Pith. sign in

REVIEW 5 major objections 5 minor 33 references

Fuzzy permutation time irreversibility for nonequilibrium analysis of complex system

T0 review · 5 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Fuzzy permutation time irreversibility ranks heartbeats young > elderly > CHF, matching complexity-loss theory.

desk verdict A plausible new fuzzy permutation TIR with a clean methodological core, but the accuracy claim leans on an unexamined ground truth and an unexplained fPEn contradiction. read the letter →

arxiv 2506.08058 v2 pith:FS5LFRV4 submitted 2025-06-09 physics.data-an cond-mat.stat-mechq-bio.QM

Wenpo Yao This is my paper
classification physics.data-ancond-mat.stat-mechq-bio.QM PACS 05.45.Tp05.70.Ln07.05.Mh02.70.Wz
keywords fuzzypermutationtimeirreversibilityentropycomplexsystemsymbolicdynamicsheartbeatanalysisnonequilibriumsurrogatedata
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 proposes fuzzy permutation time irreversibility (fpTIR), a way of measuring time asymmetry that replaces the usual ordinal pattern with an amplitude permutation carrying a fuzzy membership weight. The weight is computed from the spread of adjacent differences in each embedded vector, so that the statistic retains information about the absolute sizes of fluctuations, not just their rank order. The paper's central claim is that fpTIR characterizes the structure of a sequence more accurately than standard permutation time irreversibility, and that this improves nonequilibrium analysis of complex systems. On heartbeat recordings from PhysioNet, fpTIR orders healthy young > healthy elderly > congestive-heart-failure patients at every tested dimension and delay, consistent with the theory that physiological complexity is lost with aging and disease; standard pTIR at dimension 3 and delay 1 instead ranks CHF heartbeats above healthy elderly. The paper also defines fuzzy permutation entropy (fPEn), which often disagrees and sometimes opposes fpTIR, and treats the relationship between the two as an open question.

What carries the argument

The central object is the fuzzy permutation: an amplitude permutation paired with a membership degree $p = \exp[-c\,\sigma]$, where $\sigma$ is the standard deviation of the adjacent differences of the vector's sorted elements; equal values are merged, and a vector with all-equal elements gets membership 1. This object carries the argument because it converts the absolute spacing of time-series values into a graded weight, so that two vectors with the same ordinal ranking but different element separations contribute differently to the permutation statistics. fpTIR is then the sum over permutation types of the normalized forward–backward probability difference, and fPEn is the Shannon entropy of the same fuzzy-permutation distribution, normalized by $\ln N(m)$ where $N(m)$ is the maximum number of permutation types.

What would settle it

Generate a stationary Gaussian linear process, which is theoretically time-reversible, and run the fpTIR pipeline: if the statistic falls outside the 2.5–97.5% surrogate band, it is detecting something other than irreversibility; conversely, bootstrap resampling of the PhysioNet heartbeat groups that shrinks the young–elderly–CHF gap to zero would directly undermine the accuracy claim.

Watch

Extended reading notes

Core claim

The central discovery, on the paper's own terms, is that the coarse-grained nature of ordinal permutation—which ignores the absolute distances between elements—is a genuine source of error in time-irreversibility analysis, and that a fuzzy amplitude permutation corrects it. For each $m$-dimensional embedding vector, the paper sorts the values, records the amplitude permutation (the original positions in the sorted order), merges equal values so they do not dominate the spread, and assigns a membership degree $p = \exp[-c\,\sigma]$, with $\sigma$ the standard deviation of the adjacent differences of the sorted vector. The probability of each fuzzy permutation is the accumulated membership divided by the number of vectors, and fpTIR is the summed relative difference $\sum_i |p_f(i)-p_b(i)|/(p_f(i)+p_b(i))$ between forward and reversed sequences. The paper demonstrates with logistic, Hénon, and AR(1) surrogate experiments that fpTIR separates nonlinear from linear dynamics, and on PhysioNet heartbeats it yields healthy young > healthy elderly > CHF while standard pTIR at $m=3,\tau=1$ gives the anomalous CHF > elderly ordering. The claim is that fpTIR therefore enhances the accuracy of nonequilibrium analysis of complex systems.

Load-bearing premise

The paper judges fpTIR 'accurate' because it matches the assumption that heartbeat complexity falls in the order healthy young > healthy elderly > CHF, but it supplies no independent measure or significance test for that ordering.

Editorial extensions

If this is right

  • At dimension 3 and delay 1, where standard pTIR ranks CHF heartbeats above healthy elderly, fpTIR restores the complexity-loss ordering young > elderly > CHF, giving a more sensitive permutation-based nonequilibrium marker for physiological signals.
  • Because fuzzy permutation merges equal values and uses subtraction-based probability differences, it sidesteps the forbidden-permutation and tie-handling problems that complicate standard pTIR.
  • The companion fPEn is better at separating the three heartbeat groups and is insensitive to dimension and delay, so combining fpTIR and fPEn describes complex systems more comprehensively than either alone.
  • Fuzzy permutation is more sensitive to noise and requires more computation than ordinary permutation analysis, so the paper's method is preferable when signal interference is small and amplitude information matters.

Reading between the lines

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

  • If fpTIR measures directional asymmetry while fPEn measures distribution uniformity, their opposite rankings of the heartbeat groups are not a contradiction but a separation of two properties; a joint phase diagram of irreversibility versus entropy could be a useful diagnostic.
  • The amplitude-spacing weighting is generic and could be grafted onto other permutation-based tools, such as permutation entropy or transfer entropy, whenever magnitude information matters.
  • A natural extension, not pursued in the paper, is to compare fpTIR against direct entropy-production estimates on experimental nonequilibrium systems to test whether the physiological ordering generalizes.
  • The heartbeat ordering rests on a complexity-loss assumption without significance testing; a bootstrap confidence interval on the fpTIR difference between groups would make the central accuracy claim directly testable.
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

5 major / 5 minor

Summary. The paper proposes a fuzzy permutation time irreversibility (fpTIR) measure that combines amplitude permutation with a negative-exponential membership degree, and a companion fuzzy permutation entropy (fPEn). The authors test fpTIR and fPEn on logistic, Hénon, and AR(1) series with iAAFT surrogates, then apply the measures to PhysioNet heartbeat recordings from CHF patients and healthy elderly and young subjects. They report that fpTIR orders heartbeats as healthy young > healthy elderly > CHF and conclude that fpTIR accurately characterizes sequence structure and improves nonequilibrium analysis.

Significance. A fuzzy refinement of permutation time irreversibility is a reasonable and potentially useful idea, and the surrogate results for nonlinear versus linear model series behave as expected. The heartbeat ordering is a concrete, falsifiable prediction. However, the central accuracy claim is under-supported: the definition of the Ys-based measure is misprinted, the heartbeat validation lacks significance tests and relies on an assumed complexity-loss ordering, and the surrogate analysis only establishes nonlinearity, not structural accuracy. With corrected definitions and proper statistical validation, the method could make a useful contribution to nonequilibrium time-series analysis.

major comments (5)
  1. [2.3, Eq. (3)] As typeset, Eq. (3) defines fpTIR as a sum of terms of the form (p_f - p_b)/(p_f + p_b) without an absolute value. Since both probability distributions sum to 1, the raw differences sum to 0, so fpTIR would be identically 0 for any normalized forward/backward distributions. The text says Ys is "based on subtraction," but a meaningful measure must be a distance such as sum over patterns of |p_f - p_b|/(p_f + p_b). Please correct the equation and state the definition of Ys precisely; all numerical results depend on this.
  2. [3.2, Figs. 3-4] The central claim that fpTIR "accurately characterizes" heartbeat complexity is validated only by agreement with the complexity-loss ordering (young > elderly > CHF) stated in Sec. 3.2. No significance test is reported: Fig. 3 shows mean +/- standard error, and at m=2, t=1 the group error bars overlap, while at m=3, t=1 no confidence intervals or p-values are given. The opposite fPEn ordering is dismissed with "requires further study," which leaves a major inconsistency unexplained. Please add per-subject statistical tests among the three groups, report effect sizes, and either provide an independent benchmark for the complexity-loss assumption or substantially weaken the accuracy claim.
  3. [3.1, Fig. 2] The surrogate analysis establishes only that the logistic and Hénon series are significantly different from iAAFT surrogates and that the AR(1) series is not; this is a test of nonlinearity, not a demonstration that fpTIR recovers correct structural features or that it is more accurate than pTIR. Statements in the Abstract and Conclusion that fpTIR "accurately characterizes the structure of the sequences" therefore go beyond what Fig. 2 can support.
  4. [2.3, Eq. (4)] The normalization constants P(m) for fPEn are listed as 3, 13, 73, and 501 for m=2, 3, 4, and 5. The number of possible tie-allowed weak-order permutations is standardly 3, 13, 75, and 541. Please clarify the counting that yields 73 and 501, because the fPEn values for m=4 and 5 in Fig. 2 are scaled by these constants and would be miscalibrated if the constants are incorrect.
  5. [3.2, Fig. 3] The membership control parameter lambda is set to 1 in the model-series study (Sec. 3.1) but to 0.1 for the heartbeat analysis, and the assertion that "fpTIR of the heartbeats was not significantly affected by the membership parameter lambda" is not supported by any displayed sensitivity analysis. Please provide a sensitivity curve or a principled criterion for choosing lambda, since the results depend on a free parameter whose value changes between experiments.
minor comments (5)
  1. [Throughout] Many equations and inline symbols are garbled by the translation or typesetting (e.g., Eqs. (1)-(4) and the sentence defining p_i); a clean, correctly typeset version is needed for the paper to be reproducible.
  2. [2.2] The terms "standard vector" and "strange vector" are confusing: a vector whose adjacent-difference vector has zero standard deviation is called "standard," which conflicts with the usual meaning. Please define these terms explicitly with equations.
  3. [2.2] The sentence "If the vector contains only two elements, sigma(v) can be modified as the ratio of the standard deviation of the vector to the whole sequence" is unclear; specify the exact formula and the case in which it applies.
  4. [3.2] The heartbeat analysis should state how many beats or segments per subject were analyzed, whether ectopic beats or artifacts were removed, and how the reported mean and standard error were computed across subjects.
  5. [Fig. 1] Figure 1 is referenced in Sec. 2.3 but not shown in the text; please include the flowchart or remove the reference.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: fpTIR is a newly defined statistic, and its heartbeat ordering is validated against an external complexity-loss benchmark rather than being built into the definition.

full rationale

The derivation chain is self-contained and empirical. Equations (1)-(3) define the fuzzy membership, the fuzzy permutation probability, and the forward/backward probability difference fpTIR; no parameter is fitted to the target heartbeat ordering. The control parameter λ is set to 1 for model series and to 0.1 for heartbeats after inspecting results, but the paper states that 'the fpTIR of the heartbeats was not significantly affected by the membership parameter λ, especially when m=3', so the reported ordering healthy young > healthy elderly > CHF is not forced by a fitted constant. The 'accuracy' claim is benchmarked externally: iAAFT surrogates in Sec. 3.1 test nonlinearity, and Sec. 3.2 invokes the complexity-loss theory of Goldberger et al. (ref. [32]), an external physiological hypothesis, as the ground truth. The self-citations to Yao's earlier pTIR work justify the choice of the Ys subtraction statistic and amplitude permutation, but the formula is stated explicitly and tested on independent PhysioNet data, so the citations are not load-bearing in the sense of importing an unverified premise. The acknowledged discrepancy with fPEn ('the fpTIR and fPEn... requires further study') is a limitation, not a circular step. I find no equation or fitted quantity that reduces by construction to its own input.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

No new physical entities (particles, forces, dimensions, conserved quantities) are introduced. Fuzzy permutation is a mathematical construction defined in Sec. 2.2, not an entity with independent falsifiable handles. The main fitted quantity is the membership control lambda, and the central validation relies on two domain assumptions: the surrogate null model and the complexity-loss ordering.

free parameters (1)
  • lambda (membership control parameter) = 1.0 for model series; 0.1 for heartbeat analysis
    Controls the decay rate of the exponential membership exp(-lambda * sigma(v)) in Eq. (1). The value is chosen by hand, and for the heartbeat application it is set to 0.1 after inspecting results, making it a fitted rather than derived parameter.
assumptions (4)
  • domain assumption Surrogate null hypothesis: original series should fall within 2.5-97.5% surrogate percentiles if linear/stationary, outside if nonlinear.
    Used in Sec. 3.1 to label logistic and Henon as nonlinear and AR1 as linear; the validity of iAAFT surrogates for this fuzzy permutation statistic is assumed rather than proved.
  • domain assumption Complexity-loss ordering of heartbeat groups: healthy young > healthy elderly > CHF in regulatory complexity.
    Invoked in Sec. 3.2 as the ground truth against which fpTIR is judged 'accurate'; no independent metric or statistical test establishes this ordering for these specific recordings.
  • ad hoc to paper Exponential membership form exp(-lambda * sigma(v)) is a suitable fuzzification of amplitude permutation.
    Introduced in Sec. 2.2, Eq. (1), with references to fuzzy set conventions but without derivation or comparison to alternative membership functions.
  • domain assumption Ys subtraction-based probability difference is appropriate for permutation time irreversibility.
    Borrowed from the author's prior pTIR work (Refs. 1,7-9,15) and used in Sec. 2.3, Eq. (3), to avoid division by forbidden permutations; treated as established rather than re-derived here.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fuzzy permutation time irreversibility for nonequilibrium analysis of complex system." pith.science (2026). https://pith.science/paper/FS5LFRV4

@misc{pith2026250608058,
  author       = {Pith},
  title        = {Pith review of: Fuzzy permutation time irreversibility for nonequilibrium analysis of complex system},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FS5LFRV4}},
  note         = {Machine review of arXiv:2506.08058}
}
read the original abstract

Permutation time irreversibility is an important method to quantify nonequilibrium characteristics of complex systems; however, ordinal pattern is a coarse-graining alternative of temporal structure and cannot accurately represent detailed structural information. This study aims to propose a fuzzy permutation time irreversibility (fpTIR) by measuring the difference between vector elements based on a negative exponential function. The amplitude permutation of vector is constructed and its membership degree is calculated; then, the difference in probability distribution between the forward and backward sequences is measured for fpTIR. To compare and measure the system's complexity, the Shannon entropy is calculated as the average amount of information in the fuzzy permutation probability distribution, i.e., fuzzy permutation entropy (fPEn). According to the surrogate theory, mode series are generated using logistic, Henon, and first-order autoregressive systems to verify the fpTIR, which is then used to analyze the heartbeats of patients with congestive heart failure and healthy elderly and young participants from the PhysioNet database. Results suggest that the fpTIR effectively measures the system's nonequilibrium characteristics, thus improving the accuracy of heartbeat analysis. However, in analyzing probability distributions, the fpTIR and fPEn exhibit discrepancies in the chaotic series and even opposite results in the heartbeats, wherein the results of fpTIR are consistent with the theory of complexity loss in aging and disease. Overall, the fpTIR accurately characterizes the structure of the sequences and enhances the accuracy of the nonequilibrium analysis of complex systems, providing a theoretical basis for exploring complex systems from the perspectives of nonequilibrium dynamics and entropy complexity.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

33 extracted references · 33 canonical work pages

  1. [1]

    YaoWP2024 Phys. Rev. E109054104

    work page

  2. [2]

    DawC,FinneyC,TracyE2003 Rev. Sci. Instrum.74915

    work page

  3. [3]

    BandtC,PompeB2002 Phys. Rev. Lett.88174102

    work page

  4. [4]

    MartinezJ,HerreraDJ,ChavezM2018 Chaos28123111

    work page

  5. [5]

    ZaninM,RodriguezA,MenasalvasRE,PapoD2018 Entropy20665

    work page

  6. [6]

    ZaninM2021 Chaos31103118

    work page

  7. [7]

    YaoWP,DaiJF,PercM,WangJ,YaoDZ,GuoDQ2020 Nonlinear Dyn.100907

    work page

  8. [8]

    Nonlinear Sci

    YaoWP,YaoWL,XuR,WangJ2023 Commun. Nonlinear Sci. Numer. Simul.117106925

    work page

Show all 33 references
  1. [9]

    YaoWP,YaoWL,WangJ2022 Phys. Lett. A430127977

  2. [10]

    AmigoJM,ZambranoS,SanjuanMA2010 Int. J. Bifurcation Chaos202915

  3. [11]

    AmigoJM,ZambranoS,SanjuanMA2007 EPL7950001

  4. [12]

    AmigoJM,KennelMB2007 Physica D231137

  5. [13]

    ZaninM2008 Chaos18013119

  6. [14]

    KulpCW,ZuninoL,OsborneT,ZawadzkiB2017 Phys. Rev. E96022218

  7. [15]

    YaoWP,YaoWL,WangJ,DaiJ2019 Phys. Lett. A383738

  8. [16]

    YaoWP,YaoWL,YaoDZ,GuoDQ,WangJ2020 Appl. Phys. Lett.116014101

  9. [17]

    YaoWP,YaoWL,WangJ2019 Phys. Lett. A3831764

  10. [18]

    BianC,QinC,MaQ,ShenQ2012 Phys. Rev. E85021906

  11. [19]

    ZuninoL,OlivaresF,ScholkmannF,RossoOA2017 Phys. Lett. A3811883

  12. [20]

    CuestaFD,VarelaEM,MolinaPA,VargasB2018 Complexity20181324696

  13. [21]

    WeissG1975 J. Appl. Probab.12831

  14. [22]

    KellyFP1979 Reversibility and Stochastic Networks （Chichester:CambridgeUniversity Press） pp5–10

  15. [23]

    ZadehLA1999 Fuzzy Sets Syst.1009

  16. [24]

    Neural Syst

    ChenW,WangZ,XieHB,YuWX2007 IEEE Trans. Neural Syst. Rehabil. Eng.15266

  17. [25]

    Fuzzy Syst.303785

    RostaghiM,KhatibiMM,AshoryMR,AzamiH2021 IEEE Trans. Fuzzy Syst.303785

  18. [26]

    Sin.63078704

    YaoWP,LiuTB,DaiJF,WangJ2014 Acta Phys. Sin.63078704

  19. [27]

    SchreiberT,SchmitzA1996 Phys. Rev. Lett.77635

  20. [28]

    SchreiberT,SchmitzA2000 Physica D142346

  21. [29]

    GoldbergerAL,AmaralLA,GlassL,HausdorffJM,IvanovPC,MarkRG,PengC, StanleyHE2000 Circulation101e215

  22. [30]

    BaimDS,ColucciWS,MonradES,Smith,HartonS,BraunwaldE1986 J. Am. Coll. Cardiol.7661

  23. [31]

    IyengarN,PengCK,MorinR,GoldbergerAL,LipsitzLA1996 Am. J. Physiol.271R1078

  24. [32]

    Aging2323

    GoldbergerAL,PengCK,LipsitzLA2002 Neurobiol. Aging2323

  25. [33]

    XiongW,FaesL,IvanovP2017 Phys. Rev. E95062114

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.