Pith. sign in

REVIEW 4 major objections 5 minor 34 references

Recursive Manifold Coherence for Deadtime-Aware Distributed Triggering through Geometric State Estimation

T0 review · 4 major / 5 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a simple recursive filter, applied to a low-dimensional coherence state, can preserve event-level trigger information across detector deadtime where binary coincidence logic fails.

desk verdict RMC is an IIR filter in a manifold costume; the only empirical evidence is a simulation that isn't specified, so the central claim doesn't land. read the letter →

arxiv 2601.17043 v2 pith:6QUBAYYP submitted 2026-01-20 physics.ins-det physics.comp-phphysics.data-an

Thammarat Yawisit , Pittaya Pannil This is my paper · ORCID
classification physics.ins-detphysics.comp-phphysics.data-an PACS 07.05.Hd29.40.-n
keywords deadtimepile-updistributedtriggerinformationgeometryrecursivestateestimationsoftware-definedtriggeringcoincidencelogicneutrinoobservatory
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 argues that detector deadtime need not erase trigger-level evidence. It replaces binary coincidence logic with a low-dimensional 'coherence state' updated by a first-order recursion (an exponential smoother), so during non-live intervals the state decays smoothly instead of resetting. In simulations with increasing deadtime probability, the recursion's event-recovery efficiency rolls off far slower than standard coincidence logic, recovering high-multiplicity events whose coincidence chains are broken. The authors frame this as a shift from hard vetoes to stateful, missing-data-aware triggering, with a fixed per-sample computational cost suitable for real-time pipelines.

What carries the argument

The recursive coherence state: a first-order IIR/exponential-smoothing recursion x[k] = Ax[k−1] + Bu[k], where A = ρI provides bounded exponential memory. Liveness indicators Li[k] gate observations so deadtime becomes suppressed input rather than a reset, causing the state to decay with controlled uncertainty. A correlation-based metric (Fisher information, covariance, or correlation-weighted approximation) defines distances in feature space and supplies the weighting for the state update.

What would settle it

Run RMC against a coincidence trigger on a benchmark generator with an independent signal process—e.g., Poisson-distributed hits with wideband timing jitter and per-channel random deadtime not correlated with the signal—and measure efficiency versus deadtime probability. If RMC's advantage narrows to the persistence time of ρ and vanishes when ρ is tuned to the deadtime duration, the proposed 'manifold coherence' mechanism is not the operative driver.

Watch

Extended reading notes

Core claim

The central claim is that treating deadtime as missing data in a stateful estimator—rather than as a hard veto—preserves event-level coherence. The update law x[k] = Ax[k−1] + Bu[k], with A = ρI giving exponential memory, propagates correlation evidence across gaps, and the scalar coherence score G[k] = φ(x[k]) can exceed threshold even when direct coincidence is broken. Simulation results show RMC retains higher efficiency than coincidence logic as deadtime probability rises, and the deadtime-bridging visualization shows the state re-locking to the truth once observations resume.

Load-bearing premise

The claimed advantage rests on the unstated simulation being a fair representation of real detector deadtime and signal correlation; the paper gives no event generator, background model, or baseline specification, and its own limitations section admits correlated backgrounds, afterpulsing, and calibration drifts are absent.

Editorial extensions

If this is right

  • Trigger decisions can be made with bounded memory and fixed per-sample cost, compatible with FPGA fixed-point and software-based pipelines.
  • Events with high multiplicity and extended topology survive partial deadtime where coincidence triggers lose them.
  • Deadtime becomes observable as uncertainty growth rather than silent data loss, giving operators a continuous liveness diagnostic.
  • The same recursion works across detector geometries because it only requires compact charge/timing features and liveness flags.
  • With an appropriate persistence parameter, trigger efficiency degrades gracefully with deadtime probability rather than sharply collapsing.

Reading between the lines

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

  • The mechanism is formally an IIR smoother, so the claimed advantage may reduce to choosing a persistence time constant longer than the deadtime gap; a fair baseline should include a stateful (smoothed) coincidence trigger to isolate the contribution of the coherence geometry itself.
  • The state decay rate could be calibrated to known deadtime statistics, turning the trigger into a continuous live-fraction estimator and connecting this work to standard deadtime-correction theory.
  • A testable extension would be to run the recursion on real data with afterpulsing or correlated backgrounds—explicitly absent in this study—to see whether state-dependent persistence or low-rank extensions are needed.
  • Because the recursion is linear and scalar, it could be interpreted as a matched filter for signal persistence, suggesting that the detector-agnostic claim may be better phrased as a tunable-persistence trigger rather than a manifold-geometric one.
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 proposes Recursive Manifold Coherence (RMC), a trigger-logic framework that replaces binary coincidence windows with a first-order linear recursion x[k] = Ax[k-1] + Bu[k] over validated charge/timing features u[k] = L[k]y[k], where L[k] encodes channel liveness. With the recommended choice A = rho I, the state decays smoothly during deadtime instead of resetting, so the scalar trigger score G[k] = phi(x[k]) can remain above threshold even when direct coincidence chains are broken. The authors claim, based on 'simulation studies representative of large optical detector arrays' in Section 4, that RMC recovers event-level coherence for high-multiplicity topologies and degrades more gracefully than baseline coincidence logic as deadtime probability increases. Section 2 presents the update law and an information-geometric interpretation; Section 3 discusses implementation complexity and firmware compatibility; Section 6 concedes that correlated backgrounds, afterpulsing, and calibration drifts are not modeled.

Significance. If the performance claim were established, the proposal would offer a simple, bounded-memory alternative to coincidence-based triggers, with an implementation cost as low as O(m) per sample when A = rho I. The algebraic recursion is correct as a linear filter, and the paper usefully identifies that stateless binary logic discards correlated evidence during deadtime. However, the central demonstration is not supported. The simulation study is unspecified, no false-alarm-rate control is provided, and the 'deadtime bridging' behavior is definitionally built into the exponential-smoothing update law. The geometric apparatus (Eq. 2.2) is never used in the evaluation. As a result, the claimed 'superior robustness' is not established, and the contribution currently reduces to an exponential smoother with a decorational information-geometry layer.

major comments (4)
  1. [Section 4, Figs. 5-6] The central performance claim rests entirely on the simulation comparison, but the simulation is not specified. There is no generative model for true events, no noise or background rate, no detector response function, no liveness/deadtime process, no definition of the baseline 'coincidence logic', no trigger threshold Gamma, and no false-alarm or trigger-rate operating point. Efficiency alone is not a valid trigger metric: any acceptance rule can achieve high efficiency by lowering its threshold. Because RMC with rho in [0.85, 0.95] is a long-memory accumulator, it will also accumulate background fluctuations; without matched false-alarm rates, the slower efficiency roll-off in Figs. 5-6 may simply reflect a higher false-trigger rate.
  2. [Section 2.3, Remark 2.1 and Section 4.2] The claimed 'recovery' of coherence through deadtime is definitional, not empirical. When u[k] is approximately zero, Eq. (2.3) reduces to x[k] = A x[k-1], so with A = rho I the state decays exponentially by construction. The deadtime-bridging visualization (Fig. 1 and Fig. 3) is therefore a plot of the update law, not evidence that RMC 'recovers event-level coherence'. To substantiate the claim, the authors need an independent definition of the true event process and a detection rule whose probability of detection is evaluated at a fixed false-alarm probability.
  3. [Section 6] The limitations section concedes that spatially heterogeneous noise, correlated backgrounds, afterpulsing, and time-dependent calibration drifts are not modeled. These are exactly the effects that would penalize a recursive accumulator with fixed persistence, because they create correlated or slowly varying input that the IIR memory would retain. The absence of these effects in the simulation, combined with the lack of matched false-alarm control, makes the claim of 'superior robustness against data fragmentation' in the abstract unsubstantiated for realistic detector conditions.
  4. [Sections 2.2, 2.4, and 4] The information-geometric metric G_ij(theta) in Eq. (2.2) and the scoring functional phi in Eq. (2.4) are never instantiated in the evaluation. The simulation results in Figs. 5-6 depend only on the linear recursion with A = rho I; the manifold geometry does not enter the update or the decision rule. Either the geometric component should be operationalized and tied to the reported results, or the paper should be repositioned as a study of recursive persistence for deadtime handling, without the manifold-coherence framing.
minor comments (5)
  1. [Eq. (2.2)] The notation 'dTheta^2 = G_ij(theta) dtheta^i dtheta^j' is nonstandard; usually one writes ds^2 or explicitly sums over repeated indices. This should be corrected for clarity.
  2. [Fig. 5 and Fig. 6] The x-axis label is inconsistent: Fig. 5 uses 'Deadtime probability' in [0,0.7], while Fig. 6 uses 'deadtime probability' in [0,50] (apparently percent). The y-axis 'Event recovery efficiency' is undefined; a definition in the text or caption is needed.
  3. [Fig. 3] The plotted 'uncertainty' is not defined. Since the recursion is deterministic, the uncertainty envelope presumably comes from an external model, but the text does not specify how it is computed.
  4. [Fig. 2 caption] The caption contains a typo: 'weights / exp(¡1 2 d 2 M)' appears to be a character-encoding artifact; should read something like 'exp(-(1/2) d_M^2)'.
  5. [References] Reference [23] is a self-citation to an arXiv preprint whose connection to the present work is not explained; if it is background, it should be cited in context, not listed without discussion.

Circularity Check

1 steps flagged · score 6.0 of 10

RMC's deadtime 'recovery' is built into Eq. (2.3): the simulation demonstration restates the recursion rather than testing an independent prediction.

  1. self definitional [Sec. 2.3 (Eq. 2.3, Remark 2.1), Sec. 2.4, Secs. 4.1–4.2 (Figs. 3, 5–6)]
    "When many channels are non-live so that u[k]≈0, equation (2.3) reduces to x[k]=Ax[k−1]. Rather than resetting, the coherence state decays smoothly, representing increasing uncertainty instead of immediate information loss ... This mechanism allows correlated structure to persist across deadtime gaps without requiring explicit coincidence recovery. ... Because the coherence state propagates through deadtime, G[k] can exceed threshold even when direct coincidence chains are broken, enabling recovery of high-multiplicity or partially obscured event topologies."

    The claimed ability to 'recover event-level coherence even when direct coincidence chains are broken' is not an independent outcome of the simulations; it is literally the update law (2.3) with A=ρI. During deadtime, u[k]≈0 forces x[k]=A x[k−1], and the trigger statistic G[k]=φ(x[k]) inherits that persistence by construction. The 'demonstrations' in Figs. 3, 5, and 6 therefore plot the algorithm's own constitutive recursion. No independent event generator, detector response, background/noise model, baseline coincidence parameters, threshold Γ, or matched false-alarm rate is supplied (Sec. 6 concedes correlated backgrounds, afterpulsing, and calibration drifts are absent). The empirical claim reduces to restating the definition of the state update.

full rationale

The paper contains no load-bearing self-citation chain or imported uniqueness theorem; reference [23] is a self-citation but is not used as the argument's support. The circularity is narrower and definitional: the central contribution—coherence persisting through deadtime—is exactly the chosen recursion x[k]=ρx[k−1] when u[k]≈0. The qualitative deadtime-bridging plot and the efficiency curves simulate that same law, and because Γ, the baseline, the noise model, and false-alarm rates are never specified, the 'demonstration' cannot be distinguished from the construction. This is partial, not total, circularity: the framework remains a coherent algorithmic proposal, and its complexity, firmware-suitability, and detector-agnostic design claims are independent of the circular step. But the primary empirical validation as presented reduces to the definition, warranting a score of 6.

Assumptions & free parameters 3 free parameters · 4 assumptions · 2 invented entities

The central performance claim rests on hand-picked ρ and an unreported threshold Γ, plus assumptions that liveness is exactly known, that a first-order linear state is sufficient, and that decay during deadtime equals uncertainty growth. The 'manifold' metric is never used. No external benchmark or real data enters the evaluation.

free parameters (3)
  • Persistence parameter ρ (A = ρI) = 0.85, 0.90, 0.95 (Fig. 5)
    Chosen by hand; controls how long the state persists through deadtime and therefore largely determines the bridging performance.
  • Trigger threshold Γ = not reported
    Any efficiency comparison depends on a threshold; without reporting how Γ was set (fixed or per deadtime), the curves in Figures 5–6 are not reproducible.
  • Input/aggregation matrix B and metric G_ij = not specified
    B determines how validated observations update the state; G_ij in Eq. (2.2) is never instantiated, but if used would require additional free choices.
assumptions (4)
  • domain assumption Liveness indicator L_i[k] is known exactly and deadtime is a binary event per sample.
    Section 2.1 defines u_i[k] = L_i[k] y_i[k]; in real detectors liveness may be uncertain or known only after data arrives.
  • ad hoc to paper A first-order linear recursion with fixed persistence captures event-level coherence.
    Sections 2.3 and 6 choose the update x[k] = Ax[k−1] + Bu[k] for simplicity; no argument is given that coherence has exactly this dynamics.
  • ad hoc to paper Decaying state during deadtime represents increasing uncertainty rather than information loss.
    Remark 2.1 asserts this interpretation; it is a modeling choice, not a derived result.
  • domain assumption A scalar function of the state is sufficient for trigger decisions.
    Section 2.4 assumes G[k] = φ(x[k]) captures all decision-relevant information without testing against a multi-dimensional discriminator or full likelihood.
invented entities (2)
  • Coherence state x[k]
    purpose: A low-dimensional internal state intended to carry correlated charge/timing evidence through deadtime.
    The paper postulates x[k] as the core estimator, but it has no observable handle outside the trigger score; it is a reformulation of an IIR filter state.
  • Information metric G_ij(θ) in Eq. (2.2)
    purpose: Gives geometric/statistical meaning to distances in parameter space.
    Introduced in the framework but never instantiated or computed; it does not appear in the simulations or the decision rule.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Recursive Manifold Coherence for Deadtime-Aware Distributed Triggering through Geometric State Estimation." pith.science (2026). https://pith.science/paper/6QUBAYYP

@misc{pith2026260117043,
  author       = {Pith},
  title        = {Pith review of: Recursive Manifold Coherence for Deadtime-Aware Distributed Triggering through Geometric State Estimation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6QUBAYYP}},
  note         = {Machine review of arXiv:2601.17043}
}
read the original abstract

Large-scale neutrino observatories operate under unavoidable detector deadtime and signal pile-up, leading to systematic inefficiencies in conventional coincidence-based trigger systems. Such triggers typically rely on binary temporal windows and assume continuous sensor availability, causing partial or complete loss of correlated signal information during non-live intervals. We introduce Recursive Manifold Coherence (RMC), a geometric framework that reformulates distributed trigger logic as a continuous state estimation problem in a low-dimensional information space defined by correlated charge and timing observables. Instead of applying hard vetoes during deadtime, the proposed method employs a recursive update rule that propagates a coherence state across sensor nodes, allowing partially obscured signals to be retained and evaluated consistently. Using simulation studies representative of large optical detector arrays, we demonstrate that RMC successfully recovers event-level coherence for high-multiplicity topologies even when direct coincidence chains are broken. By treating the detector response as a smooth manifold rather than discrete hits, the framework achieves superior robustness against data fragmentation compared to standard binary logic. The framework is detector-agnostic and compatible with software-defined trigger pipelines, providing a flexible foundation for deadtime-aware analysis and triggering strategies in future distributed detector systems.

Figures

Figures reproduced from arXiv: 2601.17043 by the authors.

Figure 1
Figure 1. Deadtime-bridging visualization. (A) Nominal operation without deadtime: baseline coincidence logic and RMC track the underlying truth. (B) Deadtime overlapping the signal peak: baseline coincidence collapses, while RMC propagates a coherence estimate with controlled decay across the non-live interval and re-locks when observations resume. Shaded region indicates detector deadtime. a problem of stateful information … view at source ↗
Figure 2
Figure 2. Feature-space geometry induced by RMC weighting. (A) Live observations in the raw charge–timing feature space. (B) Effective geometry after deadtime-aware RMC weighting, where statistically consistent signal-like structure is enhanced while background and deadtime-removed observations are suppressed. Distances reflect statistical distinguishability rather than physical separation. 0 2 4 6 8 10 time (s) -0.25 0.00 0.… view at source ↗
Figure 3
Figure 3. Evolution of the RMC coherence state and associated uncertainty. During detector deadtime (shaded region), validated input is suppressed and the state propagates according to the recursive update law, exhibiting controlled decay and uncertainty growth. When observations resume, the state rapidly re-locks to the underlying trajectory. deadtime is handled algorithmically rather than through pipeline-level vetoes, rese… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Integration of Recursive Manifold Coherence into a streaming trigger pipeline. RMC operates on compact charge and timing features with explicit liveness tagging, enabling deadtime-aware coherence estimation without buffering raw waveforms. The recursive state update ru…
Figure 5
Figure 5. Figure 5: Event recovery efficiency as a function of deadtime probability for different persistence parameters ρ. RMC remains stable across a wide range of ρ values, while baseline coincidence logic rapidly degrades. 0 10 20 30 40 50 deadtime probability 0.0 0.2 0.4 0.6 0.8 1.0 …
Figure 6
Figure 6. Figure 6: Event recovery efficiency as a function of deadtime probability. RMC: coherent structure can be retained, propagated, and evaluated under detector deadtime, precisely in the regimes where conventional coincidence-based trigger logic tends to fragment or discard events …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 6 canonical work pages

  1. [27]

    Kalman filtering with intermittent observations,

    Y. Zhang and X. R. Li, “Kalman filtering with intermittent observations,”IEEE Transactions on Automatic Control, vol. 45, no. 5, pp. 879–885, 2000. doi:10.1109/9.855552

    work page doi:10.1109/9.855552

  2. [1]

    The IceCube Neutrino Observatory: Instrumenta- tion and Online Systems,

    M. G. Aartsenet al.(IceCube Collaboration), “The IceCube Neutrino Observatory: Instrumenta- tion and Online Systems,”Journal of Instrumen- tation, vol. 12, P03012, 2017. doi:10.1088/1748- 0221/12/03/P03012

    doi:10.1088/1748- 2017

  3. [2]

    Triggering at the LHC,

    W. H. Smith, “Triggering at the LHC,”An- nual Review of Nuclear and Particle Science, vol. 66, pp. 123–150, 2016. doi:10.1146/annurev- nucl-102115-044655

    doi:10.1146/annurev- 2016

  4. [3]

    The ATLAS Trigger Sys- tem,

    ATLAS Collaboration, “The ATLAS Trigger Sys- tem,”European Physical Journal C, vol. 77, 317,

  5. [4]

    The CMS Trigger System,

    CMS Collaboration, “The CMS Trigger System,” Journal of Instrumentation, vol. 12, P01020, 2017. doi:10.1088/1748-0221/12/01/P01020

    doi:10.1088/1748-0221/12/01/p01020 2017

  6. [5]

    A Roadmap for HEP Software and Computing R&D,

    J. Albrechtet al., “A Roadmap for HEP Software and Computing R&D,”Computing and Software for Big Science, vol. 3, 7, 2019. doi:10.1007/s41781- 018-0018-8

    doi:10.1007/s41781- 2019

  7. [6]

    Real-Time Event Selection and Trigger Systems in High Energy Physics,

    R. Bauret al., “Real-Time Event Selection and Trigger Systems in High Energy Physics,”Comput- ing and Software for Big Science, vol. 5, 1, 2021. doi:10.1007/s41781-020-00052-3

    work page doi:10.1007/s41781-020-00052-3 2021

  8. [7]

    G. F. Knoll,Radiation Detection and Measurement, 4th ed., Wiley, 2010

    2010

Show all 34 references
  1. [8]

    Dead-timecorrectionsforcountingsys- tems,

    J.C.Davis, “Dead-timecorrectionsforcountingsys- tems,”Nuclear Instruments and Methods, vol. 18, pp. 1–7, 1962. doi:10.1016/0029-554X(62)90127-6

    1962 doi

  2. [9]

    Pileup subtraction using jet areas,

    M. Cacciari, G. P. Salam, and G. Soyez, “Pileup subtraction using jet areas,”Physics Letters B, vol. 659, pp. 119–126, 2008. doi:10.1016/j.physletb.2007.09.077

    2008 doi

  3. [10]

    A New Approach to Linear Fil- tering and Prediction Problems,

    R. E. Kalman, “A New Approach to Linear Fil- tering and Prediction Problems,”ASME Journal of Basic Engineering, vol. 82, pp. 35–45, 1960. doi:10.1115/1.3662552

    1960 doi

  4. [11]

    B. D. O. Anderson and J. B. Moore,Optimal Fil- tering, Prentice Hall, 1979

    1979

  5. [12]

    Ljung,System Identification: Theory for the User, 2nd ed., Prentice Hall, 1999

    L. Ljung,System Identification: Theory for the User, 2nd ed., Prentice Hall, 1999

    1999

  6. [13]

    Särkkä,Bayesian Filtering and Smooth- ing, Cambridge University Press, 2013

    S. Särkkä,Bayesian Filtering and Smooth- ing, Cambridge University Press, 2013. doi:10.1017/CBO9781139344203 7

    2013 doi

  7. [14]

    A.V.OppenheimandR.W.Schafer,Discrete-Time Signal Processing, 3rd ed., Prentice Hall, 2010

    2010

  8. [15]

    S.K.Mitra,Digital Signal Processing: A Computer- Based Approach, McGraw-Hill, 2011

    2011

  9. [16]

    Lyons,Understanding Digital Signal Processing, 3rd ed., Pearson, 2011

    R. Lyons,Understanding Digital Signal Processing, 3rd ed., Pearson, 2011

    2011

  10. [17]

    Meyer-Baese,Digital Signal Processing with Field Programmable Gate Arrays, Springer, 2014

    U. Meyer-Baese,Digital Signal Processing with Field Programmable Gate Arrays, Springer, 2014. doi:10.1007/978-3-642-45309-6

    2014 doi

  11. [18]

    Amari,Information Geometry and Its Applica- tions, Springer, Tokyo, 2016

    S. Amari,Information Geometry and Its Applica- tions, Springer, Tokyo, 2016. doi:10.1007/978-4-431- 55978-8

    2016 doi

  12. [19]

    Information and the Accuracy At- tainable in the Estimation of Statistical Parame- ters,

    C. R. Rao, “Information and the Accuracy At- tainable in the Estimation of Statistical Parame- ters,”Bulletin of the Calcutta Mathematical Society, vol. 37, pp. 81–91, 1945

    1945

  13. [20]

    T. M. Cover and J. A. Thomas,Elements of Infor- mation Theory, 2nd ed., Wiley, 2006

    2006

  14. [21]

    Cohen,Time-Frequency Analysis, Prentice Hall, 1995

    L. Cohen,Time-Frequency Analysis, Prentice Hall, 1995

    1995

  15. [22]

    H. L. Van Trees,Detection, Estimation, and Mod- ulation Theory, Wiley, 2001

    2001

  16. [23]

    Synchromodulametry: A Hardware-First, Metric-Aware Measure- ment Interface for Multimessenger Coher- ence,

    T. Yawisit, “Synchromodulametry: A Hardware-First, Metric-Aware Measure- ment Interface for Multimessenger Coher- ence,”arXiv preprintarXiv:2512.20678, 2025. doi:10.48550/arXiv.2512.20678

    2025 doi

  17. [24]

    Reconfigurable Com- puting: A Survey of Systems and Software,

    K. Compton and S. Hauck, “Reconfigurable Com- puting: A Survey of Systems and Software,”ACM Computing Surveys, vol. 34, pp. 171–210, 2002. doi:10.1145/508352.508353

    2002

  18. [25]

    Kopetz,Real-Time Systems: Design Principles for Distributed Embedded Applications, Springer, 2011

    H. Kopetz,Real-Time Systems: Design Principles for Distributed Embedded Applications, Springer, 2011

    2011

  19. [26]

    The Tail at Scale,

    J. Dean and L. A. Barroso, “The Tail at Scale,” Communications of the ACM, vol. 56, no. 2, pp. 74– 80, 2013. doi:10.1145/2408776.2408794

    2013

  20. [28]

    Exact matrix com- pletion via convex optimization,

    E. J. Candès and B. Recht, “Exact matrix com- pletion via convex optimization,”Foundations of Computational Mathematics, vol. 9, pp. 717–772,

  21. [29]

    P. K. Varshney,Distributed Detection and Data Fu- sion, Springer, 1997. doi:10.1007/978-1-4757-2519-4

    1997 doi

  22. [30]

    The Fermi Gamma-ray Burst Monitor,

    C. A. Meeganet al., “The Fermi Gamma-ray Burst Monitor,”Astrophysical Journal, vol. 702, pp. 791– 804, 2009. doi:10.1088/0004-637X/702/1/791

    2009 doi

  23. [31]

    Spectra of Some Self-Exciting and Mutually Exciting Point Processes,

    A. G. Hawkes, “Spectra of Some Self-Exciting and Mutually Exciting Point Processes,” Biometrika, vol. 58, no. 1, pp. 83–90, 1971. doi:10.1093/biomet/58.1.83

    1971 doi

  24. [32]

    Accelerator-rich architectures: Opportunities and challenges,

    J. Conget al., “Accelerator-rich architectures: Opportunities and challenges,”Proceedings of the IEEE, vol. 104, no. 5, pp. 939–962, 2016. doi:10.1109/JPROC.2016.2526818 8

    2016

  25. [2009]

    doi:10.1007/s10208-009-9045-5

    doi

  26. [2017]

    doi:10.1140/epjc/s10052-017-4852-3

    doi

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.