{"id":"c53ceb96-a351-46c3-a95c-299a6af708f7","arxiv_id":"2506.08058","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Fuzzy permutation time irreversibility, built from amplitude rankings and exponential membership weights, orders heartbeats as healthy young greater than healthy elderly greater than heart failure, matching the complexity-loss hypothesis.","lead":"Researchers propose a new way to measure time irreversibility in complex signals by replacing coarse ordinal patterns with fuzzy amplitude rankings and then test it on heartbeats from heart-failure patients and healthy volunteers. If it works, it gives physiologists a sharper marker for detecting complexity loss and nonequilibrium dynamics.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed gain in accuracy rests entirely on agreement with an assumed complexity-loss ordering; no significance test or independent benchmark supports it, and fPEn's opposite ordering remains unexplained.","rationale":"The paper is a coherent methodological proposal: the fuzzy permutation construction is defined clearly, the surrogate checks in Sec. 3.1 behave as expected for distinguishing nonlinear from linear model series, and the physiological application is relevant. I found no internal mathematical contradiction in Eqs. (1)-(4). However, the headline claim that fpTIR 'enhances the accuracy' of nonequilibrium analysis is supported only by agreement with an assumed complexity-loss ordering of heartbeats, and that support is weak for three concrete reasons. First, the group ordering in Figs. 3-4 is presented as mean plus or minus standard error, with no significance tests; at several parameter settings the error bars overlap, so it is unclear whether the reported ordering is statistically real. Second, the intended ground truth, complexity-loss theory, is stated rather than demonstrated for these specific recordings, and the companion fPEn gives the opposite ordering, which the paper itself cannot explain. Third, the surrogate validation does not provide a quantitative benchmark for structural accuracy; it only separates known nonlinear from linear model series, so it cannot validate the specific claim that fpTIR is more accurate than pTIR at characterizing sequence structure. These issues are addressable with additional statistical analysis and benchmark experiments, so they do not require rejection, but they do make the current evidence conditional. My read agrees with the Reader's weakest_assumption, and the recommended CONDITIONAL verdict matches the Reader's assessment.","tokens_in":8061,"tokens_out":2525,"duration_ms":31962,"concrete_test":"Re-analyze the raw heartbeat data with per-subject fpTIR values rather than group means: compute pairwise Mann-Whitney U tests and bootstrap confidence intervals at each (m, tau, lambda) grid point, report effect sizes and subject-level distribution overlap, and verify whether young > elderly > CHF holds significantly for fpTIR. Repeat the analysis after excluding ectopic beats and matching groups by recording length and protocol. Independently, build surrogate benchmarks with known time-irreversibility ground truth (e.g., Gaussian vs nonlinear stochastic processes with known TIR) and check whether fpTIR recovers the true ordering while fPEn does not.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper's central claim (Abstract, Sec. 3.2, Conclusion) is that fpTIR 'accurately characterizes the structure of sequences' and improves nonequilibrium analysis, with heartbeat fpTIR ordering young > elderly > CHF at m=3, t=1 (Fig. 3). The only evidence for 'accurate' is that this ordering matches the complexity-loss assumption stated in Sec. 3.2 as 'consistent with the complexity losing theory'. That theory is used as an uncontested ground truth, not as a falsifiable benchmark. No significance test separates the groups: Fig. 3 and Fig. 4 report mean ± standard error only, and at m=2, t=1 the group error bars visibly overlap, so the ordering may not be statistically distinguishable. The companion fPEn gives the opposite trend (CHF > elderly > young, Fig. 4), and the paper dismisses this as 'requires further study'. If the complexity-loss assumption is not a reliable ground truth for these recordings—or if group differences are confounded by recording protocol, age, heart rate, or ectopy—the fpTIR accuracy claim has no independent support. The surrogate validation in Sec. 3.1 only establishes nonlinearity (logistic/Henon outside surrogate band, AR1 inside); it does not test whether fpTIR recovers correct structural features, so it cannot underwrite the accuracy claim.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","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.","tokens_in":8306,"tokens_out":7252,"duration_ms":88240,"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":[{"comment":"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.","section":"2.3, Eq. (3)"},{"comment":"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.","section":"3.2, Figs. 3-4"},{"comment":"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.","section":"3.1, Fig. 2"},{"comment":"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.","section":"2.3, Eq. (4)"},{"comment":"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.","section":"3.2, Fig. 3"}],"minor_comments":[{"comment":"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.","section":"Throughout"},{"comment":"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.","section":"2.2"},{"comment":"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.","section":"2.2"},{"comment":"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.","section":"3.2"},{"comment":"Figure 1 is referenced in Sec. 2.3 but not shown in the text; please include the flowchart or remove the reference.","section":"Fig. 1"}],"recommendation":"major_revision","confidential_remarks":"The English translation has corrupted many equations, which I assume is a production issue, but the author must supply a clean version. The more substantive concern is that the paper's central accuracy claim is validated only by agreement with an assumed complexity-loss ordering and lacks statistical testing; if the author cannot add appropriate tests and an independent benchmark, rejection may be warranted even after the definitional issues are fixed."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's my take. The new piece is the specific construction: amplitude permutation, membership from an exponential of local standard deviation, and forward-backward probability difference. That combination is not in the existing pTIR literature, and the paper makes a good case that it handles equal values more cleanly than ordinal permutation. The surrogate tests in Sec. 3.1 behave exactly as they should—logistic and Henon are outside the surrogate band, AR1 inside—so the measure is at least a plausible nonlinearity detector.\n\nWhere it gets soft is the heartbeat validation. The central claim that fpTIR improves accuracy rests on agreement with the complexity-loss ordering (young > elderly > CHF). That ordering is assumed, not independently established for these recordings, and the figures show mean ± SE only. At m=2, t=1 the groups overlap; no significance test is reported. More importantly, the companion fPEn gives the opposite ordering, and the paper dismisses it as 'requires further study.' That is not a dismissal you can afford when the only evidence for 'accuracy' is agreement with one particular theory. The stress-test note is right: the surrogate tests only establish nonlinearity, not structural accuracy.\n\nOther issues are minor. Equation (3) as printed omits the absolute value; probably a typesetting issue but should be fixed. The choice λ=0.1 for Fig. 4 appears post-hoc, although the paper claims insensitivity. No code or data is provided, which hampers exact reproduction. The self-citations are not a problem; they are relevant prior work.\n\nWho is this for? People working on permutation-based irreversibility and complexity measures. It is a legitimate extension with a plausible but unproven practical gain. I would send it to peer review, but the review should demand significance tests, an independent benchmark for the heartbeat ordering, and an explanation (or at least a proper test) of the fPEn contradiction before the accuracy claim is accepted.","headline":"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.","tokens_in":8819,"tokens_out":3282,"would_cite":false,"duration_ms":38035,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["05.45.Tp","05.70.Ln","07.05.Mh","02.70.Wz"],"model":"deepseek-v4-flash","headline":"Fuzzy permutation time irreversibility ranks heartbeats young > elderly > CHF, matching complexity-loss theory.","keywords":["fuzzy permutation","time irreversibility","permutation entropy","complex system","symbolic dynamics","heartbeat analysis","nonequilibrium dynamics","surrogate data"],"falsifier":"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.","tokens_in":7815,"feed_emoji":"🫀","tokens_out":12613,"duration_ms":120780,"temperature":0.7,"pith_summary":"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.","feed_headline":"Fuzzy timing measure orders heartbeats: young > elderly > CHF","feed_subtitle":"Restores the expected young-elderly-CHF order where the standard permutation measure fails.","key_machinery":"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.","core_discovery":"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.","pith_inferences":["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."],"forward_implications":["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."],"supporting_citations":[{"why":"supplies the pTIR framework identifying forbidden permutations and equal values as bias sources that fpTIR corrects.","marker":"[1]"},{"why":"introduces permutation probability-difference as a baseline measure of time irreversibility that fpTIR extends.","marker":"[4]"},{"why":"distinguishes original from amplitude permutation, the structural distinction at the core of fuzzy permutation.","marker":"[9]"},{"why":"provides the equal-value modification and the count of permutation types used in fPEn normalization.","marker":"[18]"},{"why":"supplies the improved amplitude-adjusted Fourier transform surrogate method used for fpTIR validation.","marker":"[27]"},{"why":"is the PhysioNet source of the heartbeat recordings analyzed.","marker":"[29]"},{"why":"provides the congestive-heart-failure heartbeat datasets (chfdb, chfdb2).","marker":"[30]"},{"why":"provides the Fantasia healthy elderly and young heartbeat datasets.","marker":"[31]"},{"why":"states the complexity-loss theory that the heartbeat ordering is tested against.","marker":"[32]"}],"fun_headline_variants":["Fuzzy amplitude measure fixes heartbeat irreversibility ranking","New irreversibility index restores young-elderly-CHF order","fpTIR outperforms ordinal permutations in heartbeat analysis","Amplitude-based fuzzy entropy improves nonequilibrium detection","Fuzzy permutations correct heart rhythm irreversibility errors"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"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.","fun_headline_variants_meta":{"raw":{"variants":["Fuzzy amplitude measure fixes heartbeat irreversibility ranking","New irreversibility index restores young-elderly-CHF order","fpTIR outperforms ordinal permutations in heartbeat analysis","Amplitude-based fuzzy entropy improves nonequilibrium detection","Fuzzy permutations correct heart rhythm irreversibility errors"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000224,"raw_usage":{"total_tokens":1538,"prompt_tokens":1103,"completion_tokens":435,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":719,"completion_tokens_details":{"reasoning_tokens":364}},"tokens_in":719,"tokens_out":435,"duration_ms":5818,"temperature":1.0,"reasoning_tokens":364,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T05:33:19.695103+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"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.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the pTIR framework identifying forbidden permutations and equal values as bias sources that fpTIR corrects."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"introduces permutation probability-difference as a baseline measure of time irreversibility that fpTIR extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"distinguishes original from amplitude permutation, the structural distinction at the core of fuzzy permutation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the equal-value modification and the count of permutation types used in fPEn normalization."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the improved amplitude-adjusted Fourier transform surrogate method used for fpTIR validation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"is the PhysioNet source of the heartbeat recordings analyzed."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the congestive-heart-failure heartbeat datasets (chfdb, chfdb2)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"provides the Fantasia healthy elderly and young heartbeat datasets."},{"cited_title":"Aging2323","cited_arxiv_id":null,"evidence_quote":"states the complexity-loss theory that the heartbeat ordering is tested against."}],"review_version":1}