on the structure of the autoencoder, and combine with logistic regression (LR) to construct a network structure framework suitable for pulsar candidate identification. RCAE is used as the primary model to fit the data distribution of the non-pulsar sample, the process does not need to consider the positive and negative pulsar sample imbalance. LR is used as an auxiliary classification model to test the final results. The experimental results on the HTRU Medlat and PMPS-26k data sets show that the best performance is achieved without the use of data generation and complex enhancement methods.

A Preliminary Study of Large Scale Pulsar Candidate Sifting Based on Parallel Hybrid Clustering

Pulsar candidate sifting is an essential part of pulsar analysis pipelines for discovering new pulsars. To solve the problem of data mining of a large number of pulsar data using a Five-hundred-meter Aperture Spherical radio Telescope (FAST), a parallel pulsar candidate sifting algorithm based on semi-supervised clustering is proposed, which adopts a hybrid clustering scheme based on density hierarchy and the partition method, combined with a Spark-based parallel model and a sliding window-based partition strategy. Experiments on the two datasets, HTRU (The High Time-Resolution Universe Survey) 2 and AOD-FAST (Actual Observation Data from FAST), show that the algorithm can excellently identify the pulsars with high performance: On HTRU2, the Precision and Recall rates are 0.946 and 0.905, and those on AOD-FAST are 0.787 and 0.994, respectively; the running time on both datasets is also significantly reduced compared with its serial execution mode. It can be concluded that the proposed algorithm provides a feasible idea for astronomical data mining of FAST observation.

Synchro-curvature emitting regions in high-energy pulsar models

ABSTRACT The detected high-energy pulsars’ population is growing in number, and thus, having agile and physically relevant codes to analyse it consistently is important. Here, we update our existing synchro-curvature radiation model by including a better treatment of the particle injection, particularly where the large pitch angle particles dominate the spectra, and by implementing a fast and accurate minimization technique. The latter allows a large improvement in computational cost, needed to test model enhancements, and to apply the model to a larger pulsar population. We successfully fit the sample of pulsars with X-ray and gamma-ray data. Our results indicate that, for every emitting particle, the spatial extent of their trajectory where the pitch angle is large and most of the detected X-ray radiation is produced is a small fraction of the light cylinder. We also confirm with this new approach that synchrotron radiation is not negligible for most of the gamma-ray pulsars detected. In addition, with the results obtained, we argue that J0357+3205 and J2055+2539 are MeV pulsar candidates and are suggested for exhaustive observations in this energy band.

Pulsar Candidate Recognition Using Deep Neural Network Model

With an improvement in the performance of radio telescopes, the number of pulsar candidates has increased rapidly, which makes selecting valuable pulsar signals from the candidates challenging. It is imperative to improve the recognition efficiency of pulsars. Therefore, we solved this problem from the perspective of intelligent image processing and a deep neural network model AR_Net was proposed in this paper. A single time–phase-subgraph or frequency-phase-subgraph was used as the judgment basis in the recognition model. The convolution blocks can be obtained by combining the attention mechanism module, feature extractor and residual connection. Then, different convolution blocks were superimposed to constitute the AR_Net to screen pulsars. The attention mechanism module was used to calculate the weight through an additional feedforward neural network and the important features in the sample were identified by weight, so the ability of the model to learn pivotal information was improved. The feature extractor was used to gain the high-dimensional features in the samples and the residual connection was introduced to alleviate the problem of network degradation and intensify feature reuse. The experimental results show that AR_Net has higher F1-score, recall and accuracy, and our method produces a competitive result compared with previous methods.