Pith. sign in

REVIEW 4 major objections 5 minor 34 references

Performance Study of a Position-sensitive Plastic Scintillator Detector

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read The paper claims a flat 200×200×6 mm plastic scintillator with 64 SiPMs can reconstruct simulated muon hits to about 1.5 mm using a CNN trained only on timing data, with 22.29 ps timing resolution.

desk verdict Simulation-only design study with a new geometry and a flawed geometric reconstruction; the headline numbers are simulation outputs, not measured performance, but the work is coherent and deserves a referee. read the letter →

arxiv 2504.19083 v1 pith:DL5VA2ES submitted 2025-04-27 physics.ins-det hep-ex

S.K.Lv , C.S.Dai , D.D.Hu , T.C.Zhong , W.F.Wu , X.J.Wang This is my paper · ORCID
classification physics.ins-dethep-ex
keywords plasticscintillatorsiliconphotomultiplierposition-sensitivedetectortimeresolutionspatialconvolutionalneuralnetworkGEANT4simulationbeampositionmonitoring
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to show that a flat, unwrapped plastic scintillator panel surrounded by 64 silicon photomultipliers can serve as a cheap, compact position-sensitive detector, replacing expensive photomultiplier tubes and complex fiber couplings used in earlier designs. All reported numbers come from a GEANT4 simulation of 3 GeV/c muons: a timing resolution of 22.29 ps, a geometric position reconstruction at about 3.4 mm, and a convolutional-neural-network position reconstruction at about 1.5 mm in both X and Y. A sympathetic reader would care because, if the simulation is faithful, this geometry gives millimeter-level beam or particle positioning with straightforward mechanics and low-cost readout.

What carries the argument

The mechanism that carries the argument is the spatial encoding of photon arrival times. Photons created at the muon track reach nearby SiPMs sooner, so the 64 readout times form a pattern that shifts with hit position, and the detector's fourfold symmetry makes this pattern translationally invariant. The geometric branch uses $L = tc/n$ to turn each arrival time into a circle, pairs the 64 SiPMs to draw 2016 intersecting circles, and locates the hit at the center of the largest circle empty of intersection points. The learning branch reshapes the 64 times into a $(2,16,2)$ tensor that respects the left/right and up/down symmetry and feeds it through two convolutional layers with ReLU, batch normalization, average pooling, and dropout, followed by fully connected regression trained with MSE loss and the Adam optimizer. The timing branch uses a photoelectron-weighted average, Eq. (2), to suppress the strong dependence of single-channel timing on hit position.

What would settle it

Build the described 200×200×6 mm scintillator with 64 edge-coupled SiPMs, send a collimated source or scanned beam to known positions, and apply the same weighted-time and CNN algorithms: measured residuals far from 1.5 mm, or a time spread far from 22.29 ps, would falsify the central claim. A faster check is to measure the actual single-photoelectron waveform and transit-time jitter and rerun the simulation with those measured inputs; if the predicted resolution degrades to match the measurement, the simulation assumptions are the identified cause.

Watch

Extended reading notes

Core claim

The central claim is that the timing pattern across the 64 edge-mounted SiPMs contains enough spatial information to localize a hit well below the detector's 200 mm scale. The paper shows that a CNN trained to regress hit coordinates from the per-channel times reconstructs both coordinates with about 1.5 mm resolution, roughly twice as good as a geometric algorithm that converts arrival-time differences into intersecting circles and finds the largest empty circle. The paper also claims that a photoelectron-weighted average of the 64 channel times removes position-dependent timing jitter and yields a single-peaked arrival-time distribution with 22.29 ps resolution. These results are presented as simulation outcomes, demonstrating that a bare scintillator with symmetric SiPM readout can combine fast timing with precise position in one simple structure.

Load-bearing premise

The whole case rests on the GEANT4 simulation faithfully representing the real detector; if the simulated optical surfaces, SiPM photon detection efficiency, transit-time jitter, or electronic noise are optimistic, the measured position and timing resolutions will be worse than the reported 1.5 mm and 22.29 ps.

Editorial extensions

If this is right

  • A detector of this geometry could act as a beam-position monitor with millimeter-level accuracy while avoiding the mechanical complexity of slotted scintillators and fiber readout.
  • Because the CNN uses only timing information, position reconstruction could in principle proceed without precise per-channel gain or charge calibration.
  • The 22.29 ps weighted-average timing resolution would allow the same panel to serve as a time-of-flight counter while simultaneously reporting hit position.
  • The two reconstruction methods cross-check each other: the geometric result anchors the interpretation, and the CNN improvement quantifies the nonlinear information available in the timing pattern.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The 1.5 mm figure is a simulation ceiling; dark counts, crosstalk, imperfect optical coupling, and electronic noise will add real-world position jitter, so the measured resolution should be expected to be somewhat worse until the simulation is tuned to bench data.
  • The CNN may be fitting simulation-specific artifacts such as the assumed 20 ps transit-time spread and idealized optical surfaces, so transfer to a real detector may require retraining on measured waveforms or domain adaptation.
  • Following the paper's own suggestions, adding photoelectron count or pulse amplitude as extra input channels, or replacing the largest-empty-circle fit with Hough circle detection, are natural next steps that should improve both methods.
  • The paper's motivating application demands better than 1 mm, so the simulated 1.5 mm result does not yet meet that target; the contribution is evidence that the compact geometry is in the right regime, with the final push left to further algorithms or hardware refinements.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper presents a simulation study of a position-sensitive plastic scintillator detector consisting of a 200 mm × 200 mm × 6 mm EJ-200 scintillator read out by 64 SiPMs arranged on its four sides. Using GEANT4 simulations of 3 GeV/c muons, the authors simulate optical photon transport, synthesize SiPM waveforms from individual photoelectron responses, and extract channel times with a constant-fraction discriminator. They report a weighted-average time resolution of 22.29 ps (Section 3), a geometric reconstruction based on circle intersections and a "largest empty circle" search with roughly 3.4 mm resolution (Section 4.1), and a CNN regressor trained on simulated timing data that achieves approximately 1.5 mm resolution in both horizontal and vertical directions (Section 4.2). The paper concludes that the design provides millimeter-level spatial resolution and sub-30 ps timing performance.

Significance. If the claimed performance were demonstrated on a physical detector, the design would be a compact, low-cost option for beam position monitoring at facilities like XiPAF. The manuscript gives a concrete geometry, a clear simulation chain, and enough detail to reproduce the simulation setup, which is a positive feature. However, the significance is strongly limited by the fact that all headline results are single-point outputs of an unvalidated simulation; no experimental data, benchmark comparison, or uncertainty quantification is provided. In addition, the geometric algorithm in Section 4.1 contains an internal inconsistency, and the CNN evaluation in Section 4.2 measures only the network's ability to reproduce the simulation's mapping rather than the detector's physical resolution. The paper is best read as a preliminary simulation study, not as a demonstration of achieved detector performance.

major comments (4)
  1. [Section 4.1, Eq. (3)] The geometric reconstruction converts photon arrival times into straight-line distances using L = tc/n, assuming line-of-sight propagation. This directly contradicts the same subsection's statement that "the vast majority undergo multiple reflections within the scintillator." Since the arrival time is dominated by multiply-reflected paths, Eq. (3) does not measure the distance from the emission point to the SiPM, and the circle-intersection construction in Figs. 8-11 is therefore built on an invalid premise. The 3.4 mm resolution claim and Table 1 results are not supported by a consistent physical model.
  2. [Section 4.2, Figs. 15-16] The reported ~1.5 mm CNN position resolution is obtained by training the network on GEANT4-simulated timing data with simulated truth labels and then evaluating on the same simulation model. This is a self-consistency check of the network's ability to fit the simulation's input-output mapping; it provides no evidence about performance on a physical detector. A meaningful resolution claim would require experimental data, or at minimum a demonstration that the simulation reproduces measured quantities and an estimate of the simulation-to-reality domain gap.
  3. [Section 2.2 and all results] Every headline number (22.29 ps in Section 3, ~3.4 mm in Section 4.1, ~1.5 mm in Section 4.2) depends on an unvalidated simulation chain: GEANT4 optical transport, the SPE waveform superposition of Eq. (1), CFD timing, and the CNN training. The manuscript provides no comparison of simulated NPE spectra, waveform shapes, or timing distributions to any measurement, no benchmark against a known detector, and no uncertainty analysis on critical optical parameters such as the EJ-200 refractive index, surface treatment, or SiPM PDE. Without such validation, the results are assumption-laden simulation outputs and cannot be presented as achieved detector performance.
  4. [Section 3, Fig. 6] The 22.29 ps time resolution is presented as a single Gaussian fit to a simulated distribution, but the paper does not state the number of simulated events, the fit range, or the statistical uncertainty on the resolution. In addition, the claim that the SiPM transit-time spread (standard deviation 20 ps) "does not affect the outcome" is not justified, which is concerning for a result whose central quantity is a time resolution. The paper should provide the statistical and systematic uncertainty on the quoted value.
minor comments (5)
  1. [Section 2.1 and Ref. [16]] The text specifies an S13360-6025PE SiPM, but the cited reference [16] links to the S13360-3025PE model. The authors should verify which model is used and correct the reference or the text accordingly.
  2. [Table 1] The table header contains a stray "/2" and the column labels are poorly formatted, making the table harder to read than necessary.
  3. [Section 4.1] The largest-empty-circle algorithm depends on an unspecified square search-region side length and a 200×200 grid resolution; these parameters affect the reconstructed positions in Table 1 and Figs. 10-13 and should be stated explicitly.
  4. [Abstract and Section 5] The wording "a position resolution of mm level has been achieved" and "delivering exceptional temporal sensitivity" overstates results that are purely simulated. The paper should consistently refer to simulated performance until experimental validation is available.
  5. [Throughout] There are several minor grammatical and typographical errors, e.g., "timimg" in Section 3, "Relation" instead of "Equation" for Eq. (1), and inconsistent use of "SiPM" versus "SiPMs." A careful proofread is needed.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the simulated timing, geometric, and CNN results are computed from the stated simulation chain rather than returned as inputs by construction; the one self-citation is not load-bearing.

full rationale

Neither headline number reduces to its own input by construction. The 22.29 ps timing resolution is obtained by applying Eq. (2), an NPE-weighted average, to simulated per-channel timestamps and Gaussian-fitting the resulting distribution; no fitted parameter is renamed as a prediction. The CNN result is a supervised regression: the inputs are GEANT4-simulated SiPM timing features and the labels are the same simulation's true hit positions. Evaluating the trained network on validation data measures how well the simulated timing features encode position; that is a consistency/interpolation benchmark, not a quantity forced to equal the label by definition. The geometric algorithm does combine simulated times with Eq. (3) L = tc/n, but Eq. (3) is an assumed straight-line propagation model, and the reconstruction output is the fitted blank-region center, which is not algebraically identical to the timing inputs. The paper's own admission that most detected photons undergo multiple reflections is a modeling inconsistency and a simulation-fidelity concern, not a circular reduction. The only self-citation, ref. [20] from overlapping authors on a prior T0 detector, motivates the averaging strategy, but the present numerical timing result is recomputed in this paper's own simulation, so the citation is not load-bearing for the 22.29 ps claim. No uniqueness theorem, ansatz, or external constraint is imported from the authors' prior work. The central vulnerability is external validity: all performance numbers are simulation-only and are not benchmarked against a measurement, an uncertainty analysis, or released code and data. That is a correctness and validation risk, but it is not circularity under the required definition.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities; it relies on standard simulation tools, detector components, and algorithm parameters. The free parameters are algorithmic choices for reconstruction and training, not physical constants fitted to measurements. The dominant underdetermination is the fidelity of the GEANT4 simulation, which is a domain assumption rather than an invented entity.

free parameters (5)
  • Search region side length for the largest empty circle algorithm
    The paper refers to 'a square region centered at the mean value with a certain side length' without specifying how this length is chosen; it is a hand-tuned parameter of the reconstruction algorithm.
  • Search grid resolution = 200x200
    The center is searched on a 200x200 grid, a discretization choice that affects the precision of the fitted center.
  • Constant fraction discriminator threshold = 20%
    The timing point is defined at 20% of maximum amplitude on the falling edge, a chosen setting for all channels.
  • CNN early stopping patience = 2000 epochs
    Training stops when validation loss shows no improvement over 2000 epochs; this is a hyperparameter.
  • SiPM transit time jitter sigma = 20 ps
    The transit time is sampled from a Gaussian with sigma 20 ps; the paper claims this value does not affect the time resolution, but it is a modeling choice.
assumptions (4)
  • domain assumption The GEANT4 simulation faithfully models the detector's optical and electronic response.
    All results depend on the simulation's accuracy; no experimental comparison is provided.
  • domain assumption 3 GeV/c muons are minimum ionizing particles and are incident uniformly and perpendicularly.
    The simulation uses this beam condition; real beams may have different angular and energy distributions.
  • domain assumption The refractive index of the scintillator is 1.58 and the speed of light in the material is c/n for time-to-distance conversion.
    The geometric algorithm uses this relation, but the paper notes most photons undergo multiple reflections, making straight-line time-of-flight invalid.
  • domain assumption The SiPM dark count, crosstalk, and afterpulsing are not included in the digitization.
    The simulation only uses SPE waveforms and photoelectron statistics; real SiPM noise would degrade time and position resolution.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Performance Study of a Position-sensitive Plastic Scintillator Detector." pith.science (2026). https://pith.science/paper/DL5VA2ES

@misc{pith2026250419083,
  author       = {Pith},
  title        = {Pith review of: Performance Study of a Position-sensitive Plastic Scintillator Detector},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DL5VA2ES}},
  note         = {Machine review of arXiv:2504.19083}
}
read the original abstract

For a long time, scintillator detectors have suffered from relatively weak spatial resolution due to various influencing factors. Additionally, the high cost of photomultiplier tubes (PMTs) has limited the widespread adoption of scintillator detectors as position-sensitive detectors in particle and nuclear physics experiments. In recent years, thanks to the rapid development of silicon photomultipliers (SiPMs), their excellent cost-performance ratio has led to a renewed interest in scintillator detectors in particle and nuclear physics. This project provides a detailed discussion of a detector based on scintillators coupled with SiPMs, focusing on how to improve the detector's position accuracy. By developing algorithms based on traditional optical propagation, a position resolution of mm level has been achieved. Furthermore, the introduction of a machine learning CNN algorithm has further enhanced the detector's position resolution.

Figures

Figures reproduced from arXiv: 2504.19083 by the authors.

Figure 1
Figure 1. Schematic of the plastic scintillator detector with SiPMs array [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Fluorescence transmission in the detector simulated by Geant4 [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Number of photoelectrons generated by the 64 SiPMs array [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (13 more)
Figure 4
Figure 4. Figure 4: Simulated output signal of SiPM the simulated output waveform from one SiPM when a muon traverses the detector. A constant fraction timing method is applied to the signal, with the timing point defined at the 20% level of the maximum amplitude on the falling edge, ther…
Figure 5
Figure 5. Figure 5: The time resolution of different channels. [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: The arrival time of muons using weighted average algorithm. [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Schematic of the photon transmission algorithm. [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: Schematic of the Photon transmission algorithm. [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: Schematic of the Photon transmission algorithm. [PITH_FULL_IMAGE:figures/full_fig_p018_9.png]
Figure 10
Figure 10. Figure 10: Reconstructed X-axis Position at (80 mm, 80 mm). [PITH_FULL_IMAGE:figures/full_fig_p020_10.png]
Figure 11
Figure 11. Figure 11: Reconstructed Y-axis Position at (80 mm, 80 mm). [PITH_FULL_IMAGE:figures/full_fig_p021_11.png]
Figure 12
Figure 12. Figure 12: Reconstructed X-axis Position under Random Incidence Condi [PITH_FULL_IMAGE:figures/full_fig_p022_12.png]
Figure 13
Figure 13. Figure 13: Reconstructed Y-axis Position under Random Incidence Condi [PITH_FULL_IMAGE:figures/full_fig_p023_13.png]
Figure 14
Figure 14. Figure 14: An overview of a convolutional neural network (CNN) architecture [PITH_FULL_IMAGE:figures/full_fig_p027_14.png]
Figure 15
Figure 15. Figure 15: Reconstructed X-axis Position under Random Incidence Condi [PITH_FULL_IMAGE:figures/full_fig_p030_15.png]
Figure 16
Figure 16. Figure 16: Reconstructed Y-axis Position under Random Incidence Condi [PITH_FULL_IMAGE:figures/full_fig_p031_16.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 23 canonical work pages

  1. [1]

    Beam dynamics design of a proton/heavy-ion injector for the synchrotron of the Xi- PAF upgrading project[J]

    Yue,C.B., Ma,P.F, Wang,B.C., et al. Beam dynamics design of a proton/heavy-ion injector for the synchrotron of the Xi- PAF upgrading project[J]. Nuclear Instruments and Meth- ods in Physics Research Section A: Accelerators, Spectrom- eters, Detectors and Associated Equipment, 2025: 170537. doi:10.1016/j.nima.2025.170537

    work page arXiv 2025

  2. [2]

    Construction and beam commissioning of a compact proton synchrotron for space radiation environment simulation[J]

    Wang,Z.M., Chen,W., Qiu,M.T., et al. Construction and beam commissioning of a compact proton synchrotron for space radiation environment simulation[J]. Nuclear Instru- ments and Methods in Physics Research Section A: Acceler- 33 ators, Spectrometers, Detectors and Associated Equipment, 2022,1027:166283. doi:10.1016/j.nima.2021.166283

    arXiv 2022

  3. [3]

    Plastic scintil- lation detectors for time-of-flight mass measurements[J]

    Wang,K., Estrade,A., Neupane,S., et al. Plastic scintil- lation detectors for time-of-flight mass measurements[J]. Nuclear Instruments and Methods in Physics Re- search Section A: Accelerators, Spectrometers, Detec- tors and Associated Equipment, 2020, 974: 164199. doi:10.1016/j.nima.2020.164199

    work page arXiv 2020

  4. [4]

    Reaching time res- olution of less than 10 ps with plastic scintillation de- tectors[J]

    Zhao,J.W., Sun,B.H., Tanihata,I., et al. Reaching time res- olution of less than 10 ps with plastic scintillation de- tectors[J]. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, De- tectors and Associated Equipment, 2016, 823: 41-46. doi:10.1016/j.nima.2016.03.106

    work page doi:10.1016/j.nima.2016.03.106 2016

  5. [5]

    Design of a high dynamic range photomultiplier base board for the BGO ECAL of DAMPE[J]

    Zhang,Z., Zhang,Y., Dong,J., et al. Design of a high dynamic range photomultiplier base board for the BGO ECAL of DAMPE[J]. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrome- ters, Detectors and Associated Equipment, 2015, 780: 21-

    work page 2015

  6. [6]

    Study on the optimized 34 energy resolution of scintillator detectors based on SiPMs and LYSO: Ce[J]

    Hu,P., Hua,Z.H., Ma,L.S., et al. Study on the optimized 34 energy resolution of scintillator detectors based on SiPMs and LYSO: Ce[J]. Journal of Instrumentation, 2022, 17(09): T09010. doi:10.1088/1748-0221/17/09/T09010

    work page doi:10.1088/1748-0221/17/09/t09010 2022

  7. [7]

    Design and testing of a position-sensitive plastic scintilla- tor detector for fast neutron imaging[J]

    Bravar,U., Bruillard,P.J., Flckiger,E.O., et al. Design and testing of a position-sensitive plastic scintilla- tor detector for fast neutron imaging[J]. IEEE trans- actions on nuclear science, 2006, 53(6): 3894-3903. doi:10.1109/TNS.2006.886046

    work page arXiv 2006

  8. [8]

    High-spatial-resolution position-sensitive plastic scintillation optical fiber bun- dle detector[C]//Photonics

    Song,S., Kim,J., Park,J.H., et al. High-spatial-resolution position-sensitive plastic scintillation optical fiber bun- dle detector[C]//Photonics. MDPI, 2021, 8(2): 26. doi:10.3390/photonics8020026

    work page doi:10.3390/photonics8020026 2021

Show all 34 references
  1. [9]

    Anode position and last dynode timing circuits for dual-layer BGO scintilla- tor with PS-PMT based modular PET detectors[J]

    Zhang,N., Thompson,C.J., Togane,D., et al. Anode position and last dynode timing circuits for dual-layer BGO scintilla- tor with PS-PMT based modular PET detectors[J]. IEEE Transactions on Nuclear Science, 2002, 49(5): 2203-2207. doi:10.1109/TNS.2002.803815

    2002

  2. [10]

    Sub-3mm spatial resolution from a large monolithic LaBr3 (Ce) scin- tillator[J]

    Liprandi,S., Mayerhofer,M., Aldawood,S., et al. Sub-3mm spatial resolution from a large monolithic LaBr3 (Ce) scin- tillator[J]. Current Directions in Biomedical Engineering, 2017, 3(2): 655-659. doi:10.1515/cdbme-2017-0138. 35

    2017 doi

  3. [11]

    LAL Orsay, 2012

    Collazuol,G., The sipm physics and technology-a review[J]. LAL Orsay, 2012

    2012

  4. [12]

    Development of a muon detector based on a plastic scintillator and WLS fibers to be used for muon tomography system[J]

    Park,C., Kim,K.B., Baek,M.K., et al. Development of a muon detector based on a plastic scintillator and WLS fibers to be used for muon tomography system[J]. Nu- clear Engineering and Technology, 2023, 55(3): 1009-1014. doi:10.1016/j.net.2022.11.016

    2023 doi

  5. [13]

    A cosmic ray imaging sys- tem based on plastic scintillator detector with SiPM read- out[J]

    Liang,Z., Hu,T., Li,X., et al. A cosmic ray imaging sys- tem based on plastic scintillator detector with SiPM read- out[J]. Journal of Instrumentation, 2020, 15(07): C07033. doi:10.1088/1748-0221/15/07/C07033

    2020 doi

  6. [14]

    https://eljentechnology.com/products/plastic- scintillators/ej-200-ej-204-ej-208-ej-212

  7. [15]

    IEEE Transactions on Nuclear Science, 2015, 62(5): 1972-

    ter Weele,D.N., Schaart,D.R., Dorenbos,P., Scintillation de- tector timing resolution; a study by ray tracing software[J]. IEEE Transactions on Nuclear Science, 2015, 62(5): 1972-

    2015

  8. [16]

    https://www.hamamatsu.com/us/en/product/optical- sensors/mppc/S13360-3025PE.html

  9. [17]

    GEANT4—a 36 simulation toolkit[J]

    Agostinelli,S., Allison,J., Amako,K., et al. GEANT4—a 36 simulation toolkit[J]. Nuclear instruments and methods in physics research section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 2003, 506(3): 250-

    2003

  10. [18]

    Geant4 developments and applications[J]

    Allison,J., Amako,K., Apostolakis,J.E.A., et al. Geant4 developments and applications[J]. IEEE Transac- tions on nuclear science, 2006, 53(1): 270-278. doi:10.1109/TNS.2006.869826

    2006

  11. [19]

    CDF Note 5358, 2000

    Jones,M., PMT base design for the CDF-II time-of-flight system[R]. CDF Note 5358, 2000

    2000

  12. [20]

    Design and performance testing of a T0 detector for the CSR External-target Ex- periment[J]

    Hu,D., Wang,X., Shao,M., et al. Design and performance testing of a T0 detector for the CSR External-target Ex- periment[J]. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detec- tors and Associated Equipment, 2023, 1057: 168773. do...

    2023

  13. [22]

    Improved calorimetric particle identification in NA62 using machine learning techniques[J]

    Cortina Gil,E., Kleimenova, A., Minucci, E., et al. Improved calorimetric particle identification in NA62 using machine learning techniques[J]. Journal of High Energy Physics, 2023, 2023(11): 1-15. doi:10.1007/JHEP11(2023)138

    2023 doi

  14. [23]

    Machine learning method for 12C event classification and reconstruction in the active target time-projection chamber[J]

    Wu,H., Wang, Y., Wang, Y., et al. Machine learning method for 12C event classification and reconstruction in the active target time-projection chamber[J]. Nuclear In- struments and Methods in Physics Research Section A: Ac- celerators, Spectrometers, Detectors and Associated E...

    2023

  15. [24]

    NATURE COMMUNICATIONS, 2025, 16(1).doi:10.1038/s41467- 024-55042-y

    Bianchi,F., Observation of a rare beta decay of the charmed baryon with a Graph Neural Network[J]. NATURE COMMUNICATIONS, 2025, 16(1).doi:10.1038/s41467- 024-55042-y

    2025 doi

  16. [25]

    Measurements of decay branch- ing fractions of the Higgs boson to hadronic final states at the CEPC[J]

    Ma,X., Wu,Z., Wu,J., et al. Measurements of decay branch- ing fractions of the Higgs boson to hadronic final states at the CEPC[J]. Chinese Physics C, 2025. doi:10.1088/1674- 1137/adacc5

    2025 doi

  17. [26]

    doi:10.1016/j.nima.2015.01.036

    2015 doi

  18. [27]

    First appli- cation of machine learning algorithms to the position re- construction in Resistive Silicon Detectors[J]

    Siviero,F., Arcidiacono,R., Cartiglia,N., et al. First appli- cation of machine learning algorithms to the position re- construction in Resistive Silicon Detectors[J]. Journal of Instrumentation, 2021, 16(03): P03019. doi:10.1088/1748- 0221/16/03/P03019

    2021 doi

  19. [28]

    Deep learn- ing[M]

    Goodfellow,I., Bengio,Y., Courville,A., et al. Deep learn- ing[M]. Cambridge: MIT press, 2016

    2016

  20. [29]

    Using deep neural networks to improve the precision of fast-sampled particle 38 timing detectors[J]

    Kocot,M., Misan,K., Avati,V., et al. Using deep neural networks to improve the precision of fast-sampled particle 38 timing detectors[J]. arXiv preprint arXiv:2312.05883, 2023. doi:10.7494/csci.2024.25.1.5784

    2023 arXiv

  21. [30]

    Jour- nalofMachineLearningResearch, 2014, 15(1): 1929

    SrivastavaN,H.G., Krizhevsky,A., Dropout: A simple- waytopreventneuralnetworksfromoverfitting[J]. Jour- nalofMachineLearningResearch, 2014, 15(1): 1929. doi:10.5555/2627435.2670313

    2014

  22. [31]

    arXiv preprint arXiv:1609.04747, 2016

    Ruder,S., An overview of gradient descent optimiza- tion algorithms[J]. arXiv preprint arXiv:1609.04747, 2016. doi:10.4855-/arXiv.1609.04747. 39

    2016 arXiv

  23. [32]

    pmlr, 2015: 448-456

    Ioffe,S., Szegedy,C., Batch normalization: Accelerating deep network training by reducing internal covariate shift[C]//International conference on machine learning. pmlr, 2015: 448-456. doi:10.5555/3045118.3045167

    2015

  24. [35]

    arXiv preprint arXiv:1412.6980, 2014

    Kingma,D.,P., Ba,J., Adam: A method for stochas- tic optimization[J]. arXiv preprint arXiv:1412.6980, 2014. doi:10.48550/arXiv.1412.6980. 40

    2014 doi arXiv

  25. [303]

    doi:10.1016/S0168-9002(03)01368-8

    doi

  26. [1980]

    doi:10.1109/TNS.2015.2460266

    2015

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.