Pith. sign in

REVIEW 3 major objections 6 minor 23 references

A two-stage 1D neural network cascade can detect fast radio bursts in four-channel RATAN-600 records, recovering nearly all synthetic typical bursts above a total SNR of 4 while keeping false positives to about two per day.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-04 16:52 UTC pith:XEKSIUOB

load-bearing objection Honest, reproducible methods paper; strong in-distribution results but the practical FRB search claim still rests on unvalidated synthetic-to-real transfer. the 3 major comments →

arxiv 2509.11215 v1 pith:XEKSIUOB submitted 2025-09-14 astro-ph.IM astro-ph.HE

Neural networks in the search for fast radio bursts with RATAN-600

D.O. Kudryavtsev , S.A. Trushkin , P.G. Tsybulev , V.A. Stolyarov This is my paper
classification astro-ph.IM astro-ph.HE
keywords fast radio burstsconvolutional neural networks1D time series classificationRATAN-600radio transient searchsynthetic FRB injectionfalse positive suppressiondispersion measure
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper aims to show that fast radio bursts can be found in radio observations that have only a handful of frequency channels, a regime where classical dedispersion loses the signal. It develops a compact 1D convolutional network, EfficientNet1d-XS, that reads four time series directly and classifies each one-second stretch as containing an FRB or not. To make the search practical, the authors add a second network trained only on the first network's mistakes, cutting the false-positive rate from thousands per day to roughly two. On synthetic bursts injected into real noise, the cascade recovers nearly all 'typical' FRBs (5 ms width, DM 500) with total SNR above 4, and it successfully flags a giant pulse from the Crab Pulsar. The method matters because few-channel broadband radiometers like RATAN-600's Western Sector are otherwise blind to FRBs.

Core claim

The central claim is that frequency–time information from just four broad radio channels is sufficient, after dispersion smearing, to recognize an FRB if the classifier is a 1D convolutional network trained on realistic synthetic pulses embedded in real radiometer noise and RFI. With a cascade of two such networks, the false-positive rate becomes the product of the stages' individual rates, about 5e-6 at the medium threshold, corresponding to roughly two false events per day across four radiometers. The paper reports detection rates near unity for synthetic FRBs with w=5 ms and DM=500 pc cm^-3 once total SNR exceeds 4, and a working detection of the Crab giant pulse with DM about 57 and an e

What carries the argument

EfficientNet1d-XS, a 1D adaptation of the EfficientNetV2 convolutional architecture in which 2D convolutions are replaced with 1D ones, the network is scaled down to about 1.13 million parameters, and MBConv blocks carry Squeeze-and-Excitation attention, GELU activations, stochastic depth, and residual connections. It processes 1-second, 4-channel time series (4080 samples each) and outputs a single logit. A second identical network is trained only on the first network's false positives, and the two form a cascade whose false-positive rate is the product of the two stages' rates; probability thresholds on the second stage trade off detection rate against false alarms.

Load-bearing premise

The training set assumes real FRBs resemble the synthetic Gaussian-profile pulses with the listed parameter distributions; if actual bursts have complex non-Gaussian time structures, the reported detection rates will not transfer to real data.

What would settle it

Measure the cascade's recall on synthetic bursts whose time profiles are non-Gaussian (e.g., two-component or asymmetric) injected into the same real noise records; a substantial drop below the Gaussian-curve recall would falsify the transferability claim. Alternatively, run the cascade on a year of real RATAN-600 archive data and compare the number of visually confirmed FRB candidates with the rate predicted from the synthetic recall curves.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • RATAN-600's archived Western Sector data (since 2017) can be searched for FRBs even though the telescope has only four 150-MHz channels.
  • Small radio telescopes with few-channel broadband backends gain a viable FRB search route without large filterbank spectrometers.
  • The cascade scheme can be reused for any classifier with a high raw false-positive rate, since training stage 2 on stage-1 mistakes multiplies the rates.
  • Detection efficiency depends on pulse width and DM: narrow, low-DM pulses (like giant pulses) are recovered only at high SNR, so follow-up searches for such events should use the low-probability threshold.
  • Adding simple coincidence filters (detection in two half-interval scans, and rejection of pulses seen in multiple beams) brings yearly false positives down to about a hundred, making manual review feasible.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the synthetic-to-real transfer holds, the same 1D cascade approach could be applied to other few-channel radiometers (e.g., total-power monitors with a handful of bands) with only re-training on their noise and RFI.
  • The paper's dependence on a Gaussian intrinsic profile is the main transfer risk; testing the cascade on non-Gaussian simulated bursts (e.g., double-peaked or multi-component profiles) would quantify how much real recall drops.
  • One could also use the network's own confidence distribution to scan a large archive and then visually inspect only the high-probability candidates initially; the paper's manual step could later be replaced by a second-stage candidate filter such as a DM–time fit on re-examined data.
  • The reported ~2/day false-positive rate assumes stationary RFI statistics; a seasonal or year-long RFI catalog would likely increase that number, so archival searches may need periodic retraining on newly collected RFI.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper presents a 1D convolutional neural network (EfficientNet1d-XS) for detecting fast radio bursts in RATAN-600 four-channel, 4.7 GHz time-series data, where classical dedispersion with only four broad channels is ineffective. Training data comprise real radiometer noise, real RFI, and synthetic FRBs generated with a Gaussian intrinsic profile plus scattering, dispersion smearing, spectral indexing, and scintillation. A two-expert cascade is used to suppress false positives. Evaluation reports false-positive rates at several thresholds and recall as a function of total SNR for selected width/DM combinations, with bootstrap confidence intervals, plus a successful detection of a Crab giant pulse. The main quantitative claims are that for 'typical' synthetic FRBs (w = 5 ms, DM = 500 pc cm^-3) the cascade detects nearly all events with total SNR > 4, while the medium threshold yields roughly 2 false positives per day.

Significance. If the reported performance transfers to real observations, the work would enable blind FRB searches with broadband radiometers having very few frequency channels, an instrumental regime that is increasingly relevant and for which standard imaging-based deep-learning pipelines are unsuited. The paper is commendably concrete: the code is released, the architecture and training are described in detail, bootstrap uncertainties are provided, real noise and RFI records are used, and the Crab-pulse test checks the full preprocessing and inference chain. The central limitation is explicit in the manuscript itself: the quantitative detection rates are measured on synthetic events drawn from the same generator used for training, and no real FRB has yet been detected with this instrument. The result is therefore best understood as an internally consistent proof of concept under a stated signal model, not as an externally validated detection pipeline.

major comments (3)
  1. [§5.3, §7, Table 4] The headline recall claim ('nearly all events with SNR > 4 ' for w = 5 ms, DM = 500) is an in-distribution result: the test events are generated by the same recipe (Gaussian profile, scattering, dispersion smearing, spectral index, scintillation) used to build the training sample. This does not test generalization to real FRB morphologies, which the manuscript itself acknowledges can be non-Gaussian and complex. To make the practical claim load-bearing, I ask for a robustness experiment: inject synthetic FRBs with alternative morphologies (e.g., sub-burst structure, frequency-dependent drift, asymmetric profiles) into real noise records and report recall and FP rates, or explicitly restrict the abstract/conclusion claims to 'under the Table 4 model.' Without one of these, the phrase 'we expect to detect nearly all events' overstates what has been demonstrated.
  2. [§5.2, Table 6] The cascade FP rate is computed as the product of the Expert 1 and Expert 2 FP rates. This identity is only valid if Expert 2's quoted FP rate is the conditional probability P(FP | Expert 1 positive) on the actual candidate population, or if the two classifiers' errors are independent on non-FRB data. The manuscript does not specify the test population used for Expert 2's FP rate. Please state this explicitly, measure the cascade end-to-end on a continuous test stream, and report a confidence interval on the resulting per-day rate. This is directly load-bearing for the '2 events per day' claim.
  3. [Table 6, high-threshold row] The high-threshold row reports a cascade FP rate of '~0' because no false positives appeared in the test sample. A zero count provides an upper bound, not a point estimate; with a finite test sample the 95% confidence upper limit is nonzero and can correspond to a non-negligible number of events per day when extrapolated to 345,600 daily instances. Please report a Clopper-Pearson or Poisson upper bound for this row, or state the effective number of independent test instances used.
minor comments (6)
  1. [§3.2.4, Eq. (3)] The notation uses s_nu(t) on both sides of the amplitude-calibration equation, which is self-referential and confusing. Use a normalized profile symbol (e.g., p_nu(t)) for the convolved shape before rescaling.
  2. [Table 4] Typo in the header: 'Distibution' should be 'Distribution'. Also, the scattering distribution range 'Uniform [0,1]' at tau_1GHz is not obvious; a brief justification of the physical range would help.
  3. [§5.3, Fig. 7] The figure caption says 'panels (left to right) correspond to pulses with certain characteristics,' but the three (w, DM) combinations are only given in the text. Add the parameters directly to the panel titles or legend for readability.
  4. [§4.2, Table 5] The terms 'Time flip' and 'Channel permutation' in Table 5 are not defined until the surrounding text; a one-line definition in the caption would improve clarity.
  5. [§2.2] The statement that width and depth were scaled 'proportionally to the minimum' is vague. Give the exact scaling factors relative to EfficientNetV2-S, or state which blocks were reduced by how much.
  6. [References] The citation 'Snell et al. 2019' refers to a textbook by three authors, which is fine, but the reference entry appears to list only the title; add the publisher and page/chapter information.

Circularity Check

0 steps flagged

No significant circularity: evaluation is in-distribution synthetic testing, and the paper explicitly conditions its claims on synthetic FRBs.

full rationale

The paper's quantitative claims are honest in-distribution measurements. The network is trained on synthetic FRBs generated by the model of Sec. 3.2 and evaluated on independently drawn synthetic events from the same Table 4 distributions; held-out test performance is not guaranteed by construction, so the reported recall has empirical content. The cascade FP calculation (0.0065 × 0.0008) is simply the definition of the two-stage conditional false-positive rate, not a hidden re-use of the target result. Thresholds are disclosed in Table 6 and are set on the test sample, which is a standard (if somewhat optimistic) evaluation choice rather than a circular reduction. The paper repeatedly and explicitly limits its claim: 'Synthetic FRBs may not fully replicate real FRB characteristics, potentially affecting real-world performance' and 'Currently there are no real FRBs to be presented and validate the performance on the example of real events.' The only real-event check is the Crab giant pulse, which the authors themselves note is not FRB-like. No load-bearing self-citation, uniqueness import, or renamed empirical result occurs; the external-validity gap is a limitation, not circularity.

Axiom & Free-Parameter Ledger

3 free parameters · 4 axioms · 0 invented entities

The central claim rests on the synthetic FRB model used for training and evaluation, plus operational thresholds chosen from test data. There are no fitted physical constants; the free parameters are ML thresholds and hand-chosen simulation distributions. No invented physical entities are proposed.

free parameters (3)
  • Expert 2 logit thresholds = 0, -4.1, +4.1 (medium, low, high)
    Thresholds selected from test-sample FP rates to trade recall against false positives; the reported cascade FP rates (about 20, 2, and ~0 per day) are direct functions of these chosen values (Table 6, Section 5.2).
  • Synthetic FRB parameter distributions = A_int log-normal (mu=3, sigma=1); w_int uniform [0.02,50] ms; DM uniform [30,5000] pc cm^-3; etc.
    Chosen by hand to span plausible FRB parameters; recall curves in Section 5.3 are measured on events drawn from these distributions, so the quoted detection rates depend on this choice.
  • Network hyperparameters = dropout/stochastic depth p=0.2, SE ratio 1/16, XS width/depth scaling
    Not tuned (stated in Section 7); they affect the reported FP and recall numbers, but no ablation shows robustness to these choices.
axioms (4)
  • standard math Dispersion delay formula (Eq. 1) and DM-time behavior of FRBs at 4.7 GHz.
    Standard radio astronomy formula used for synthetic signal generation; assumed valid for the pulse model.
  • ad hoc to paper Synthetic FRBs with Gaussian intrinsic profile, exponential scattering, dispersion smearing, spectral index, and scintillation are representative of real FRB signals at RATAN-600.
    This is the load-bearing modeling assumption. The training and test sets are built from this generator; if real FRBs have complex non-Gaussian structure, the recall estimates do not transfer (acknowledged in Section 7).
  • domain assumption RFI records collected during one month (11 pm to 6 am) represent the variety of RFI in the multi-year archive.
    Section 3.1 and Section 7: novel RFI types not in training may raise the FP rate.
  • domain assumption Expert 2 FP rates measured on expert 1's false positives combine with expert 1's FP rate as a product.
    Section 4.2 and Table 6: valid if expert 2's FP rate is conditional on expert 1 FPs, but small-sample variance and threshold dependence remain.

pith-pipeline@v1.3.0-alltime-deepseek · 14358 in / 12828 out tokens · 147210 ms · 2026-08-04T16:52:58.127276+00:00 · methodology

0 comments
read the original abstract

We present a technique to search for fast radio bursts in records obtained with broadband radiometers having few radio channels. The technique is applied to the RATAN-600 surveys carried out at its Western Sector since the year 2017. A 1D convolutional neural network for multichannel time series classification is developed based on the EfficientNet family of models. The procedure to generate synthetic FRB signals needed for the training dataset is described. We implement a two-stage cascade scheme to effectively suppress the rate of false positive detections. Evaluation of the trained model is provided based on the synthetic events and the giant pulse of the Crab Pulsar.

Figures

Figures reproduced from arXiv: 2509.11215 by D.O. Kudryavtsev, P.G. Tsybulev, S.A. Trushkin, V.A. Stolyarov.

Figure 1
Figure 1. Figure 1: A synthetic FRB signal (w = 5 ms, DM = 500 cm−3 pc) in the time–frequency domain for different frequency ranges: L-band (1–2 GHz, top panel) and RATAN-600 radiometers (4.4–5 GHz, bot￾tom panel). The horizontal red lines designate the boundaries of the RATAN-600 four subchannels [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Comparison of synthetic DM–time diagrams in the case [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: MBConv and Fused MBConv blocks. The transformations [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: An example of a synthetic FRB event: Aint = 3 σnoise, DM = 500 cm−3 pc, wint = 5 ms. The four lower panels demonstrate the records in the four radiometer channels, in the top panel the records are combined. The nulling effect is seen in the second channel. but is trained not on the all variety of RFI events but only on those that are FP detections from expert 1. The results is that the overall FP rate is r… view at source ↗
Figure 5
Figure 5. Figure 5: Variation of the learning rate at the beginning of each [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The training and validation loss curves for experts 1 (left) and 2 (right). The validation loss is estimated at the end of an epoch. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Detection rates in the cascade scheme depending on the total SNR. The panels (left to right) correspond to pulses with certain [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: A giant pulse from the Crab Pulsar observed on 30 June [PITH_FULL_IMAGE:figures/full_fig_p010_8.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

23 extracted references · 7 linked inside Pith

  1. [1]

    , author Aggarwal , K

    author Agarwal , D. , author Aggarwal , K. , author Burke-Spolaor , S. , author Lorimer , D.R. , author Garver-Daniels , N. , year 2020 a. title FETCH: A deep-learning based classifier for fast transient classification . journal Monthly Notices of the Royal Astronomical Society volume 497 , pages 1661--1674 . :10.1093/mnras/staa1856, http://arxiv.org/abs/...

    Pith/arXiv arXiv 2020

  2. [2]

    , author Lorimer , D.R

    author Agarwal , D. , author Lorimer , D.R. , author Surnis , M.P. , author Pei , X. , author Karastergiou , A. , author Golpayegani , G. , author Werthimer , D. , author Cobb , J. , author McLaughlin , M.A. , author White , S. , author Armour , W. , author MacMahon , D.H.E. , author Siemion , A.P.V. , author Foster , G. , year 2020 b. title Initial resul...

    Pith/arXiv arXiv 2020

  3. [3]

    , author van Leeuwen , J

    author Connor , L. , author van Leeuwen , J. , year 2018 . title Applying Deep Learning to Fast Radio Burst Classification . journal Astronomical Journal volume 156 , pages 256 . :10.3847/1538-3881/aae649, http://arxiv.org/abs/1803.03084 arXiv:1803.03084

    Pith/arXiv arXiv 2018

  4. [4]

    , author Borovec , J

    author Falcon , W. , author Borovec , J. , author W \"a lchli , A. , author Eggert , N. , author Schock , J. , author Jordan , J. , author Skafte , N. , author Ir1dXD , author Bereznyuk , V. , author Harris , E. , author Murrell , T. , author Yu , P. , author Pr sius , S. , author Addair , T. , author Zhong , J. , author Lipin , D. , author Uchida , S. , ...

    2020

  5. [5]

    , author Forestier, G

    author Fawaz, H.I. , author Forestier, G. , author Weber, J. , author Idoumghar, L. , author Muller, P.A. , year 2018 . title Deep learning for time series classification: a review . journal Data Mining and Knowledge Discovery volume 33 , pages 917 -- 963 . https://api.semanticscholar.org/CorpusID:52195012

    2018

  6. [6]

    , author Zhang, X

    author He, K. , author Zhang, X. , author Ren, S. , author Sun, J. , year 2016 . title Deep residual learning for image recognition , in: booktitle 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pp. pages 770--778 . :10.1109/CVPR.2016.90

    doi:10.1109/cvpr.2016.90 2016

  7. [7]

    , author Gimpel, K

    author Hendrycks, D. , author Gimpel, K. , year 2023 . title Gaussian error linear units (gelus) . http://arxiv.org/abs/1606.08415 arXiv:1606.08415

    Pith/arXiv arXiv 2023

  8. [8]

    , author Shen, L

    author Hu, J. , author Shen, L. , author Sun, G. , year 2018 . title Squeeze-and-excitation networks , in: booktitle 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp. pages 7132--7141 . :10.1109/CVPR.2018.00745

    arXiv 2018

  9. [9]

    , author Sun, Y

    author Huang, G. , author Sun, Y. , author Liu, Z. , author Sedra, D. , author Weinberger, K.Q. , year 2016 . title Deep networks with stochastic depth , in: editor Leibe, B. , editor Matas, J. , editor Sebe, N. , editor Welling, M. (Eds.), booktitle Computer Vision -- ECCV 2016 , publisher Springer International Publishing , address Cham . pp. pages 646--661

    2016

  10. [10]

    , author Szegedy, C

    author Ioffe, S. , author Szegedy, C. , year 2015 . title Batch normalization: accelerating deep network training by reducing internal covariate shift , in: editor Francis, B. , editor Blei, D. (Eds.), booktitle Proceedings of the 32nd International Conference on Machine Learning , publisher PMLR . pp. pages 448--456

    2015

  11. [11]

    , author Avci, O

    author Kiranyaz, S. , author Avci, O. , author Abdeljaber, O. , author Ince, T. , author Gabbouj, M. , author Inman, D.J. , year 2021 . title 1d convolutional neural networks and applications: A survey . journal Mechanical Systems and Signal Processing volume 151 , pages 107398 . https://www.sciencedirect.com/science/article/pii/S0888327020307846, :https:...

    arXiv 2021

  12. [12]

    , author Sutskever, I

    author Krizhevsky, A. , author Sutskever, I. , author Hinton, G.E. , year 2012 . title Imagenet classification with deep convolutional neural networks , in: editor Pereira, F. , editor Burges, C. , editor Bottou, L. , editor Weinberger, K. (Eds.), booktitle Advances in Neural Information Processing Systems , publisher Curran Associates, Inc. https://proce...

    2012

  13. [13]

    , author Chen, Q

    author Lin, M. , author Chen, Q. , author Yan, S. , year 2014 . title Network in network . http://arxiv.org/abs/1312.4400 arXiv:1312.4400

    Pith/arXiv arXiv 2014

  14. [14]

    , author Shelhamer, E

    author Long, J. , author Shelhamer, E. , author Darrell, T. , year 2015 . title Fully convolutional networks for semantic segmentation , in: booktitle 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pp. pages 3431--3440 . :10.1109/CVPR.2015.7298965

    arXiv 2015

  15. [15]

    , author Gross, S

    author Paszke, A. , author Gross, S. , author Massa, F. , author Lerer, A. , author Bradbury, J. , author Chanan, G. , author Killeen, T. , author Lin, Z. , author Gimelshein, N. , author Antiga, L. , author Desmaison, A. , author Kopf, A. , author Yang, E. , author DeVito, Z. , author Raison, M. , author Tejani, A. , author Chilamkurthy, S. , author Stei...

    2019

  16. [16]

    , author Smith , A.J

    author Perera , B.B.P. , author Smith , A.J. , author Vaddi , S. , author Carballo-Rubio , R. , author McGilvray , A. , author Venkataraman , A. , author Roshi , D.A. , author Manoharan , P.K. , author Perillat , P. , author Lieb , E. , author Lorimer , D.R. , author McLaughlin , M.A. , author Agarwal , D. , author Aggarwal , K. , author Ransom , S.M. , y...

    Pith/arXiv arXiv 2022

  17. [17]

    , author Howard, A

    author Sandler, M. , author Howard, A. , author Zhu, M. , author Zhmoginov, A. , author Chen, L.C. , year 2018 . title Mobilenetv2: Inverted residuals and linear bottlenecks , in: booktitle 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp. pages 4510--4520 . :10.1109/CVPR.2018.00474

    arXiv 2018

  18. [18]

    , author Kurtz , S.E

    author Snell , R.L. , author Kurtz , S.E. , author Marr , J.M. , year 2019 . title Fundamentals of Radio Astronomy (Series in Astronomy and Astrophysics)

    2019

  19. [19]

    , author Hinton, G

    author Srivastava, N. , author Hinton, G. , author Krizhevsky, A. , author Sutskever, I. , author Salakhutdinov, R. , year 2014 . title Dropout: A simple way to prevent neural networks from overfitting . journal Journal of Machine Learning Research volume 15 , pages 1929--1958 . http://jmlr.org/papers/v15/srivastava14a.html

    2014

  20. [20]

    , author Le, Q

    author Tan, M. , author Le, Q. , year 2019 . title E fficient N et: Rethinking model scaling for convolutional neural networks , in: editor Chaudhuri, K. , editor Salakhutdinov, R. (Eds.), booktitle Proceedings of the 36th International Conference on Machine Learning , publisher PMLR . pp. pages 6105--6114 . https://proceedings.mlr.press/v97/tan19a.html

    2019

  21. [21]

    , author Le, Q

    author Tan, M. , author Le, Q. , year 2021 . title Efficientnetv2: Smaller models and faster training , in: editor Meila, M. , editor Zhang, T. (Eds.), booktitle Proceedings of the 38th International Conference on Machine Learning , publisher PMLR . pp. pages 10096--10106 . https://proceedings.mlr.press/v139/tan21a.html

    2021

  22. [22]

    , author Yan, W

    author Wang, Z. , author Yan, W. , author Oates, T. , year 2017 . title Time series classification from scratch with deep neural networks: A strong baseline , in: booktitle 2017 International Joint Conference on Neural Networks (IJCNN) , pp. pages 1578--1585 . :10.1109/IJCNN.2017.7966039

    arXiv 2017

  23. [23]

    , author Gajjar , V

    author Zhang , Y.G. , author Gajjar , V. , author Foster , G. , author Siemion , A. , author Cordes , J. , author Law , C. , author Wang , Y. , year 2018 . title Fast Radio Burst 121102 Pulse Detection and Periodicity: A Machine Learning Approach . journal Astrophysical Journal volume 866 , pages 149 . :10.3847/1538-4357/aadf31, http://arxiv.org/abs/1809....

    Pith/arXiv arXiv 2018