pith. sign in

arxiv: 2508.02750 · v2 · submitted 2025-08-03 · 💻 cs.LG · cs.AI· nucl-ex· physics.app-ph· physics.atom-ph

Pulse Shape Discrimination Algorithms: Survey and Benchmark

Haoran Liu , Yihan Zhan , Mingzhe Liu , Yanhua Liu , Peng Li , Zhuo Zuo , Bingqi Liu , Runxi Liu This is my paper

Pith reviewed 2026-05-19 01:19 UTC · model grok-4.3

classification 💻 cs.LG cs.AInucl-exphysics.app-phphysics.atom-ph
keywords pulse shape discriminationradiation detectiondeep learningmachine learningbenchmarkneural networksfigure of merit
0
0 comments X

The pith

Deep learning models, especially MLPs and hybrids, outperform traditional methods in pulse shape discrimination for radiation detection.

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

The paper conducts a survey and benchmark of nearly sixty pulse shape discrimination algorithms for radiation detection. It groups the methods into statistical techniques based on time or frequency domains and prior-knowledge techniques that include machine learning and deep learning. All algorithms are implemented and run on two fixed datasets, one unlabeled from an americium-beryllium source and one time-of-flight labeled from a plutonium-beryllium source. Results measured by Figure of Merit, F1-score, and ROC-AUC indicate that deep learning models and hybrids often deliver better separation than conventional statistical approaches. The work also supplies an open toolbox and the datasets to support further comparisons.

Core claim

After implementing and testing nearly sixty PSD algorithms on the two standardized datasets, the analysis shows that deep learning models, particularly Multi-Layer Perceptrons and hybrid approaches that combine statistical features with neural regression, often achieve higher performance than traditional methods according to the chosen metrics.

What carries the argument

A uniform benchmarking pipeline that applies all algorithms to two fixed radiation datasets and scores them with Figure of Merit, F1-score, ROC-AUC, and inter-method correlations.

If this is right

  • Deep learning and hybrid methods may maintain advantages at lower energy thresholds where traditional methods degrade.
  • Hybrid statistical-neural approaches can combine explicit domain features with learned representations for improved results.
  • Reliance on a single metric such as FOM has limitations, so multiple metrics give a more complete picture.
  • The released toolbox and datasets allow direct reproduction and extension of the comparisons.

Where Pith is reading between the lines

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

  • Top-performing models could improve real-time neutron-gamma separation in portable or security-related detectors.
  • Correlations among methods point to possible ensemble techniques that combine several algorithms.
  • Further tests on varied energy regimes or detector materials would clarify how broadly the ranking holds.

Load-bearing premise

That fair, equivalent implementations of all algorithms were achieved and that the two chosen datasets plus the chosen metrics adequately capture real-world PSD performance across varying hardware and energy regimes.

What would settle it

A new dataset from different detector hardware or a different neutron source where multiple traditional statistical methods score higher than the top deep learning models on the same set of metrics.

Figures

Figures reproduced from arXiv: 2508.02750 by Bingqi Liu, Haoran Liu, Mingzhe Liu, Peng Li, Runxi Liu, Yanhua Liu, Yihan Zhan, Zhuo Zuo.

Figure 1
Figure 1. Figure 1: Example pulse signals shapes and can subsequently classify individual incoming pulses, often in real-time. Within this paradigm, one can distinguish between classical machine learning methods, such as the Support Vector Machine [19], K-Nearest Neighbors [20], and Tempotron [9], and more recent deep learning models, including Multi-Layer Perceptrons [21], Convolutional Neural Networks [22], and the Transfor… view at source ↗
Figure 2
Figure 2. Figure 2: PSD Factor Histograms with Gaussian Fits for Selected High- and Low-Performing Methods on the 241Am9 Be Dataset regress on PSD factors from the ZC method) [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: PSD Factor Histograms with Gaussian Fits for Selected High- and Low-Performing Methods on the 238Pu9 Be Dataset respectively, reflecting consistent performance. In contrast, GP and PR lag with F1-scores of 0.637 and 0.621 and FOMs of 0.5580 and Failed, indicating weaker discriminative feature capture. FD methods also exhibit variability: DFT and FGA achieve F1-scores of 0.931–0.932 but differ in FOMs (0.54… view at source ↗
Figure 4
Figure 4. Figure 4: ROC Curves for (a) Best- and (b) Worst-Performing DL and (c) Best-performing ML Classifiers on the 238Pu9 Be Dataset [PITH_FULL_IMAGE:figures/full_fig_p016_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Correlation Heatmap of PSD Factors from High- and Low-Performing Methods on the 238Pu9 Be Dataset (best: ENN, worst: RNN), MLP (best: MLP2STFT, worst: MLP2WT), and advanced networks (best: TRAN, worst: MAM). Additionally, it includes curves for the four best￾performing ML classifiers: BDT, GMM, KNN, and LRSTFT. This figure illustrates the trade-offs between sensitivity and specificity, highlighting how top… view at source ↗
Figure 6
Figure 6. Figure 6: Division of the 238Pu9 Be Dataset into Software Energy Thresholds: (a) 2 MeV, (b) 3 MeV, (c) 4 MeV, and (d) 5 MeV calculated the Pearson correlation coefficients between their PSD factors on the 238Pu9 Be dataset. Based on [PITH_FULL_IMAGE:figures/full_fig_p017_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Radar Chart of PSD Method Performance Across Different Energy Thresholds performing methods for each prior-knowledge-based modality. In general, as the software energy threshold increases from 2 MeV to 4 MeV, the performance of most PSD methods improves, with metrics often rising by 0.01 to 0.03 points, reflecting better discrimination capabilities at higher energies where signal characteristics are more d… view at source ↗
read the original abstract

This review presents a comprehensive survey and benchmark of pulse shape discrimination (PSD) algorithms for radiation detection, classifying nearly sixty methods into statistical (time-domain, frequency-domain, neural network-based) and prior-knowledge (machine learning, deep learning) paradigms. We implement and evaluate all algorithms on two standardized datasets: an unlabeled set from a 241Am-9Be source and a time-of-flight labeled set from a 238Pu-9Be source, using metrics including Figure of Merit (FOM), F1-score, ROC-AUC, and inter-method correlations. Our analysis reveals that deep learning models, particularly Multi-Layer Perceptrons (MLPs) and hybrid approaches combining statistical features with neural regression, often outperform traditional methods. We discuss architectural suitabilities, the limitations of FOM, alternative evaluation metrics, and performance across energy thresholds. Accompanying this work, we release an open-source toolbox in Python and MATLAB, along with the datasets, to promote reproducibility and advance PSD research.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. The paper surveys nearly sixty pulse shape discrimination (PSD) algorithms for radiation detection, classifying them into statistical (time-domain, frequency-domain, neural network-based) and prior-knowledge (machine learning, deep learning) paradigms. It implements and evaluates all algorithms on two standardized datasets—an unlabeled set from a 241Am-9Be source and a time-of-flight labeled set from a 238Pu-9Be source—using metrics including Figure of Merit (FOM), F1-score, ROC-AUC, and inter-method correlations. The central claim is that deep learning models, particularly Multi-Layer Perceptrons (MLPs) and hybrid approaches combining statistical features with neural regression, often outperform traditional methods. The work discusses architectural suitabilities, limitations of FOM, performance across energy thresholds, and releases an open-source Python/MATLAB toolbox with the datasets.

Significance. If the benchmark implementations prove fair and representative across hardware and energy regimes, the manuscript would offer a valuable standardized comparison of PSD methods and a reusable toolbox that promotes reproducibility in radiation detection research. The explicit discussion of FOM limitations and use of multiple metrics (F1, ROC-AUC) alongside two datasets strengthens the empirical contribution relative to prior single-metric surveys.

major comments (2)
  1. [§4 (Benchmark Results) and §3.2 (Implementation Details)] §4 (Benchmark Results) and §3.2 (Implementation Details): The central claim that MLPs and hybrid statistical+neural models often outperform the ~60 traditional methods (e.g., charge comparison, zero-crossing, frequency-domain filters) in FOM/F1/ROC-AUC rests on the assumption of equivalent optimization effort. Traditional methods have multiple dataset- and hardware-specific tunable parameters (integration windows, thresholds, filter coefficients) whose exhaustive search is feasible but not automatically equivalent to architecture and learning-rate sweeps for neural models; without explicit documentation of search spaces, grid/random search details, or validation procedures applied uniformly, the reported performance gaps could be artifacts of unequal tuning rather than intrinsic superiority.
  2. [Table 2 or equivalent results table (energy-thresholded rows)] Table 2 or equivalent results table (energy-thresholded rows): The FOM values for traditional methods appear lower than for MLPs, but if post-hoc energy thresholds or data-cleaning choices were applied differently across algorithm classes, this would undermine the cross-method and cross-energy-regime comparisons that support the outperformance conclusion.
minor comments (3)
  1. [§2] The classification of methods into 'statistical' vs. 'prior-knowledge' paradigms could be clarified with a explicit decision tree or table in §2, as some neural-network-based statistical methods blur the boundary.
  2. [Figures in §4] Figure captions for correlation matrices or ROC curves should explicitly state the number of runs or seeds used to generate error bars, if any.
  3. [Introduction] A few references to prior PSD surveys appear to be missing from the introduction; adding them would better situate the novelty of the ~60-algorithm scope.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive report. The comments raise important points about benchmarking fairness that we address point by point below. We have revised the manuscript accordingly to improve clarity and documentation.

read point-by-point responses

  1. Referee: [§4 (Benchmark Results) and §3.2 (Implementation Details)] §4 (Benchmark Results) and §3.2 (Implementation Details): The central claim that MLPs and hybrid statistical+neural models often outperform the ~60 traditional methods (e.g., charge comparison, zero-crossing, frequency-domain filters) in FOM/F1/ROC-AUC rests on the assumption of equivalent optimization effort. Traditional methods have multiple dataset- and hardware-specific tunable parameters (integration windows, thresholds, filter coefficients) whose exhaustive search is feasible but not automatically equivalent to architecture and learning-rate sweeps for neural models; without explicit documentation of search spaces, grid/random search details, or validation procedures applied uniformly, the reported performance gaps could be artifacts of unequal tuning rather than intrinsic superiority.

    Authors: We appreciate this observation on ensuring comparable optimization. In the original implementation, traditional methods were tuned via grid search over their principal parameters (integration windows, thresholds, and filter coefficients) on the same validation splits used for neural models, with the resulting best configurations reported in the benchmark. However, we agree that the description in §3.2 could be more explicit. In the revised manuscript we will expand §3.2 with a new paragraph (and an accompanying table in the supplement) that lists the exact search ranges, grid resolutions, and validation procedure applied uniformly to every algorithm class. This addition will make the equivalence of tuning effort transparent without altering the reported results. revision: yes

  2. Referee: [Table 2 or equivalent results table (energy-thresholded rows)] Table 2 or equivalent results table (energy-thresholded rows): The FOM values for traditional methods appear lower than for MLPs, but if post-hoc energy thresholds or data-cleaning choices were applied differently across algorithm classes, this would undermine the cross-method and cross-energy-regime comparisons that support the outperformance conclusion.

    Authors: We confirm that all methods were evaluated under identical preprocessing. The same energy thresholds, pulse-selection criteria, and data-cleaning steps were applied to every algorithm before computing the metrics shown in Table 2; this is stated in §4 but can be made more prominent. In the revision we will add an explicit sentence to the Table 2 caption and to the opening paragraph of §4 stating that “preprocessing, energy thresholding, and data-cleaning steps were performed identically for all algorithms to guarantee fair cross-method comparison.” No changes to the numerical results are required. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical benchmark on external datasets

full rationale

The paper is a survey and benchmark that implements ~60 PSD algorithms and evaluates them directly on two external radiation datasets (241Am-9Be unlabeled and 238Pu-9Be time-of-flight labeled) using standard metrics (FOM, F1, ROC-AUC). The central claim that MLPs and hybrid statistical+neural methods often outperform traditional methods is presented as the outcome of these runs, with no derivation chain, equations, fitted parameters, or self-citations invoked to produce the result. The work is self-contained against external benchmarks and releases code/datasets for reproducibility; no load-bearing step reduces to an internal input by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claims rest on the representativeness of the two radiation datasets and on the assumption that re-implementations of prior algorithms were performed without systematic bias favoring any class of method.

axioms (2)
  • domain assumption The unlabeled 241Am-9Be and time-of-flight labeled 238Pu-9Be datasets are appropriate standardized references for comparing PSD algorithms across energy ranges.
    All algorithms are evaluated on these two collections; any mismatch with real detector conditions would affect the reported performance ordering.
  • domain assumption Standard ML metrics (FOM, F1-score, ROC-AUC) and inter-method correlation provide sufficient and unbiased assessment of discrimination quality.
    The paper itself notes limitations of FOM, yet still relies on these metrics for the headline comparisons.

pith-pipeline@v0.9.0 · 5730 in / 1544 out tokens · 40046 ms · 2026-05-19T01:19:28.499716+00:00 · methodology

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

65 extracted references · 65 canonical work pages · 1 internal anchor

  1. [1]

    In contrast, GP and PR lag with F1 - scores of 0.637 and 0.621 and FOMs of 0.5580 and Failed, indicating weaker discriminative feature capture

    PSD Factor Histograms with Gaussian Fits for Selected High - and Low - Performing Methods on the 2 38 Pu 9 Be Dataset respectively, reflecting consistent performance. In contrast, GP and PR lag with F1 - scores of 0.637 and 0.621 and FOMs of 0.5580 and Failed, indicating weaker discriminative feature capture. FD methods also exhibit variability: DFT and F...

    work page 2005

  2. [3]

    Non - linearity assessment of neutron detection systems using zero - power reactor transients,

    T. Bily and L. Keltnerova, "Non - linearity assessment of neutron detection systems using zero - power reactor transients," Appl. Radiat. Isot., vol. 157, p. 109016, 2020, doi: 10.1016/j.apradiso.2019.109016

    work page doi:10.1016/j.apradiso.2019.109016 2020

  3. [4]

    Delayed Neutron Detection with graphite moderator for clad failure detection in Sodium - Cooled Fast Reactors,

    E. Rohée et al. , "Delayed Neutron Detection with graphite moderator for clad failure detection in Sodium - Cooled Fast Reactors," Ann. Nucl. Energy, vol. 92, pp. 440 – 446, 2016, doi: 10.1016/j.anucene.2016.02.003

    work page doi:10.1016/j.anucene.2016.02.003 2016

  4. [5]

    Calculation of Production Reaction Cross Section of Some Radiopharmaceuticals Used in Nuclear Medicine by New Density Dependent Parameters,

    Y. Kavun, T. Eyyup, M. Şahan, and A. Salan, "Calculation of Production Reaction Cross Section of Some Radiopharmaceuticals Used in Nuclear Medicine by New Density Dependent Parameters," Süleyman Demirel Üniversitesi Fen Edebiyat Fakültesi Fen Dergisi, vol. 14, no. 1, pp. 57 – 61, 2019, doi: 10.29233/sdufeffd.477539

    work page doi:10.29233/sdufeffd.477539 2019

  5. [6]

    Meteorologically Driven Neutron Background Prediction for Homeland Security,

    D. VanDerwerken et al. , "Meteorologically Driven Neutron Background Prediction for Homeland Security," IEEE Trans. Nucl. Sci., vol. 65, no. 5, pp. 1187 – 1195, 2018, doi: 10.1109/TNS.2018.2821630

    work page doi:10.1109/tns.2018.2821630 2018

  6. [7]

    Pulse pileup rejection methods using a two - component Gaussian Mixture Model for fast neutron detection with pulse shape discriminating scintillator,

    A. Glenn, Q. Cheng, A. D. Kaplan, and R. Wurtz, "Pulse pileup rejection methods using a two - component Gaussian Mixture Model for fast neutron detection with pulse shape discriminating scintillator," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 988, p. 164905, 2021, doi: 10.1016/j.nima.2020.164905

    work page doi:10.1016/j.nima.2020.164905 2021

  7. [8]

    Value - Assigned Pulse Shape Discrimination for Neutron Detectors,

    F. C. E. Teh et al. , "Value - Assigned Pulse Shape Discrimination for Neutron Detectors," IEEE Trans. Nucl. Sci., vol. 68, no. 8, pp. 2294 – 2300, 2021, doi: 10.1109/TNS.2021.3091126

    work page doi:10.1109/tns.2021.3091126 2021

  8. [9]

    NeutralNet: Development and testing of a machine learning solution for pulse shape discrimination,

    R. L. Garnett and S. H. Byun, "NeutralNet: Development and testing of a machine learning solution for pulse shape discrimination," Appl. Radiat. Isot., vol. 211, p. 111384, 2024, doi: 10.1016/j.apradiso.2024.111384

    work page doi:10.1016/j.apradiso.2024.111384 2024

  9. [10]

    Pulse Shape Discrimination Based on the Tempotron: A Powerful Classifier on GPU,

    H. Liu, P. Li, M. Liu, K. Wang, Z. Zuo, and B. Liu, "Pulse Shape Discrimination Based on the Tempotron: A Powerful Classifier on GPU," IEEE Trans. Nucl. Sci., vol. 71, no. 10, pp. 2297 – 2308, 2024, doi: 10.1109/TNS.2024.3444888

    work page doi:10.1109/tns.2024.3444888 2024

  10. [12]

    A new linear method of discrimination between elementary particles in scintillation counters,

    E. Gatti and F. D. Martini, "A new linear method of discrimination between elementary particles in scintillation counters," in Nuclear Electronics II. Proceedings of the Conference on Nuclear Electronics. V. II , 1962, pp. 265 –

    work page 1962

  11. [13]

    Digital discrimination of neutrons and γ - rays in liquid scintillators using pulse gradient analysis,

    B. D’Mellow, M. D. Aspinall, R. O. Mackin, M. J. Joyce, and A. J. Peyton, "Digital discrimination of neutrons and γ - rays in liquid scintillators using pulse gradient analysis," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 578, no. 1, pp. 191 – 197, 2007, doi: 10.1016/j.nima.2007.04.174

    work page doi:10.1016/j.nima.2007.04.174 2007

  12. [14]

    A digital method for the discrimination of neutrons and gamma rays with organic scintillation detectors using frequency gradient analysis,

    G. Liu, M. J. Joyce, X. Ma, and M. D. Aspinall, "A digital method for the discrimination of neutrons and gamma rays with organic scintillation detectors using frequency gradient analysis," IEEE Trans. Nucl. Sci., vol. 57, no. 3, pp. 1682 – 1691, 2010, doi: 10.1109/TNS.2010.2044246

    work page doi:10.1109/tns.2010.2044246 2010

  13. [15]

    Discrete Fourier Transform Method for Discrimination of Digital Scintillation Pulses in Mixed Neutron - Gamma Fields,

    M. J. Safari, F. A. Davani, H. Afarideh, S. Jamili, and E. Bayat, "Discrete Fourier Transform Method for Discrimination of Digital Scintillation Pulses in Mixed Neutron - Gamma Fields," IEEE Trans. Nucl. Sci., vol. 63, no. 1, pp. 325 – 332, 2016, doi: 10.1109/TNS.2016.2514400

    work page doi:10.1109/tns.2016.2514400 2016

  14. [16]

    Digital discrimination of neutrons and gamma - rays in liquid scintillators using wavelets,

    S. Yousefi, L. Lucchese, and M. D. Aspinall, "Digital discrimination of neutrons and gamma - rays in liquid scintillators using wavelets," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 598, no. 2, pp. 551 – 555, 2009, doi: 10.1016/j.nima.2008.09.028

    work page doi:10.1016/j.nima.2008.09.028 2009

  15. [19]

    Discrimination of neutrons and gamma - rays in plastic scintillator based on spiking cortical model,

    B. - Q. Liu, H. - R. Liu, L. Chang, Y. - X. Cheng, Z. Zuo, and P. Li, "Discrimination of neutrons and gamma - rays in plastic scintillator based on spiking cortical model," Nucl. Eng. Technol., vol. 55, no. 9, pp. 3359 – 3366, 2023, doi: 10.1016/j.net.2023.04.032

    work page doi:10.1016/j.net.2023.04.032 2023

  16. [20]

    Machine learning for digital pulse shape discrimination,

    T. S. Sanderson, C. D. Scott, M. Flaska, J. K. Polack, and S. A. Pozzi, "Machine learning for digital pulse shape discrimination," in 2012 IEEE Nuclear Science Symposium and Medical Imaging Conference Record (NSS/MIC) , 27 Oct. – 3 Nov. 2012 2012, pp. 199 – 202, doi: 10.1109/NSSMIC.2012.6551092

    work page doi:10.1109/nssmic.2012.6551092 2012

  17. [21]

    K - Nearest Neighbors regression for the discrimination of gamma rays and neutrons in organic scintillators,

    M. Durbin, M. A. Wonders, M. Flaska, and A. T. Lintereur, "K - Nearest Neighbors regression for the discrimination of gamma rays and neutrons in organic scintillators," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 987, p. 164826, 2021, doi: 10.1016/j.nima.2020.164826

    work page doi:10.1016/j.nima.2020.164826 2021

  18. [22]

    Pulse Shape Discrimination of n/γ in Liquid Scintillator at PMT Nonlinear Region Using Artificial Neural Network Technique,

    E. Yun, J. Y. Choi, S. Y. Kim, and K. K. Joo, "Pulse Shape Discrimination of n/γ in Liquid Scintillator at PMT Nonlinear Region Using Artificial Neural Network Technique," Sensors, vol. 24, no. 24, p. 8060, 2024, doi: 10.3390/s24248060

    work page doi:10.3390/s24248060 2024

  19. [23]

    Pulse shape discrimination and exploration of scintillation signals using convolutional neural networks,

    J. Griffiths, S. Kleinegesse, D. Saunders, R. Taylor, and A. Vacheret, "Pulse shape discrimination and exploration of scintillation signals using convolutional neural networks," Mach. Learn.: Sci. Technol., vol. 1, no. 4, p. 045022, 2020, doi: 10.1088/2632 - 2153/abb781

    work page doi:10.1088/2632 2020

  20. [24]

    Attention is all you need,

    A. Vaswani et al. , "Attention is all you need," in Proceedings of the 31st International Conference on Neural Information Processing Systems , Long Beach, California, USA, 2017: Curran Associates Inc., pp. 6000 – 6010, doi: 10.5555/3295222.3295349

    work page doi:10.5555/3295222.3295349 2017

  21. [25]

    Pulse shape discrimination between (fast or thermal) neutrons and gamma rays with plastic scintillators: State of the art,

    G. H. V. Bertrand, M. Hamel, S. Normand, and F. Sguerra, "Pulse shape discrimination between (fast or thermal) neutrons and gamma rays with plastic scintillators: State of the art," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 776, pp. 114 – 128, 2015, doi: 10.1016/j.nima.2014.12.024

    work page doi:10.1016/j.nima.2014.12.024 2015

  22. [26]

    Recent developments in plastic scintillators with pulse shape discrimination,

    N. P. Zaitseva et al. , "Recent developments in plastic scintillators with pulse shape discrimination," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 889, pp. 97 – 104, 2018, doi: 10.1016/j.nima.2018.01.093

    work page doi:10.1016/j.nima.2018.01.093 2018

  23. [27]

    On the fabrication and characterization of heterogeneous composite neutron detectors with triple - pulse - shape - discrimination capability,

    A. Foster et al. , "On the fabrication and characterization of heterogeneous composite neutron detectors with triple - pulse - shape - discrimination capability," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 954, p. 161681, 2020, doi: 10.1016/j.nima.2018.11.140

    work page doi:10.1016/j.nima.2018.11.140 2020

  24. [28]

    A Review of Neutron – Gamma - Ray Discrimination Methods Using Organic Scintillators,

    I. Ahnouz, H. Arahmane, and R. Sebihi, "A Review of Neutron – Gamma - Ray Discrimination Methods Using Organic Scintillators," Nucl. Sci. Eng., vol. 198, no. 12, pp. 2241 – 2273, 2024, doi: 10.1080/00295639.2024.2316946

    work page doi:10.1080/00295639.2024.2316946 2024

  25. [29]

    Studies of neutron – γ pulse shape discrimination in EJ - 309 liquid scintillator using charge integration method,

    I. A. Pawełczak, S. A. Ouedraogo, A. M. Glenn, R. E. Wurtz, and L. F. Nakae, "Studies of neutron – γ pulse shape discrimination in EJ - 309 liquid scintillator using charge integration method," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 711, pp. 21 – 26, 2013, doi: 10.1016/j.nima.2013.01.028

    work page doi:10.1016/j.nima.2013.01.028 2013

  26. [30]

    Discrimination of neutrons and gamma - rays in plastic scintillator based on falling - edge percentage slope method,

    Z. Zuo, Y. Xiao, Z. Liu, B. Liu, and Y. Yan, "Discrimination of neutrons and gamma - rays in plastic scintillator based on falling - edge percentage slope method," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 1010, p. 165483, 2021, doi: 10.1016/j.nima.2021.165483

    work page doi:10.1016/j.nima.2021.165483 2021

  27. [31]

    Improving photoelectron counting and particle identification in scintillation detectors with Bayesian techniques,

    M. Akashi - Ronquest et al. , "Improving photoelectron counting and particle identification in scintillation detectors with Bayesian techniques," Astropart. Phys., vol. 65, pp. 40 – 54, 2015, doi: 10.1016/j.astropartphys.2014.12.006

    work page doi:10.1016/j.astropartphys.2014.12.006 2015

  28. [32]

    Pulse - shape discrimination against low - energy Ar - 39 beta decays in liquid argon with 4.5 tonne - years of DEAP - 3600 data,

    P. Adhikari et al. , "Pulse - shape discrimination against low - energy Ar - 39 beta decays in liquid argon with 4.5 tonne - years of DEAP - 3600 data," The European Physical Journal C, vol. 81, no. 9, p. 823, 2021, doi: 10.1140/epjc/s10052 - 021 - 09514 - w

    work page doi:10.1140/epjc/s10052 2021

  29. [34]

    Principal Component Analysis for pulse - shape discrimination of scintillation radiation detectors,

    T. Alharbi, "Principal Component Analysis for pulse - shape discrimination of scintillation radiation detectors," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 806, pp. 240 – 243, 2016, doi: 10.1016/j.nima.2015.10.030

    work page doi:10.1016/j.nima.2015.10.030 2016

  30. [35]

    Development of neutron - gamma discrimination technique using pattern - recognition method with digital signal processing,

    D. Takaku, T. Oishi, and M. Baba, "Development of neutron - gamma discrimination technique using pattern - recognition method with digital signal processing," Prog. Nucl. Sci. Technol, vol. 1, pp. 210 – 213, 2011, doi: 10.15669/pnst.1.210

    work page doi:10.15669/pnst.1.210 2011

  31. [36]

    A simple pulse - shape discrimination circuit,

    P. Sperr, H. Spieler, M. R. Maier, and D. Evers, "A simple pulse - shape discrimination circuit," Nucl. Instrum. Methods, vol. 116, no. 1, pp. 55 – 59, 1974, doi: 10.1016/0029 - 554X(74)90578 -

    work page doi:10.1016/0029 1974

  32. [39]

    Efficient pulse shape discrimination using scalogram image masking and decision tree,

    A. Abdelhakim and E. Elshazly, "Efficient pulse shape discrimination using scalogram image masking and decision tree," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 1050, p. 168140, 2023, doi: 10.1016/j.nima.2023.168140

    work page doi:10.1016/j.nima.2023.168140 2023

  33. [40]

    A Wavelet Packet Transform Inspired Method of Neutron - Gamma Discrimination,

    D. I. Shippen, M. J. Joyce, and M. D. Aspinall, "A Wavelet Packet Transform Inspired Method of Neutron - Gamma Discrimination," IEEE Trans. Nucl. Sci., vol. 57, no. 5, pp. 2617 – 2624, 2010, doi: 10.1109/TNS.2010.2044190

    work page doi:10.1109/tns.2010.2044190 2010

  34. [41]

    Pulse Shape Discrimination Algorithms, Figures of Merit, and Gamma - Rejection for Liquid and Solid Scintillators,

    W. G. J. Langeveld, M. J. King, J. Kwong, and D. T. Wakeford, "Pulse Shape Discrimination Algorithms, Figures of Merit, and Gamma - Rejection for Liquid and Solid Scintillators," IEEE Trans. Nucl. Sci., vol. 64, no. 7, pp. 1801 – 1809, 2017, doi: 10.1109/TNS.2017.2681654

    work page doi:10.1109/tns.2017.2681654 2017

  35. [42]

    Anti - noise performance of the pulse coupled neural network applied in discrimination of neutron and gamma - ray,

    H. Liu, Z. Zuo, P. Li, B. Liu, L. Chang, and Y. Yan, "Anti - noise performance of the pulse coupled neural network applied in discrimination of neutron and gamma - ray," Nucl. Sci. Tech., vol. 33, no. 6, p. 75, 2022, doi: 10.1007/s41365 - 022 - 01054 -

    work page doi:10.1007/s41365 2022

  36. [43]

    Random - Coupled Neural Network,

    H. Liu et al. , "Random - Coupled Neural Network," Electronics, vol. 13, no. 21, p. 4297, 2024, doi: 10.3390/electronics13214297

    work page doi:10.3390/electronics13214297 2024

  37. [44]

    2015 , journal =

    Y. LeCun, Y. Bengio, and G. Hinton, "Deep learning," Nature, vol. 521, no. 7553, pp. 436 – 444, 2015, doi: 10.1038/nature14539

    work page doi:10.1038/nature14539 2015

  38. [45]

    IEEE Transactions on Neural Networks and Learning Systems , volume=

    Z. Li, F. Liu, W. Yang, S. Peng, and J. Zhou, "A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects," IEEE Trans. Neural Networks Learn. Syst., vol. 33, no. 12, pp. 6999 – 7019, 2022, doi: 10.1109/TNNLS.2021.3084827

    work page doi:10.1109/tnnls.2021.3084827 2022

  39. [46]

    Neutron - gamma pulse shape discrimination for organic scintillation detector using 2D CNN based image classification,

    A. Karmakar, A. Pal, G. Anil Kumar, Bhavika, Vivek, and M. Tyagi, "Neutron - gamma pulse shape discrimination for organic scintillation detector using 2D CNN based image classification," Appl. Radiat. Isot., vol. 217, p. 111653, 2025, doi: 10.1016/j.apradiso.2024.111653

    work page doi:10.1016/j.apradiso.2024.111653 2025

  40. [47]

    Advanced supervised learning in multi - layer perceptrons — From backpropagation to adaptive learning algorithms,

    M. Riedmiller, "Advanced supervised learning in multi - layer perceptrons — From backpropagation to adaptive learning algorithms," Comput. Stand. Interfaces, vol. 16, no. 3, pp. 265 – 278, 1994, doi: 10.1016/0920 - 5489(94)90017 -

    work page doi:10.1016/0920 1994

  41. [48]

    Pulse shape simulation and discrimination using machine learning techniques,

    S. Dutta, S. Ghosh, S. Bhattacharya, and S. Saha, "Pulse shape simulation and discrimination using machine learning techniques," J. Instrum., vol. 18, no. 03, p. P03038, 2023, doi: 10.1088/1748 - 0221/18/03/P03038

    work page doi:10.1088/1748 2023

  42. [49]

    A Review of Recurrent Neural Networks: LSTM Cells and Network Architectures,

    Y. Yu, X. Si, C. Hu, and J. Zhang, "A Review of Recurrent Neural Networks: LSTM Cells and Network Architectures," Neural Comput., vol. 31, no. 7, pp. 1235 – 1270, 2019, doi: 10.1162/neco_a_01199

    work page doi:10.1162/neco_a_01199 2019

  43. [50]

    Discrimination of neutrons and γ - rays in liquid scintillator based on Elman neural network*,

    C. - X. Zhang et al. , "Discrimination of neutrons and γ - rays in liquid scintillator based on Elman neural network*," Chin. Phys. C, vol. 40, no. 8, p. 086204, 2016, doi: 10.1088/1674 - 1137/40/8/086204

    work page doi:10.1088/1674 2016

  44. [51]

    Mamba: Linear-Time Sequence Modeling with Selective State Spaces

    A. Gu and T. Dao, "Mamba: Linear - time sequence modeling with selective state spaces," arXiv, 2023, doi: 10.48550/arXiv.2312.00752

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2312.00752 2023

  45. [52]

    Machine learning: Trends, perspectives, and prospects,

    M. I. Jordan and T. M. Mitchell, "Machine learning: Trends, perspectives, and prospects," Science, vol. 349, no. 6245, pp. 255 – 260, 2015, doi: 10.1126/science.aaa8415

    work page doi:10.1126/science.aaa8415 2015

  46. [53]

    Comparison of n - γ discrimination by zero - crossing and digital charge comparison methods,

    D. Wolski, M. Moszyński, T. Ludziejewski, A. Johnson, W. Klamra, and Ö. Skeppstedt, "Comparison of n - γ discrimination by zero - crossing and digital charge comparison methods," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 360, no. 3, pp. 584 – 592, 1995, doi: 10.1016/0168 - 9002(95)00037 -

    work page doi:10.1016/0168 1995

  47. [54]

    Neutron/Gamma Discrimination Utilizing Fuzzy C - Means Clustering of the Signal from the Liquid Scintillator,

    X. Luo, G. Liu, and J. Yang, "Neutron/Gamma Discrimination Utilizing Fuzzy C - Means Clustering of the Signal from the Liquid Scintillator," in 2010 First International Conference on Pervasive Computing, Signal Processing and Applications , 17 – 19 Sept. 2010 2010, pp. 994 – 997, doi: 10.1109/PCSPA.2010.245

    work page doi:10.1109/pcspa.2010.245 2010

  48. [55]

    Advanced pulse shape discrimination via machine learning for applications in thermonuclear fusion,

    M. Gelfusa et al. , "Advanced pulse shape discrimination via machine learning for applications in thermonuclear fusion," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 974, p. 164198, 2020, doi: 10.1016/j.nima.2020.164198

    work page doi:10.1016/j.nima.2020.164198 2020

  49. [56]

    Neutron/gamma pulse shape discrimination using short - time frequency transform,

    A. Abdelhakim and E. Elshazly, "Neutron/gamma pulse shape discrimination using short - time frequency transform," Analog Integr. Circuits Signal Process., vol. 111, no. 3, pp. 387 – 402, 2022, doi: 10.1007/s10470 - 022 - 02009 - y

    work page doi:10.1007/s10470 2022

  50. [57]

    Pulse shape discrimination of neutrons and gamma rays using Kohonen artificial neural networks,

    T. Tambouratzis, D. Chernikova, and I. Pzsit, "Pulse shape discrimination of neutrons and gamma rays using Kohonen artificial neural networks," J. Artif. Intell. Soft Comput. Res., vol. 3, 2013, doi: 10.2478/jaiscr - 2014 -

    work page doi:10.2478/jaiscr 2013

  51. [58]

    Dataset for neutron and gamma - ray pulse shape discrimination,

    K. Wang, H. Liu, P. Li, M. Liu, and Z. Zuo, "Dataset for neutron and gamma - ray pulse shape discrimination," arXiv, 2023, doi: 10.48550/arXiv.2305.18242

    work page doi:10.48550/arxiv.2305.18242 2023

  52. [60]

    An NE213A TOF spectrometer for high resolution (γ,n) reaction measurements,

    J. R. M. Annand, B. E. Andersson, I. Akkurt, and B. Nilsson, "An NE213A TOF spectrometer for high resolution (γ,n) reaction measurements," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 400, no. 2, pp. 344 – 355, 1997, doi: 10.1016/S0168 - 9002(97)01021 -

    work page doi:10.1016/s0168 1997

  53. [61]

    Technique for the measurement of intrinsic pulse - shape discrimination for organic scintillators using tagged neutrons,

    N. Mauritzson et al. , "Technique for the measurement of intrinsic pulse - shape discrimination for organic scintillators using tagged neutrons," Nucl. Instrum. Methods Phys. Res., Sect. A, vol. 1039, p. 167141, 2022, doi: 10.1016/j.nima.2022.167141

    work page doi:10.1016/j.nima.2022.167141 2022

  54. [62]

    Tagging fast neutrons from an 241Am/9Be source,

    J. Scherzinger et al. , "Tagging fast neutrons from an 241Am/9Be source," Appl. Radiat. Isot., vol. 98, pp. 74 – 79, 2015, doi: 10.1016/j.apradiso.2015.01.003

    work page doi:10.1016/j.apradiso.2015.01.003 2015

  55. [63]

    Morphological filters -- Part II: Their relations to median, order - statistic, and stack filters,

    P. Maragos and R. Schafer, "Morphological filters -- Part II: Their relations to median, order - statistic, and stack filters," IEEE Trans. Acoust. Speech Signal Process., vol. 35, no. 8, pp. 1170 – 1184, 1987, doi: 10.1109/TASSP.1987.1165254

    work page doi:10.1109/tassp.1987.1165254 1987

  56. [64]

    E., & Raftery, A

    D. L. Donoho and I. M. Johnstone, "Adapting to Unknown Smoothness via Wavelet Shrinkage," J. Am. Stat. Assoc., vol. 90, no. 432, pp. 1200 – 1224, 1995, doi: 10.1080/01621459.1995.10476626

    work page doi:10.1080/01621459.1995.10476626 1995

  57. [65]

    On the use of windows for harmonic analysis with the discrete Fourier transform,

    F. J. Harris, "On the use of windows for harmonic analysis with the discrete Fourier transform," Proc. IEEE, vol. 66, no. 1, pp. 51 – 83, 1978, doi: 10.1109/PROC.1978.10837

    work page doi:10.1109/proc.1978.10837 1978

  58. [66]

    Rumelhart, Geoffrey E

    D. E. Rumelhart, G. E. Hinton, and R. J. Williams, "Learning representations by back - propagating errors," Nature, vol. 323, no. 6088, pp. 533 – 536, 1986, doi: 10.1038/323533a0

    work page doi:10.1038/323533a0 1986

  59. [67]

    Multilayer feedforward networks are universal approximators,

    K. Hornik, M. Stinchcombe, and H. White, "Multilayer feedforward networks are universal approximators," Neural Netw., vol. 2, no. 5, pp. 359 – 366, 1989, doi: 10.1016/0893 - 6080(89)90020 -

    work page doi:10.1016/0893 1989

  60. [68]

    Lecun, L

    Y. Lecun, L. Bottou, Y. Bengio, and P. Haffner, "Gradient - based learning applied to document recognition," Proc. IEEE, vol. 86, no. 11, pp. 2278 – 2324, 1998, doi: 10.1109/5.726791

    work page doi:10.1109/5.726791 1998

  61. [69]

    Imagenet classification with deep convolutional neural networks,

    A. Krizhevsky, I. Sutskever, and G. E. Hinton, "Imagenet classification with deep convolutional neural networks," Adv. Neural Inf. Process. Syst., vol. 25, 2012, doi: 10.5555/2999134.2999257

    work page doi:10.5555/2999134.2999257 2012

  62. [70]

    Mamba - 360: Survey of state space models as transformer alternative for long sequence modelling: Methods, Applications, and Challenges,

    B. N. Patro and V. S. Agneeswaran, "Mamba - 360: Survey of state space models as transformer alternative for long sequence modelling: Methods, Applications, and Challenges," Eng. Appl. Artif. Intell., vol. 159, p. 111279, 2025, doi: 10.1016/j.engappai.2025.111279

    work page doi:10.1016/j.engappai.2025.111279 2025

  63. [71]

    The Influence of Mamba Model Hyperparameters on Training Time and Accuracy,

    Y. Deng, Y. Zhang, and J. Zhang, "The Influence of Mamba Model Hyperparameters on Training Time and Accuracy," in 2024 IEEE International Conference on Smart Internet of Things (SmartIoT) , 14 – 16 Nov. 2024 2024, pp. 293 – 300, doi: 10.1109/SmartIoT62235.2024.00052

    work page doi:10.1109/smartiot62235.2024.00052 2024

  64. [72]

    Transformers in Vision: A Survey,

    S. Khan, M. Naseer, M. Hayat, S. W. Zamir, F. S. Khan, and M. Shah, "Transformers in Vision: A Survey," ACM Comput. Surv., vol. 54, no. 10s, p. Article 200, 2022, doi: 10.1145/3505244

    work page doi:10.1145/3505244 2022

  65. [73]

    Quantifying the Hyperparameter Sensitivity of Neural Networks for Character - level Sequence - to - Sequence Tasks,

    A. Wiemerslage, K. Gorman, and K. von der Wense, "Quantifying the Hyperparameter Sensitivity of Neural Networks for Character - level Sequence - to - Sequence Tasks," St. Julian’s, Malta, March 2024: Association for Computational Linguistics, in Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Vo...

    work page doi:10.18653/v1/2024.eacl 2024