Pith. sign in

REVIEW 4 major objections 5 minor 43 references

Uncovering deterministic behavior of black hole IGR J17091-3624: A twin of GRS 1915+105

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The black hole X-ray binary IGR J17091-3624, previously classified as purely stochastic, shows four temporal classes that are deterministic after removing Poisson noise, strengthening its status as a dynamical twin of GRS 1915+105.

desk verdict Careful multi-method analysis suggests determinism in IGR J17091-3624 after denoising, but the missing synthetic ground-truth control leaves the central claim unproven. read the letter →

arxiv 2411.17810 v2 pith:WHLLC6LS submitted 2024-11-26 astro-ph.HE

Anindya Guria , Banibrata Mukhopadhyay This is my paper · ORCID
classification astro-ph.HE PACS 97.60.Lf95.75.Wx05.45.-a
keywords blackholeX-raybinaryIGRJ17091-3624GRS1915+105Poissonnoisedenoisingcorrelationintegralsingularvaluedecompositionprincipalcomponentanalysis
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

This paper argues that IGR J17091-3624, long thought to be an outlier because its X-ray flickering looked purely random, actually hides deterministic dynamics. The apparent stochasticity, the authors claim, is an artifact of Poisson photon noise swamping a faint source. Applying four different denoising procedures (moving average, Gaussian, adaptive polynomial, and non-local means) and then testing with three independent statistical probes, they find that four of the nine temporal classes are non-stochastic after cleaning. That would make IGR J17091-3624 a dynamical twin of GRS 1915+105, and it would mean the earlier stochastic-only classification needs to be revised.

What carries the argument

The argument runs on the conjunction of four denoising filters and three classification tests. The filters are the boxcar (moving average), Gaussian convolution, an adaptive denoising algorithm (ADA) that fits overlapping low-order polynomials to windows of the lightcurve, and non-local means (NLM), which averages similar short patches wherever they occur in time. The tests are the Grassberger-Procaccia correlation integral with surrogate analysis (quantified by the normalized mean sigma deviation, nmsd, with nmsd > 3 rejecting the stochastic null), singular value decomposition of the delay-embedded data matrix read through Betti numbers (β0 + β1 > 1 means structure), and principal component analysis of the eigenvalue-ratio curve clustered with DBSCAN. Agreement among at least three of the four filters is required to label a class stochastic or non-stochastic.

What would settle it

Apply the same four filters to synthetic lightcurves generated from a stationary stochastic process with the same length, mean count, and RMS as each IGR class, then run the same CI/surrogate, SVD, and PCA pipeline; if these pure-noise controls are classified as non-stochastic, the denoising is imprinting determinism and the paper's conclusion fails.

Watch

Extended reading notes

Core claim

The central claim is that temporal classes IV, V, VII, and VIII of IGR J17091-3624 are deterministic (non-stochastic) once Poisson noise is removed; classes I, II, III, and IX remain stochastic, and class VI is ambiguous. The authors show that in every unfiltered lightcurve the correlation-integral dimension matches the embedding dimension and the surrogate test says 'stochastic', reproducing the earlier conclusion. After any of the four filters, the same four classes produce correlation dimensions that saturate below the embedding dimension, SVD phase portraits with multiple topological features, and PCA eigenvalue-ratio outliers, with all three tests agreeing, while the stochastic classes stay stochastic. They conclude that the source hosts a complex underlying dynamical system similar to GRS 1915+105, and that prior findings of pure stochasticity were noise artifacts.

Load-bearing premise

The whole classification rests on the assumption that the denoising filters remove Poisson noise without creating artificial low-dimensional structure in a signal that is in fact purely random.

Editorial extensions

If this is right

  • IGR J17091-3624 displays at least as much dynamical complexity as GRS 1915+105, so the two sources can be studied as the same physical class of accreting black holes.
  • The four non-stochastic classes (IV, V, VII, VIII) can now be mapped onto specific accretion-flow states: a Keplerian disk for VII and VIII, a transitional advection-dominated flow for IV, and a mixed disk/advection flow for V.
  • The stochastic classes I, II, III, and IX all correspond to power-law-dominated, general advective accretion flow states, tightening the coupling between spectral state and temporal dynamics.
  • Poisson-noise removal should become a standard pre-processing step before non-linear analysis of any faint X-ray source, not just this object.

Reading between the lines

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

  • The same pipeline could be applied to other faint X-ray binaries whose 'stochastic' classifications come from noise-dominated RXTE or NICER lightcurves, potentially overturning several published results.
  • Because the NLM filter preserves stochasticity when the signal is genuinely random (the S classes stay S), the method itself carries a built-in control; a natural next test is to benchmark all four filters on simulated chaotic signals embedded in Poisson noise to quantify detection thresholds.
  • If the determinism is real, the underlying dynamical system should be identifiable in higher-cadence future observations, for example in the phase-space topology of the attractor or in dimension estimates that match one of the GRS 1915+105 classes.
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

4 major / 5 minor

Summary. The paper analyzes nine RXTE (and one Chandra) light curves of IGR J17091–3624, applies four denoising filters (BOX, GAU, ADA, NLM), and uses three classification diagnostics (correlation integral with IAAFT surrogates, SVD with Betti numbers, and PCA with DBSCAN) to distinguish stochastic from non-stochastic behavior. The authors report that temporal classes IV, V, VII, and VIII, which were previously classified as stochastic by Adegoke et al. (2020), become non-stochastic after denoising under a majority rule over methods. They interpret this as evidence that IGR J17091–3624 is dynamically similar to GRS 1915+105 and that the earlier stochastic conclusion was an artifact of Poisson noise contamination.

Significance. If the central claim were established, it would overturn a published classification result and materially strengthen the twin-source interpretation of IGR J17091–3624 and GRS 1915+105. The paper also contributes a practical multi-filter, multi-diagnostic pipeline and is transparent about its parameter choices and about disagreements among methods; the use of public archival data and the reference to the NoLiTSA codebase are helpful for reproducibility. However, the claim is currently conditioned on an unvalidated assumption about the denoising filters and on test-specific threshold choices, so the significance is high but not yet secure.

major comments (4)
  1. [Secs. 3.1–3.3, Tables 3 and 6] Section 3 states as a design criterion that the filters should not introduce additional local correlation, yet the convolution filters in Eqs. (1)–(5) are explicitly low-pass smoothers and ADA fits overlapping local polynomials (Eq. 6). No synthetic ground-truth control is reported: Poisson-only realizations matched to the observed count rates and lengths are never passed through the four filters and then through the CI/IAAFT, SVD/Betti, and PCA/DBSCAN pipeline. Because these operations can convert white noise into a locally smooth, correlated series, the central claim that classes IV, V, VII, and VIII are non-stochastic after denoising is not yet supported. The authors should add false-positive rates for each filter–test combination on filtered pure noise and on known chaotic and stochastic benchmark systems.
  2. [Sec. 4.1.1, Eq. (12)] The IAAFT surrogate test is applied to the filtered series rather than to the original observation, and for NLM the filtered series is constructed from patch redundancy with a sensitivity parameter h tuned per lightcurve (Eq. 9, Fig. 1). IAAFT surrogates preserve the power spectrum and amplitude distribution of this already-filtered series, but they do not preserve the patch-redundancy structure that NLM specifically builds on; hence nmsd > 3 can reject the null for stochastic input if the filter imprints such structure. The Table 3 classifications therefore need to be accompanied by a surrogate or synthetic test applied to the unfiltered data, or by a surrogate scheme consistent with the filtering model.
  3. [Sec. 5.3, Fig. 12] The PCA/DBSCAN hyperparameters r_cut = 124 and epsilon = 40 are selected by maximizing the silhouette score on the very set of 45 signals that is subsequently labeled S/NS. This makes the Figure 12 split partly a product of thresholds fitted to the target data rather than an independent classification. With only nine source classes and multiple filtering variants, the reported silhouette score of 0.693 is not, by itself, evidence of robustness. A synthetic benchmark, cross-validation, or thresholds fixed a priori would be needed to show that the S/NS separation is not an artifact of threshold optimization.
  4. [Secs. 5.1 and 6, Tables 3 and 6] The statement in Sec. 5.1 that classes IV, V, VII, and VIII are determined to be NS 'unanimously by our filtering methods' is contradicted by the paper's own Table 6, where SVD gives S/NS for class IV and where class VI is listed as S/NS overall. The final assignment of IV = NS in Sec. 6 is justified by the NLM SVD portrait, even though the stated formal criterion ('if three or more methods agree') is not met. The headline claim should be revised to distinguish robustly NS classes from ambiguous ones, and the internal inconsistency between the narrative and Table 6 should be resolved.
minor comments (5)
  1. [Sec. 3.1.1] The sentence 'We move on to more involved methods that address these issues. explicitly preserved during boxcar denoising.' appears unfinished; please revise it.
  2. [Secs. 1 and 4.1] The source name is misspelled as 'IGR J17091–362' in two places; it should be 'IGR J17091–3624'.
  3. [Sec. 4.3] The heading uses 'Principle Component Analysis'; the standard term is 'Principal Component Analysis'.
  4. [Table 3 and Eq. (12)] The manuscript does not state the embedding dimension range Mmin–Mmax, the delay used for each class, or the number of surrogates per lightcurve; these values are needed to interpret the nmsd statistics and D2 saturations.
  5. [Secs. 3 and 4] For full reproducibility, the authors should provide the custom code for the filters and for the Betti-number and DBSCAN classification, rather than only citing the NoLiTSA package.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the S/NS classification is an empirical analysis whose labels are not defined in terms of the filters' outputs, and the self-citations are background, not load-bearing.

full rationale

The paper's central claim is that after applying four denoising filters, CI/surrogate, SVD, and PCA tests label IGR classes IV, V, VII, and VIII as non-stochastic. This is an inference from data, not a derivation, and I find no step where a quantity is defined in terms of the target result or where a fitted parameter is renamed as a prediction. The CI test uses IAAFT surrogates generated from each filtered series and rejects the null only if nmsd > 3; this is a conventional statistical test, and the fact that classes I, II, III, and IX remain S under all filters provides an internal control that the filters do not automatically imprint NS structure. SVD uses FNN to choose embedding dimension independently of CI, and PCA/DBSCAN is an unsupervised clustering method; the rcut = 124 and epsilon = 40 are tuned via silhouette score on the same data, which is an in-sample validation limitation but not a definitional circularity. The paper explicitly acknowledges missing ground truth ('we do not have simulated lightcurves...', 'ideally some model of the signal is essential'), and this is a validation gap rather than a circular reduction. Self-citations (Mukhopadhyay 2004; Misra et al. 2004; Adegoke et al. 2018, 2020) supply background and the spectral-state interpretation in Table 7, but the S/NS detection itself rests on the present data and tests, so the self-citations are not load-bearing. No equation in the paper equates an output to an input by construction, and no prior 'uniqueness theorem' is invoked to force a choice. The strongest criticism is methodological: denoising may imprint smoothness on stochastic inputs, but that is a falsifiability/robustness concern, not circularity by the standard used here.

Assumptions & free parameters 8 free parameters · 6 assumptions · 0 invented entities

The central claim rests on per-lightcurve tuning of denoising parameters, silhouette-optimized clustering hyperparameters, and the unvalidated assumption that denoising does not manufacture determinism. No new physical entities are introduced.

free parameters (8)
  • NLM sensitivity factor h = per-lightcurve value at lower knee of RMS-h curve
    Controls the NLM decay parameter; chosen from RMS vs h S-curve for each light curve (Section 3.3, Figure 1). Classification of NS classes depends on it.
  • ADA segment size w = ~15-20 per lightcurve
    Chosen as the segment size at the least slope of the RMS deviation plot (Section 3.2).
  • ADA polynomial order k = 5
    Chosen by hand to maximize polynomial fit closeness; authors note k>3 does not change RMS much (Section 3.2).
  • BOX kernel length 2k+1 = 9
    Selected as shortest kernel that still improves SNR (Section 3.1.1, Table 2).
  • GAU sigma and kernel length = sigma=3, kernel length 9
    Chosen to match boxcar kernel length, with spread sigma=3 (Section 3.1.2).
  • PCA eigenvalue-ratio cutoff r_cut = 124
    Optimized with DBSCAN epsilon via silhouette score over the observed signals (Section 5.3). Affects which signals are split into ER values and therefore the S/NS classification.
  • DBSCAN neighborhood epsilon = 40
    Tuned together with r_cut to maximize silhouette score; controls whether points are clustered as S or labeled NS outliers (Section 5.3).
  • PCA minimum segment size = 10
    Set following Chakka and Sinha (2024); stops recursive splitting of the lightcurve (Section 4.3).
assumptions (6)
  • standard math Delay embedding reconstructs the underlying dynamics from a scalar time series (Takens embedding theorem).
    Used implicitly in the correlation integral method (Section 4.1).
  • domain assumption The first minimum of mutual information gives a valid delay for embedding.
    Standard heuristic; chosen following Fraser and Swinney (1986) (Section 4.1).
  • domain assumption IAAFT surrogates preserve the null hypothesis of a stationary linear stochastic process, and NMSD > 3 rejects it.
    Surrogate analysis relies on this to distinguish deterministic from linearly correlated noise (Section 4.1.1).
  • ad hoc to paper Denoising filters (BOX, GAU, ADA, NLM) remove Poisson noise without creating spurious deterministic structure.
    The central conclusion depends on this; no synthetic-data test is provided to validate it (Sections 3, 5, and 6).
  • domain assumption The temporal classes of IGR J17091-3624 (Altamirano et al. 2011) and the spectral states used from Adegoke et al. (2020) are correct.
    The analysis groups data by these pre-existing classes and attaches accretion-state interpretations to them (Sections 1 and 6, Table 7).
  • domain assumption Poisson noise is the dominant noise component and can be estimated as sqrt of mean photon count.
    Used to set NLM parameter h and to argue that noise is the reason previous studies saw stochasticity (Table 1, Section 3.3).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Uncovering deterministic behavior of black hole IGR J17091-3624: A twin of GRS 1915+105." pith.science (2026). https://pith.science/paper/WHLLC6LS

@misc{pith2026241117810,
  author       = {Pith},
  title        = {Pith review of: Uncovering deterministic behavior of black hole IGR J17091-3624: A twin of GRS 1915+105},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WHLLC6LS}},
  note         = {Machine review of arXiv:2411.17810}
}
read the original abstract

Understanding nonlinear properties in accreting systems, particularly for black holes, from observation is illuminating as they are expected to be general relativistic magnetohydrodynamic flows that are nonlinear. Two features associated with nonlinear systems, used commonly, are chaos, which is deterministic, and random, which is stochastic. The differentiation between chaotic and stochastic systems is often considered to quantify the nonlinear properties of an astrophysical system. The particular emphasis is that data is often noise-contaminated and finite. We examine the dual nature of the black hole X-ray binary IGR J17091-3624, whose behavior has been closely studied in parallel to GRS 1915+105. Certain similarities in the temporal classes of these two objects are explored in literature. However, this has not been the case with their non-linear dynamics: GRS 1915+105 shows signs of determinism and stochasticity both, while IGR J17091-3624 was found to be predominantly stochastic. Here, we confront the inherent challenge of noise contamination faced by previous studies, particularly Poisson noise, which adversely impacts the reliability of non-linear results. We employ several denoising techniques to mitigate noise effects and employ methods like Principal Component Analysis, Singular Value Decomposition, and Correlation Integral to isolate the deterministic signatures. We have found signs of determinism in IGR J17091-3624, thus supporting the hypothesis of it being similar to GRS 1915+105, even as a dynamical system. Our findings not only shed light on the complex nature of IGR J17091-3624 but also pave the way for future research employing noise-reduction techniques to analyze non-linearity in observed dynamical systems.

Figures

Figures reproduced from arXiv: 2411.17810 by the authors.

Figure 1
Figure 1. The variation of RMS with increasing sensitivity proportionality constant. This helps to choose the optimal sensitivity h for NLM denoising of IGR lightcurves by finding the lower “knee”. quencies are only scaled down, while the lower frequency dynamics are preserved as it is. Since ideal Poisson noise has a uniform PSD, NLM’s effect in the frequency do￾main is, therefore, promising for mitigating this issue. This i… view at source ↗
Figure 2
Figure 2. Pre and post filtering lightcurves of IGR-VIII [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Power spectral density (PSD) of IGR-VIII pre and post denoising [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Power spectral density (PSD) of IGR-II pre and post denoising [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Comparison of IGR-V lightcurves (original and NLM) with its GRS-1915+105 counterpart µ. for denoising, which is ineffective in the case of S sig￾nals. This property works in our favor as it ensures that the NLM method will not “over-smoothen” an S signal, which is a po…
Figure 6
Figure 6. Figure 6: D2 vs. M plot for for all classes without any denoising. The red dotted curves are the surrogates, and the dark blue solid curve is the true data. The light blue straight line shows D2 = M, the expected result for an ideal S signal. As outlined in subsection 4.3, the P…
Figure 7
Figure 7. Figure 7: D2 vs. M plot for all classes after NLM denoising. The red dotted curves are the surrogates, and the dark blue solid curve is the true data. The light blue straight line shows D2 = M, the expected result for an ideal S signal. sistently NS by all denoising methods, whi…
Figure 8
Figure 8. Figure 8: SVD plots of IGR-IV original data, and after NLM denoising. The optimal embedding dimensions and delay are mentioned for each of them [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: SVD plots of IGR-VII original data, and after ADA denoising. The optimal embedding dimensions and delay are mentioned for each of them [PITH_FULL_IMAGE:figures/full_fig_p018_9.png]
Figure 10
Figure 10. Figure 10: SVD plots of IGR-II original data, and after ADA denoising. The optimal embedding dimensions and delay are mentioned for each of them [PITH_FULL_IMAGE:figures/full_fig_p019_10.png]
Figure 11
Figure 11. Figure 11: An example of a typical ER curve (IGR VIII post NLM filtering) [PITH_FULL_IMAGE:figures/full_fig_p021_11.png]
Figure 12
Figure 12. Figure 12: Scatter plot of the PCA parameters for all sig￾nals (filtered and unfiltered). Red points are the S timeseries, while green points are NS. such as CI and SVD. The PCA and SVD tests have been improved upon by making them more objective. Our findings are summarized in …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 14 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter doi edition editor eprint howpublished institution journal key month number organization pages publisher school series title misctitle type volume year version url label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts ...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION format.url url empty "" new.block "" url * "" * if FUNCTION format.eprint eprint empty "" archivePrefix empty "" archivePrefix "arXiv" = new.block " " eprint * " " * new.block " " eprint * " " * if if if FUNCTION format.doi doi empty "" " " doi * " " * if FUNCTION format.pid doi empty eprint empty ur...

  3. [3]

    - [1] #1 = = ^ ^ ^ .\!\!^ d .\!\!^ h .\!\!^ m .\!\!^ s .\!\!^ @mss

    thebibliography [1] 20pt to REFERENCES 6pt =0pt -12pt 10pt plus 3pt =0pt =0pt =1pt plus 1pt =0pt =0pt -12pt =13pt plus 1pt =20pt =13pt plus 1pt \@M =10000 =-1.0em =0pt =0pt 0pt =0pt =1.0em @enumiv\@empty 10000 10000 `\.\@m \@noitemerr \@latex@warning Empty `thebibliography' environment \@ifnextchar \@reference \@latexerr Missing key on reference command E...

    2021

  4. [4]

    A., Czerny , B., Lasota , J

    Abramowicz , M. A., Czerny , B., Lasota , J. P., & Szuszkiewicz , E. 1988, , 332, 646, 10.1086/166683

    doi:10.1086/166683 1988

  5. [5]

    C., & Bhattacharya, D

    Adegoke, O., Dhang, P., Mukhopadhyay, B., Ramadevi, M. C., & Bhattacharya, D. 2018, Monthly Notices of the Royal Astronomical Society, 476, 1581, 10.1093/mnras/sty263

    doi:10.1093/mnras/sty263 2018

  6. [6]

    K., Mukhopadhyay, B., & Misra, R

    Adegoke, O. K., Mukhopadhyay, B., & Misra, R. 2020, Monthly Notices of the Royal Astronomical Society. https://api.semanticscholar.org/CorpusID:210023531

    work page 2020

  7. [7]

    2011, The Astrophysical Journal Letters, 742, L17, 10.1088/2041-8205/742/2/L17

    Altamirano, D., Belloni, T., Linares, M., et al. 2011, The Astrophysical Journal Letters, 742, L17, 10.1088/2041-8205/742/2/L17

    doi:10.1088/2041-8205/742/2/l17 2011

  8. [8]

    2000, Astronomy and astrophysics, 355, 271

    Belloni, T., Klein-Wolt, M., M \'e ndez, M., van der Klis , M., & van Paradijs , J. 2000, Astronomy and astrophysics, 355, 271

    work page 2000

Show all 43 references
  1. [9]

    2001, A&A, 372, 551, 10.1051/0004-6361:20010480

    Belloni, T., Méndez, M., & Sánchez-Fernández, C. 2001, A&A, 372, 551, 10.1051/0004-6361:20010480

    2001 doi

  2. [10]

    1870, Annali di Matematica Pura ed Applicata, 4, 140–158, 10.1007/bf02420029

    Betti, E. 1870, Annali di Matematica Pura ed Applicata, 4, 140–158, 10.1007/bf02420029

    doi

  3. [11]

    Broomhead, D., & King, G. P. 1986, Physica D: Nonlinear Phenomena, 20, 217, https://doi.org/10.1016/0167-2789(86)90031-X

    1986 doi

  4. [12]

    Buades, A., Coll, B., & Morel, J. M. 2004, CMLA Preprint, 5, 19

    2004

  5. [13]

    2005, in 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05), Vol

    Buades, A., Coll, B., & Morel, J.-M. 2005, in 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05), Vol. 2, 60--65 vol. 2, 10.1109/CVPR.2005.38

    2005 doi

  6. [14]

    P., & Sinha, N

    Chakka, S. P., & Sinha, N. 2024, Proceedings of 15^ th International Conference on Signal Processing and Communications (SPCOM), Indian Institute of Science, https://doi.org/10.48550/arXiv.2307.07703

    2024 doi arXiv

  7. [15]

    Chen, G., Zhu, F., & Heng, P. A. 2015, in 2015 IEEE International Conference on Computer Vision (ICCV), 477--485, 10.1109/ICCV.2015.62

    2015 doi

  8. [16]

    M., & Swinney, H

    Fraser, A. M., & Swinney, H. L. 1986, Phys. Rev. A, 33, 1134, 10.1103/PhysRevA.33.1134

    1986 doi

  9. [17]

    1983, Physica D: Nonlinear Phenomena, 9, 189, https://doi.org/10.1016/0167-2789(83)90298-1

    Grassberger, P., & Procaccia, I. 1983, Physica D: Nonlinear Phenomena, 9, 189, https://doi.org/10.1016/0167-2789(83)90298-1

    1983 doi

  10. [18]

    C., Vilhu, O

    Hannikainen, D. C., Vilhu, O. R., Rodriguez, J., et al. 2003, Astronomy and Astrophysics, 411. https://api.semanticscholar.org/CorpusID:53531435

    2003

  11. [19]

    2006, Physica D: Nonlinear Phenomena, 215, 137

    Harikrishnan, K., Misra, R., Ambika, G., & Kembhavi, A. 2006, Physica D: Nonlinear Phenomena, 215, 137

    2006

  12. [20]

    2016, in 2016 International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS), 1--6, 10.1109/ISPACS.2016.7824686

    Imamura, K., Kimura, N., Satou, F., Sanada, S., & Matsuda, Y. 2016, in 2016 International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS), 1--6, 10.1109/ISPACS.2016.7824686

    2016

  13. [21]

    2015, Astronomy &; Astrophysics, 574, A92, 10.1051/0004-6361/201425003

    Janiuk, A., Grzedzielski, M., Capitanio, F., & Bianchi, S. 2015, Astronomy &; Astrophysics, 574, A92, 10.1051/0004-6361/201425003

    2015 doi

  14. [22]

    L., Miller, J

    King, A. L., Miller, J. M., Raymond, J., et al. 2012, The Astrophysical Journal, 746, L20, 10.1088/2041-8205/746/2/l20

    2012 doi

  15. [23]

    P., Pooley, G

    Klein-Wolt, M., Fender, R. P., Pooley, G. G., et al. 2002, Monthly Notices of the Royal Astronomical Society, 331, 745, 10.1046/j.1365-8711.2002.05223.x

    2002

  16. [24]

    J., & Schreiber, T

    Kostelich, E. J., & Schreiber, T. 1993, Phys. Rev. E, 48, 1752, 10.1103/PhysRevE.48.1752

    1993 doi

  17. [25]

    M., Raymond, J., Fabian, A

    Miller, J. M., Raymond, J., Fabian, A. C., et al. 2016, The Astrophysical Journal Letters, 821, L9, 10.3847/2041-8205/821/1/l9

    2016 doi

  18. [26]

    F., & Rodríguez, L

    Mirabel, I. F., & Rodríguez, L. F. 1994, Nature, 371, 46–48, 10.1038/371046a0

    1994 doi

  19. [27]

    P., Ambika, G., & Kembhavi, A

    Misra, R., aff, Harikrishnan, K. P., Ambika, G., & Kembhavi, A. K. 2006, Astrophys. J., 643, 1114, 10.1086/503094

    2006 doi

  20. [28]

    P., Mukhopadhyay, B., Ambika, G., & Kembhavi, A

    Misra, R., Harikrishnan, K. P., Mukhopadhyay, B., Ambika, G., & Kembhavi, A. K. 2004, The Astrophysical Journal, 609, 313, 10.1086/421005

    2004 doi

  21. [29]

    2004, AIP Conference Proceedings, 714, 48, 10.1063/1.1780998

    Mukhopadhyay, B. 2004, AIP Conference Proceedings, 714, 48, 10.1063/1.1780998

    2004 doi

  22. [30]

    1994, , 428, L13, 10.1086/187381

    Narayan , R., & Yi , I. 1994, , 428, L13, 10.1086/187381

    1994 doi

  23. [31]

    R., & Mukhopadhyay , B

    Rajesh , S. R., & Mukhopadhyay , B. 2010, , 402, 961, 10.1111/j.1365-2966.2009.15925.x

    2010

  24. [32]

    Rao, A., & Vadawale, S. V. 2012, The Astrophysical Journal Letters, 757, L12, 10.1088/2041-8205/757/1/L12

    2012 doi

  25. [33]

    1997, Computers & Chemical Engineering, 21, S1149, https://doi.org/10.1016/S0098-1354(97)87657-0

    Rhodes, C., & Morari, M. 1997, Computers & Chemical Engineering, 21, S1149, https://doi.org/10.1016/S0098-1354(97)87657-0

    1997 doi

  26. [34]

    Rousseeuw, P. J. 1987, Journal of Computational and Applied Mathematics, 20, 53–65, 10.1016/0377-0427(87)90125-7

    1987 doi

  27. [35]

    I., & Sunyaev , R

    Shakura , N. I., & Sunyaev , R. A. 1973, , 24, 337

    1973

  28. [36]

    Smith, S. W. 1997, The Scientist & Engineer's Guide to Digital Signal Processing, hardcover edn. (California Technical Pub), 626

    1997

  29. [37]

    Stewart, G. W. 1993, SIAM Review, 35, 551, 10.1137/1035134

    1993 doi

  30. [38]

    H., & Miller, E

    Tracey, B. H., & Miller, E. L. 2012, IEEE Transactions on Biomedical Engineering, 59, 2383, 10.1109/TBME.2012.2208964

    2012

  31. [39]

    2011, Phys

    Tung, W.-w., Gao, J., Hu, J., & Yang, L. 2011, Phys. Rev. E, 83, 046210, 10.1103/PhysRevE.83.046210

    2011 doi

  32. [40]

    M., Fabian, A

    Uttley, P., Cackett, E. M., Fabian, A. C., Kara, E., & Wilkins, D. R. 2014, The Astronomy and Astrophysics Review, 22, 10.1007/s00159-014-0072-0

    2014 doi

  33. [41]

    2009, IEEE Signal Processing Letters, 16, 973, 10.1109/LSP.2009.2027669

    Van De Ville, D., & Kocher, M. 2009, IEEE Signal Processing Letters, 16, 973, 10.1109/LSP.2009.2027669

    2009

  34. [42]

    2006, Nonlinear Processes in Geophysics, 13, 321, 10.5194/npg-13-321-2006

    Venema, V., Ament, F., & Simmer, C. 2006, Nonlinear Processes in Geophysics, 13, 321, 10.5194/npg-13-321-2006

    2006 doi

  35. [43]

    1987, Chemometrics and Intelligent Laboratory Systems, 2, 37, https://doi.org/10.1016/0169-7439(87)80084-9

    Wold, S., Esbensen, K., & Geladi, P. 1987, Chemometrics and Intelligent Laboratory Systems, 2, 37, https://doi.org/10.1016/0169-7439(87)80084-9

    1987 doi

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.