Pith. sign in

REVIEW 3 major objections 5 minor 50 references

SGNAX: a unified matched-filter and excess-power pipeline for gravitational-wave detector characterization

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read SGNAX reproduces both standard glitch searches in one streaming pipeline and exposes a $\sqrt{2}$ SNR error in production features.

desk verdict A credible unified rewrite of snax/omicron with clean injections and an honest but not fully closed case that production snax over-amplifies low-frequency SNRs. read the letter →

arxiv 2608.09804 v1 pith:26QTASYO submitted 2026-08-10 astro-ph.IM

Zach Yarbrough , Olivia Godwin , Derek Davis , Gabriela González This is my paper · ORCID
classification astro-ph.IM
keywords gravitational-wavedetectorcharacterizationglitchtriggersmatchedfilterQ-transformsine-GaussiantemplatesstreamingdataflowpipelineSNRcalibrationauxiliarychannels
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

SGNAX is presented as a single Python-native streaming pipeline that carries both standard detector-characterization analyses: the multi-rate sine-Gaussian matched-filter search of snax and the multi-resolution Q-transform excess-power search of omicron. On 24 hours of archival detector strain, the Q-transform mode reproduces the omicron trigger population in the 10–100 Hz band with a trigger-rate ratio of 0.96, a Kolmogorov–Smirnov statistic of 0.02, and a median per-trigger SNR difference of −2.4%, while the dense feature-stream mode recovers the snax loud-feature population with 80% per-bin coincidence. The same injection campaign that calibrates SGNAX shows that the production snax deployment inflates reported SNRs below 25.6 Hz by factors of $\sqrt{2}$ to $2$, an error the paper localizes to a multiband amplitude gain rather than to the analysis statistic. If the claim is right, detector-characterization groups can run both analyses from one configuration and output schema, obtain Q-transform triggers at tens-of-seconds latency, and correct previously produced low-frequency feature SNRs.

What carries the argument

The load-bearing object is a single streaming dataflow graph whose per-channel branches share one data source and whitening stage, then diverge into either a multi-rate sine-Gaussian matched filter (quadrature correlation against a mismatch-spaced template bank, a per-window peak finder, and a cross-rate aggregator) or a Q-transform excess-power scan (bisquare-windowed tiles, a per-tile SNR estimator, cross-plane deduplication, and optional clustering). The shared front end is what makes per-channel cost fall as channels are added and what lets both algorithms see identical data and configuration in head-to-head comparisons. The second machinery element is the injection-calibrated scale reference: sine-Gaussian injections with analytically known optimal SNR pass through the full whiten-resample-correlate chain, establishing that SGNAX obeys the textbook matched-filter normalization and turning the residual against production snax into evidence of an amplitude error in that deployment rather than a difference in statistics.

What would settle it

Feed a sine-Gaussian of analytically known optimal SNR at about 20 Hz through the production snax chain and through SGNAX; if the production chain reports roughly the analytic SNR while SGNAX reports a factor of $\sqrt{2}$ lower, the scale error is confirmed, whereas if the two pipelines disagree differently across octave bands, the residual is in the comparison rather than in the production gain.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that a single streaming dataflow graph with a shared read-and-whitening front end can carry both the matched-filter and excess-power analyses without changing their statistics. The Q-transform branch reproduces the omicron trigger population on archival strain in the band most relevant to detector characterization—10–100 Hz with SNR above 8—at 96% of the reference trigger rate, with a 0.02 Kolmogorov–Smirnov statistic and a median per-trigger SNR 2.4% lower. The dense feature branch recovers the snax production loud-feature population with 80% per-bin coincidence and hourly rate correlation of 0.86. Injection calibration places the SGNAX SNR scale at the textbook matched-filter normalization, and the residual against production snax then localizes to a multiband amplitude error: rate bands below the production pipeline's 128 Hz resampling floor carry an extra gain of $\sqrt{2}$ in 12.8–25.6 Hz and of $2$ below 12.8 Hz, so absolute production SNRs below 25.6 Hz should be divided by $\sqrt{2}$ before comparison.

Load-bearing premise

The diagnosis that production snax over-amplifies low-frequency SNRs assumes that SGNAX's own whitening and normalization chain is exactly the textbook matched-filter scale, so the measured 0.705 SNR ratio in the 12.8–25.6 Hz band is entirely the other pipeline's error rather than a difference in whitening, resampling, or template-bank construction.

Editorial extensions

If this is right

  • Absolute SNRs recorded by production snax below 25.6 Hz carry the multiband amplitude excess and should be divided by $\sqrt{2}$ (and by $2$ below 12.8 Hz) before being compared or combined with SGNAX features.
  • The Q-transform mode can deliver triggers within tens of seconds from shared-memory or streaming data sources, closing the gap between the batch cadence of production omicron deployments and the needs of detector commissioning.
  • A single configuration object and hdf5 output schema now serve both the matched-filter and Q-transform analyses, so downstream detector-characterization tools can consume both from one layout instead of two dependency stacks.
  • Processing many channels as parallel branches of one graph lowers per-channel cost by about 2.8× for the Q-transform between one and 32 channels, and a full day of 16 kHz strain is analyzed in about 14 minutes.
  • For band-limited white-noise bursts, a configurable non-overlapping cluster-SNR estimator recovers a median 0.66 of the injected SNR versus 0.48 for the single loudest tile, without changing localized sine-Gaussian recovery.

Reading between the lines

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

  • The step-like $\sqrt{2}$ gain pattern at octave boundaries is a generic risk for any multi-rate whitened pipeline that resamples bands down to a common floor; an audit looking for SNR-ratio steps at resampling boundaries in other deployed pipelines would be a direct testable extension.
  • Because the scale error rescales SNRs without reordering them within a band, rank-based and threshold-based consumers are largely unaffected, but any absolute-SNR comparison crossing 25.6 Hz should show a systematic step; this predicts a discontinuity in production snax SNR histograms at exactly that frequency.
  • The high-frequency near-threshold residual above 2 kHz that the paper leaves open could be probed by running both modes with a common whitening filter; persistence of the residual would point to tiling or clustering rather than normalization.
  • Hosting both algorithms in one package turns the open problem of total-SNR recovery for broadband bursts into a per-pipeline configuration choice, which may accelerate a dedicated broadband estimator beyond the cluster-SNR improvement shown here.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper describes SGNAX, a Python-native pipeline built on the Stream Graph Navigator framework, which unifies the multi-rate sine-Gaussian matched filter of snax and the Q-transform excess-power search of omicron in a single dataflow graph. It claims a shared configuration object and output schema, multiple data-source backends for offline and online use, multi-channel processing with decreasing per-channel cost, matched-filter end-to-end latency of about five seconds, and a full day of 16 kHz strain analyzed in 14 minutes by the Q-transform mode. The validation includes injection-recovery tests for both modes, false-alarm/efficiency curves, comparisons with production omicron and snax outputs on 24 hours of LIGO data, runtime and latency benchmarks, and a streaming glitch-rate frame product. The most consequential claim is that the injection-calibrated reimplementation reveals a multiband amplitude error in production snax that inflates reported SNRs below 25.6 Hz by factors of sqrt(2) to 2, with the mechanism identified as incorrect octave-rate-band amplitude compensation.

Significance. If substantiated, SGNAX would be a valuable consolidation of two heavily used detector-characterization tools, and the open-source release with its test suite is a genuine reproducibility contribution. The injection-recovery studies, the transparent omicron comparison, the false-alarm/efficiency characterization, and the measured runtime and latency data are concrete strengths that go beyond a purely architectural paper. However, the headline few-percent agreement with omicron is restricted to a band selected after inspection, and the production snax amplitude-error claim depends on attributing a cross-pipeline SNR ratio to a production-side bug without a direct injection through the production chain. These load-bearing claims need additional support before the paper's central conclusions can be accepted at face value.

major comments (3)
  1. [Sec. VH, Fig. 12(b)] The conclusion that production snax over-amplifies its 12.8-25.6 Hz band by sqrt(2) rests on the measured median SNR ratio of 0.705 between sgnax-timeseries and production snax on coincident loud features, together with the assertion that sgnax's own whitening and normalization chain has the textbook matched-filter scale of Eqs. (1)-(3). That assertion is reported in text without a figure or table for the low-frequency calibration, and the paper explicitly states that a static audit of the production chain did not localize the excess. A constant 0.705 ratio is equally consistent with an under-amplification by sqrt(2) in sgnax's low-rate resampling or whitening path, or with a factor-of-two template-normalization difference, because any band-structured gain in either pipeline produces the same boundary-stepped signature. The load-bearing step is the attribution of the entire ratio to production snax, so the authors should add a direct injection test through the production snax chain, or an equivalent independent calibration of both pipelines, before presenting the production bug as established.
  2. [Sec. VF/VG and Table IV] The abstract and conclusion state that SGNAX reproduces the omicron trigger population to a few percent, but the few-percent metrics in Table IV are computed in the 10-100 Hz, SNR>8 window that was identified after a frequency-SNR sweep over the data, and the full-band comparison shows a 12% total deficit concentrated in near-threshold high-frequency tiles whose origin is left to future work. The few-percent claim should be explicitly scoped to that band, and the band-selection process should be described as exploratory rather than as the predetermined operational window. The high-frequency residual should either be included in the headline metrics or quantified separately in the same table, with a statement of whether it affects the detector-characterization conclusions for which the omicron comparison is offered as evidence.
  3. [Sec. VH and Table IV, snax column] The snax column of Table IV reports a trigger-rate ratio of 0.46 and an SNR-distribution KS statistic of 0.49, and the text attributes all of these residuals to the production amplitude scale. The per-channel breakdown of the 80% per-bin coincidence and the number of coincident loud features underlying the SNR and time medians should be reported, and the rate/KS metrics should be recomputed after applying the proposed 1/sqrt(2) correction to the production SNRs in the affected band. Without that demonstration, the statement that threshold- and rank-based consumers of the production features are unaffected by the scale error is not directly supported by the table.
minor comments (5)
  1. [Throughout] The typesetting has many missing spaces and ligature artifacts, for example 'We presentsgnax', 'a singlepython', 'gra vity spy', and 'a median of0.43'; a careful proofreading pass is needed before publication.
  2. [Abstract and Table V] The abstract says that a full day of 16 kHz strain is analyzed in 14 minutes, but Table V shows that this is the Q-transform figure; the matched-filter mode takes 84 minutes. Please qualify the abstract statement accordingly.
  3. [Sec. VG] The coincidence pairing for the omicron comparison uses a 0.2 s window and a factor-of-two frequency tolerance, but only the fraction of omicron triggers that find a partner is reported. The fraction of sgnax triggers that find an omicron partner should also be reported, since asymmetric pairing can bias the median per-trigger SNR difference.
  4. [Fig. 10] The bottom panel shows matched-filter SNR for sgnax-extract while the top two panels show excess-power SNR, and the caption notes that the shared color scale is only a morphological guide. This caveat should be stated more prominently in the text, since readers may otherwise over-read the cross-panel color comparison.
  5. [Sec. VH] The low-frequency calibration of sgnax is described only in prose; a table or figure showing the recovered-to-injected SNR ratio for the injections at 12.7 Hz and nearby frequencies, through the full whiten-resample-correlate path, would make the calibration auditable and would directly address the main concern raised in this report.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: the central comparisons are anchored to external references and to injection-calibrated SNR norms; the self-citations are infrastructural or comparative, not load-bearing.

full rationale

The claimed derivation chain is self-contained against external benchmarks. The omicron reproduction (Secs. VF and VG) compares sgnax-qtransform triggers with the production omicron trigger population from Robinet et al. [13] on the same archival L1 strain interval; the reported trigger-rate ratio of 0.96, KS statistic of 0.02, and median SNR difference of -2.4% are measured residuals, not parameters fitted into sgnax and then renamed as predictions. The snax comparison (Sec. VH) is a direct head-to-head feature-stream comparison with the production snax archive; the 80% per-bin coincidence and the median 0.70 SNR ratio are reported measurements. The inference that production snax over-amplifies the 12.8-25.6 Hz band by a factor of sqrt(2) is not circular: the sgnax SNR scale is independently calibrated with sine-Gaussian injections of analytically known optimal SNR, at central frequencies down to 12.7 Hz and through the full whiten-resample-correlate path, so the cross-pipeline ratio is a measurement against the textbook matched-filter normalization of Eqs. (1)-(3) rather than a quantity defined by the comparison. The proposed mechanism, octave rate-band compensation below the 128 Hz resampling floor, is an independent explanation of the observed band-stepped residual, not an input to it. The self-referential elements are citations to the sgn framework [22] and to snax [18,19] by overlapping authors; these supply infrastructure and the comparison target, but the central validation is anchored to externally established omicron triggers and to analytically known injections. No load-bearing claim reduces to a self-citation or to a fitted parameter renamed as a prediction, so no specific circular step is identified.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The analysis leans on standard matched-filter and Q-transform statistics from predecessor pipelines, on synthetic-noise injection calibration, and on a selected 10-100 Hz comparison band. No new physical entities are introduced. The main unproven input is that the sgnax SNR scale is textbook-correct, which makes the production snax residual a deployment bug.

free parameters (3)
  • SNR analysis threshold = 5.5
    Used for injection recovery, false-alarm, and production comparisons; chosen as a standard operating point, not fitted, but it defines what is called recoverable.
  • Injection coincidence window = 0.25 s
    Time window for matching injected to recovered triggers; a hand-chosen tolerance for the recovery fraction.
  • Omicron comparison band and threshold = 10-100 Hz, SNR > 8
    Selected after a frequency-SNR sweep as the window of closest agreement (Sec. VG); this selection drives the headline few-percent agreement with omicron.
assumptions (5)
  • domain assumption The sine-Gaussian matched-filter statistic of Eqs. (1)-(3) and the Q-transform excess-power statistic of Eqs. (5)-(7) are the correct SNR references.
    Adopted from snax and omicron; not re-derived in this paper.
  • domain assumption The synthetic stationary-Gaussian auxiliary channel is representative enough to calibrate the real whitening chain's SNR scale.
    Used in the Sec. V injection calibration; real channels are non-stationary, so the calibration may not transfer exactly.
  • ad hoc to paper The production snax SNR excess below 25.6 Hz is caused by octave-rate-band amplitude compensation, not by a difference in sgnax's whitening or bank.
    Sec. VH: the static audit did not localize the excess; the 1/sqrt2 step is inferred from the band structure of the residual.
  • domain assumption Omicron's production configuration as described by Robinet et al. is the correct baseline for Q-plane construction and normalization.
    The paper matches Q-range and mismatch to reproduce omicron's four Q-planes but does not independently verify omicron internals.
  • ad hoc to paper The chosen 10-100 Hz, SNR>8 comparison band is operationally representative rather than a favorable post-hoc subset.
    The full-band comparison shows a 12 percent total trigger deficit, so the headline agreement is band-dependent.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SGNAX: a unified matched-filter and excess-power pipeline for gravitational-wave detector characterization." pith.science (2026). https://pith.science/paper/26QTASYO

@misc{pith2026260809804,
  author       = {Pith},
  title        = {Pith review of: SGNAX: a unified matched-filter and excess-power pipeline for gravitational-wave detector characterization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/26QTASYO}},
  note         = {Machine review of arXiv:2608.09804}
}
abstract

We present SGNAX, an open-source pipeline for gravitational-wave detector characterization that delivers matched-filter and excess-power transient triggers from a single streaming dataflow graph. Built on the Stream Graph Navigator (sgn) framework, SGNAX unifies the multi-rate sine-Gaussian matched-filter search of snax and the multi-resolution Q-transform search of omicron in one Python-native package. Auxiliary channels from an interferometer are analyzed as parallel branches sharing data-read and whitening stages, reducing per-channel processing cost as channels are added. A full day of 16 kHz strain is analyzed in 14 minutes, and Q-transform processing is 2.8 times more efficient per channel at 32 channels than at one. Matched-filter correlations use PyTorch and run on CPU or GPU. Data sources include offline frame caches, shared-memory buffers, and the arrakis distribution service, with the same configuration supporting offline and online operation. The matched filter delivers triggers at about five seconds end-to-end latency, while the Q-transform operates at latencies of tens of seconds. Injection campaigns recover 99.4% of recoverable sine-Gaussian injections with parameters within the expected template mismatch and show broadband white-noise-burst recovery consistent with established event-trigger generators. On 24 hours of archival LIGO strain, SGNAX reproduces the omicron trigger population at 10--100 Hz, with trigger rates and SNRs agreeing to a few percent. On production auxiliary channels, it recovers the snax loud-feature population with 80% per-bin coincidence. The injection-calibrated reimplementation also reveals a multiband amplitude error in production snax that inflates reported SNRs below 25.6 Hz by factors of $\sqrt{2}$--2, for which we identify the mechanism and correction.

Figures

Figures reproduced from arXiv: 2608.09804 by the authors.

Figure 1
Figure 1. FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. The [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. The [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: FIG. 4. Injection recovery for the [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5. Injection recovery for the [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Head-to-head comparison of the two [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. Recovered versus injected SNR for [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]
Figure 9
Figure 9. Figure 9: FIG. 9. Detection efficiency versus false-alarm rate for the [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]
Figure 10
Figure 10. Figure 10: FIG. 10. Twenty-four hours of triggers on [PITH_FULL_IMAGE:figures/full_fig_p015_10.png]
Figure 11
Figure 11. Figure 11: FIG. 11. Per-minute glitch rate on [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]
Figure 12
Figure 12. Figure 12: FIG. 12. The [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]
Figure 13
Figure 13. Figure 13: FIG. 13. The streaming glitch-rate frame channel written by [PITH_FULL_IMAGE:figures/full_fig_p020_13.png]
Figure 14
Figure 14. Figure 14: FIG. 14. Multi-channel throughput of the two [PITH_FULL_IMAGE:figures/full_fig_p021_14.png]
Figure 15
Figure 15. Figure 15: FIG. 15. Cumulative distribution of the measured end-to-end [PITH_FULL_IMAGE:figures/full_fig_p021_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 27 canonical work pages

  1. [1]

    Thesnaxandomicronimplementations Thesnaxpipeline is implemented as apython layer overgstreamer-1.0 [18, 21]. Each in- vocation of thesnax_extractcommand builds a gstreamergraph in which alalcachesrcor live- data source feeds aframecppchanneldemux, then per- channelwhiten,audioresample, andlal_matrixmixer elements distribute samples across rate-specific fi...

    work page

  2. [2]

    Package, dependencies, and layout sgnaxis provided as thepythonpackagesgnax, available at [44]. The package requirespython3.11 or later and depends on the foursgnecosystem packages [22, 37, 38], onnumpy,scipy,torch[23], h5py,gwpy[45],lalsuite[46],igwn-ligolw[47], igwn-segments,ligo-scald, thearrakisclient library,gpstime, andpyyaml; DAG generation additio...

    work page

  3. [3]

    Static analysis is provided bymypyin strict mode, including check_untyped_defs, and byruffwith a broad rule set covering security, complexity, and import hygiene

    Static analysis and test suite The package enforces a suite of static and dynamic checks intended to catch regressions in either algorith- mic correctness or data-source configuration. Static analysis is provided bymypyin strict mode, including check_untyped_defs, and byruffwith a broad rule set covering security, complexity, and import hygiene. Dy- namic...

    work page

  4. [4]

    Aasiet al., Advanced LIGO, Class

    J. Aasiet al., Advanced LIGO, Class. Quant. Grav.32, 074001 (2015), 1411.4547

    arXiv 2015

  5. [5]

    Acerneseet al., Advanced Virgo: a second- generation interferometric gravitational wave detector, Class

    F. Acerneseet al., Advanced Virgo: a second- generation interferometric gravitational wave detector, Class. Quant. Grav.32, 024001 (2015), 1408.3978

    arXiv 2015

  6. [6]

    KAGRA Collaboration, KAGRA: 2.5 generation interfer- ometric gravitational wave detector, Nature Astron.3, 35 (2019), 1811.08079

    arXiv 2019

  7. [7]

    B. P. Abbottet al., Observation of gravitational waves from a binary black hole merger, Phys. Rev. Lett.116, 061102 (2016), 1602.03837

    arXiv 2016

  8. [8]

    Abbottet al., GWTC-3: Compact Binary Coales- cences Observed by LIGO and Virgo during the Sec- ond Part of the Third Observing Run, Phys

    R. Abbottet al., GWTC-3: Compact Binary Coales- cences Observed by LIGO and Virgo during the Sec- ond Part of the Third Observing Run, Phys. Rev. X13, 041039 (2023), 2111.03606

    arXiv 2023

Show all 50 references
  1. [9]

    A. Abacet al., GWTC-4.0: Updating the Gravitational- WaveTransientCatalogwithObservationsfromtheFirst Part of the Fourth LIGO-Virgo-KAGRA Observing Run (2025), arXiv:2508.18082 [gr-qc]

    2025 arXiv

  2. [10]

    A. Abacet al., GWTC-5.0: Observations from the Sec- ond Part of the Fourth LIGO-Virgo-KAGRA Observing Run and Updates to the Gravitational-Wave Transient Catalog (2026), arXiv:2605.27225 [gr-qc]

    2026 arXiv

  3. [11]

    B. P. Abbottet al., Characterization of transient noise in Advanced LIGO relevant to gravitational wave sig- nal GW150914, Class. Quant. Grav.33, 134001 (2016), 1602.03844

    2016 arXiv

  4. [12]

    Daviset al., LIGO detector characterization in the second and third observing runs, Class

    D. Daviset al., LIGO detector characterization in the second and third observing runs, Class. Quant. Grav.38, 135014 (2021), 2101.11673

    2021 arXiv

  5. [13]

    Soniet al., LIGO Detector Characterization in the first half of the fourth Observing run, Class

    S. Soniet al., LIGO Detector Characterization in the first half of the fourth Observing run, Class. Quant. Grav.42, 085016 (2025), 2409.02831

    2025

  6. [14]

    Glanzer, A

    J. Glanzer, A. F. Helmling-Cornell,et al., LIGO Detector Characterization in the Second and Third Parts of the Fourth Observing Run, LIGO Document P2600357, in preparation (2026)

    2026

  7. [15]

    McIver, T

    J. McIver, T. J. Massinger, F. Robinet, J. R. Smith, and M. Walker, Diagnostic methods for gravitational-wave detectors, inAdvanced Interferometric Gravitational- wave Detectors, Vol. 1, edited by D. Reitze, P. Saulson, and H. Grote (World Scientific, 2019) pp. 373–392

    2019

  8. [16]

    Robinet, N

    F. Robinet, N. Arnaud, N. Leroy, A. Lundgren, D. Macleod, and J. McIver, Omicron: A tool to char- acterize transient noise in gravitational-wave detectors, SoftwareX12, 100620 (2020), 2007.11374

    2020 arXiv

  9. [17]

    Soniet al., Reducing scattered light in LIGO’s third observing run, Class

    S. Soniet al., Reducing scattered light in LIGO’s third observing run, Class. Quant. Grav.38, 025016 (2021), 2007.14876

    2021 arXiv

  10. [18]

    Robinet, Omicron: an algorithm to detect and char- acterize transient events in gravitational-wave detectors, Virgo Technical Document VIR-0545C-14 (2018)

    F. Robinet, Omicron: an algorithm to detect and char- acterize transient events in gravitational-wave detectors, Virgo Technical Document VIR-0545C-14 (2018)

    2018

  11. [19]

    Robinetet al., Omicron: a Q-transform analysis tool for gravitational-wave detector data,https://git.ligo

    F. Robinetet al., Omicron: a Q-transform analysis tool for gravitational-wave detector data,https://git.ligo. org/virgo/virgoapp/Omicron(2025)

    2025

  12. [20]

    J. C. Brown, Calculation of a constant Q spectral trans- form, J. Acoust. Soc. Am.89, 425 (1991)

    1991

  13. [21]

    Godwinet al., snax: Stream-based Noise Acquisi- tion and eXtraction,https://git.ligo.org/snax/snax (2022)

    P. Godwinet al., snax: Stream-based Noise Acquisi- tion and eXtraction,https://git.ligo.org/snax/snax (2022)

    2022

  14. [22]

    Godwin,Low-latency Statistical Data Quality in the Era of Multi-Messenger Astronomy, Ph.D

    P. Godwin,Low-latency Statistical Data Quality in the Era of Multi-Messenger Astronomy, Ph.D. thesis, The Pennsylvania State University (2020). 25

    2020

  15. [23]

    Messicket al., Analysis Framework for the Prompt Discovery of Compact Binary Mergers in Gravitational- wave Data, Phys

    C. Messicket al., Analysis Framework for the Prompt Discovery of Compact Binary Mergers in Gravitational- wave Data, Phys. Rev. D95, 042001 (2017), 1604.04324

    2017 arXiv

  16. [24]

    Cannonet al., GstLAL: A software framework for gravitational wave discovery, SoftwareX14, 100680 (2021), 2010.05082

    K. Cannonet al., GstLAL: A software framework for gravitational wave discovery, SoftwareX14, 100680 (2021), 2010.05082

    2021 arXiv

  17. [25]

    Huang, O

    Y.-J. Huang, O. Godwin, C. Hanna, J. Kennington, J. Rollins, M. Melching, N. E. Sovitzky, A. Viets, M. Wade, Z. Yarbrough, Y.-K. Chu, W. W. Phillips, S. Sachdev, and R. Udall, SGN: A python framework for stream-processing pipelines (2026), arXiv:2607.03575 [astro-ph.IM]

    2026 arXiv

  18. [26]

    Paszkeet al., PyTorch: An Imperative Style, High- Performance Deep Learning Library, inAdvances in Neu- ral Information Processing Systems 32(2019) pp

    A. Paszkeet al., PyTorch: An Imperative Style, High- Performance Deep Learning Library, inAdvances in Neu- ral Information Processing Systems 32(2019) pp. 8024– 8035

    2019

  19. [27]

    ArrakisDevelopers,arrakis-python: Pythonclientforthe Arrakis low-latency timeseries data distribution service, https://git.ligo.org/ngdd/arrakis-python(2024)

    2024

  20. [28]

    Macleod and A

    D. Macleod and A. Urban, pyomicron: utilities for run- ning and post-processing Omicron analyses,https:// github.com/gwpy/pyomicron(2019)

    2019

  21. [29]

    McIver,The impact of terrestrial noise on the de- tectability and reconstruction of gravitational wave sig- nals from core-collapse supernovae, Ph.D

    J. McIver,The impact of terrestrial noise on the de- tectability and reconstruction of gravitational wave sig- nals from core-collapse supernovae, Ph.D. thesis, Univer- sity of Massachusetts Amherst (2015)

    2015

  22. [30]

    S. K. Chatterji,The search for gravitational wave bursts in data from the second LIGO science run, Ph.D. thesis, Massachusetts Institute of Technology (2005)

    2005

  23. [31]

    Chatterji, L

    S. Chatterji, L. Blackburn, G. Martin, and E. Kat- savounidis, Multiresolution techniques for the detection of gravitational-wave bursts, Class. Quant. Grav.21, S1809 (2004), gr-qc/0412119

    2004 arXiv

  24. [32]

    Zevinet al., Gravity Spy: integrating Advanced LIGO detector characterization, machine learning, and citizen science, Class

    M. Zevinet al., Gravity Spy: integrating Advanced LIGO detector characterization, machine learning, and citizen science, Class. Quant. Grav.34, 064003 (2017), 1611.04596

    2017 arXiv

  25. [33]

    Glanzeret al., Data quality up to the third observing run of advanced LIGO: Gravity Spy glitch classifications, Class

    J. Glanzeret al., Data quality up to the third observing run of advanced LIGO: Gravity Spy glitch classifications, Class. Quant. Grav.40, 065004 (2023), 2208.12849

    2023 arXiv

  26. [34]

    J. R. Smith, T. Abbott, E. Hirose, N. Leroy, D. Macleod, J. McIver, P. Saulson, and P. Shawhan, A hierarchi- cal method for vetoing noise transients in gravitational- wave detectors, Class. Quant. Grav.28, 235005 (2011), 1107.2948

    2011 arXiv

  27. [35]

    Biswaset al., Application of machine learning algo- rithms to the study of noise artifacts in gravitational- wave data, Phys

    R. Biswaset al., Application of machine learning algo- rithms to the study of noise artifacts in gravitational- wave data, Phys. Rev. D88, 062003 (2013), 1303.6984

    2013 arXiv

  28. [36]

    Davis, Z

    D. Davis, Z. Yarbrough, J. Areeda, R. Macas, N. Arnaud, A. Helmling-Cornell, P. Doliva, O. Godwin, H. Yuzuri- hara, B. Mannix,et al., Rapid data quality investigations of gravitational-wave events with the Data Quality Re- port Builder toolkit (2026), arXiv:2605.16183

    2026 arXiv

  29. [37]

    Brun and F

    R. Brun and F. Rademakers, ROOT: an object-oriented data analysis framework, Nucl. Instrum. Meth. A389, 81 (1997)

    1997

  30. [38]

    Robinet, GWOLLUM: a toolkit for analysis of gravitational-wave detector data,https://git.ligo

    F. Robinet, GWOLLUM: a toolkit for analysis of gravitational-wave detector data,https://git.ligo. org/virgo/virgoapp/GWOLLUM(2025)

    2025

  31. [39]

    Frigo and S

    M. Frigo and S. G. Johnson, The Design and Implemen- tation of FFTW3, Proc. IEEE93, 216 (2005)

    2005

  32. [40]

    SGN Development Team, sgn-ligo: LIGO-specific ele- ments for the SGN framework,https://git.ligo.org/ greg/sgn-ligo(2026)

    2026

  33. [41]

    ligo.org/ngdd/sgn-arrakis(2026)

    SGN Development Team, sgn-arrakis: Arrakis data- source elements for the SGN framework,https://git. ligo.org/ngdd/sgn-arrakis(2026)

    2026

  34. [42]

    Essick, P

    R. Essick, P. Godwin, C. Hanna, L. Blackburn, and E. Katsavounidis, iDQ: Statistical inference of non- Gaussian noise with auxiliary degrees of freedom in gravitational-wave detectors, Mach. Learn.: Sci. Technol. 2, 015004 (2021)

    2021

  35. [43]

    W. G. Anderson, P. R. Brady, J. D. E. Creighton, and E. E. Flanagan, An excess power statistic for detection of burst sources of gravitational radiation, Phys. Rev. D 63, 042003 (2001), gr-qc/0008066

    2001 arXiv

  36. [44]

    Godwin, gvt: a Glitch Validation Toolkit,https:// git.ligo.org/patrick.godwin/gvt(2024)

    P. Godwin, gvt: a Glitch Validation Toolkit,https:// git.ligo.org/patrick.godwin/gvt(2024)

    2024

  37. [45]

    Huxford, R

    R. Huxford, R. George, M. Trevor, Z. Yarbrough, and P. Godwin, Performance of iDQ ahead of LIGO, Virgo, and KAGRA’s fourth observing run (2024), arXiv:2412.04638 [gr-qc]

    2024 arXiv

  38. [46]

    SGN Development Team, sgn-gwframe: gravitational- wave frame file I/O elements for the SGN framework, https://git.ligo.org/greg/sgn-gwframe(2026)

    2026

  39. [47]

    Godwin and Z

    O. Godwin and Z. Yarbrough, sgnax: a multi-channel auxiliary burst-search pipeline,https://git.ligo.org/ detchar/sgn-dq/sgnax(2026)

    2026

  40. [48]

    D. M. Macleod, J. S. Areeda, S. B. Coughlin, T. J. Massinger, and A. L. Urban, GWpy: A Python pack- age for gravitational-wave astrophysics, SoftwareX13, 100657 (2021)

    2021

  41. [49]

    LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration, LVK Algorithm Library – LAL- Suite, Free software (GPL) (2018)

    2018

  42. [50]

    org/computing/software/igwn-ligolw(2024)

    IGWN Software Working Group, igwn-ligolw: LIGO Lightweight XML I/O for Python,https://git.ligo. org/computing/software/igwn-ligolw(2024)

    2024

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.