pith. sign in

arxiv: 2604.16257 · v1 · pith:TJYMFJ3Mnew · submitted 2026-04-17 · ❄️ cond-mat.stat-mech · gr-qc

MF-toolkit: A High-Performance Python Library for Multifractal Analysis with Automated Crossover Detection, Source Identification and Application to Gravitational Waves Data

Nahuel Mendez , Maria Cristina Mariani Maria Pia Beccar-Varela , Osei Tweneboah , Sebastian Jaroszewicz This is my paper

Pith reviewed 2026-05-10 06:54 UTC · model grok-4.3

classification ❄️ cond-mat.stat-mech gr-qc
keywords multifractal detrended fluctuation analysisautomatic crossover detectionIAAFT surrogate datagravitational wave noisescaling propertiesPython librarytime series analysisLIGO data
0
0 comments X

The pith

MF-toolkit automates crossover detection and surrogate testing to identify multifractality sources in time series.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces MF-toolkit, a Python library for multifractal detrended fluctuation analysis that removes manual steps from the workflow. It supplies algorithms that automatically locate scaling regions in the fluctuation function and an integrated method to generate surrogate series for testing whether observed multifractality stems from correlations or from the shape of the value distribution. These capabilities are exercised on synthetic series for validation and then applied to the non-stationary noise present in LIGO gravitational-wave recordings. A sympathetic reader cares because manual region selection has long introduced operator-dependent results and because distinguishing the physical cause of scaling helps interpret real signals.

Core claim

MF-toolkit supplies three integrated features for MFDFA: the CDV-A and SPIC algorithms that locate crossover points without operator input, a built-in IAAFT routine that produces surrogate data to discriminate the origin of multifractality, and a generator for controlled synthetic series. When these tools are run on LIGO gravitational-wave noise, they return reproducible scaling exponents and a classification of the multifractal source.

What carries the argument

MF-toolkit library containing the CDV-A and SPIC automatic crossover detectors together with an IAAFT surrogate generator that distinguishes correlation-based from distribution-based multifractality.

Load-bearing premise

The automatic crossover detectors locate the true scaling intervals in the data without injecting new systematic errors, and this remains true for the non-stationary noise found in gravitational-wave recordings.

What would settle it

Apply CDV-A and SPIC to a collection of synthetic series whose crossover locations are known in advance; if the detected intervals deviate systematically from the known locations, the claim of unbiased automatic detection is refuted.

Figures

Figures reproduced from arXiv: 2604.16257 by Maria Cristina Mariani Maria Pia Beccar-Varela, Nahuel Mendez, Osei Tweneboah, Sebastian Jaroszewicz.

Figure 1
Figure 1. Figure 1: Representative Synthetic Time Series and Corresponding Probability Distribution Functions. Panels (1a) and (1b) show a monofractal Fractional Brownian Motion (fBm) series (H=0.5) and its Gaussian PDF, respectively. Panel (2a) illustrates a series with multifractality originating from a heavy-tailed distribution, shown in its PDF (2b); and (3a) depicts a series where multifractality stems from long-range co… view at source ↗
Figure 2
Figure 2. Figure 2: Quantile-Quantile (Q-Q) Plots of Synthetic Time Series against a Standard Gaussian Distribution. (a) fBm Series: The data points closely follow the Gaussian reference line, confirming that the monofractal Fractional Brownian Motion (fBm) series exhibits a Gaussian distribution of values. (b) Multifractal Series (Heavy￾tail PDF): A significant deviation from the linear reference is observed, particularly in… view at source ↗
Figure 3
Figure 3. Figure 3: MFDFA Curves for Synthetic Multifractal Signal (Long-Range Cor￾relations). The complete Multifractal Detrended Fluctuation Analysis workflow. (a) Generalized fluctuation function Fq(s) vs. scale s; (b) Generalized Hurst exponent h(q) vs. moment q; (c) Mass exponent τ (q) vs. q; (d) Singularity spectrum f(α) vs. α. The non-constant h(q) confirms multifractal scaling. 3.1.3. Source Identification with Surrog… view at source ↗
Figure 4
Figure 4. Figure 4: MFDFA Curves for Synthetic Multifractal Signal (Heavy-Tailed Dis￾tribution). The MFDFA analysis. (a) Generalized fluctuation function Fq(s); (b) Gener￾alized Hurst exponent h(q); (c) Mass exponent τ (q); (d) Singularity spectrum f(α). Note the dependence of h(q) on q, which is characteristic of multifractality originating from the heavy-tailed value distribution. • Shuffling: Randomly permutes the time ser… view at source ↗
Figure 5
Figure 5. Figure 5: The Effect of Random Shuffling on Multifractal Signatures. Panels (a) and (b) show the generalized fluctuation functions, Fq(s), and the generalized Hurst exponent, h(q), respectively, for the multifractal series where multifractality is due to a heavy-tailed probability distribution. Panels (c) and (d) show the corresponding plots for the multifractal series where multifractality is due to long-range corr… view at source ↗
Figure 6
Figure 6. Figure 6: The Effect of IAAFT Shuffling on Multifractal Signatures. Panels (a) and (b) show the generalized fluctuation functions, Fq(s), and the generalized Hurst exponent, h(q), respectively, for the multifractal series where multifractality is due to a heavy-tailed probability distribution. Panels (c) and (d) show the corresponding plots for the multifractal series where multifractality is due to long-range corre… view at source ↗
Figure 7
Figure 7. Figure 7: Synthetic Time Series Generated using the Fourier Filtering Method (FFM). This series was generated using the FFM algorithm to exhibit a crossover behav￾ior, resulting from the imposition of two distinct scaling regimes. The generated fractional Gaussian noise (fGn) process is characterized by two designated Hurst exponents: H1 for the long-range behavior and H2 for the short-range behavior. 22 [PITH_FULL… view at source ↗
Figure 8
Figure 8. Figure 8: Crossover Detection in Fq(s) Functions using the CDV-A Method. This figure illustrates the application of the Crossover Detection and Validation Algo￾rithm (CDV-A) to the generalized fluctuation functions, Fq(s), of the synthetic series with two scaling regimes. The dashed vertical line indicates the crossover index detected automatically by the CDV-A, determined by averaging the detection results across a… view at source ↗
Figure 9
Figure 9. Figure 9: Crossover Detection in Fq(s) Functions using the SPIC Method. This figure illustrates the application of the Sequential Permutation for Identifying Crossovers (SPIC) algorithm to the generalized fluctuation functions, Fq(s), of the synthetic series with two scaling regimes. The dashed vertical line indicates the crossover index detected automatically by the SPIC algorithm. To empirically determine the opti… view at source ↗
Figure 10
Figure 10. Figure 10: SPIC performance and statistical convergence under non-ideal con￾ditions. The algorithm was tested on a synthetic series (N = 100000) with a subtle structural break (α1 = 0.9, α2 = 1.1) and 30% additive Gaussian noise. a) Statistical reli￾ability: Rate of true detections over 10 independent trials. b) Computational cost: Mean execution time, where error bars denote ±1 standard deviation over the 10 indepe… view at source ↗
Figure 11
Figure 11. Figure 11: Robustness of the crossover detection algorithms and systematic offset evaluation. a) Monte Carlo simulations for the SPIC (blue circles) and CDV￾A (red squares) algorithms applied to synthetic multifractal time series corrupted with progressive levels of additive Gaussian white noise. Markers indicate the mean detected crossover scale across 100 independent stochastic realizations for each noise level, w… view at source ↗
Figure 12
Figure 12. Figure 12: Generalized Hurst exponent h(q) for H1 and L1 detectors. Note the similarity between ’Events’ and ’Pre-events’ [PITH_FULL_IMAGE:figures/full_fig_p028_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: The multifractal spectrum D(α). The overlapping curves indicate that events do not possess a distinct multifractal signature compared to noise. showing negligible difference between "event" and "noise" phases but signif￾icant difference between detectors. 3.2.4. Distinguishing Signal from Noise: Shuffled Data Analysis The shuffle-surrogate test ( [PITH_FULL_IMAGE:figures/full_fig_p028_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Hurst exponent for shuffled data. The collapse to a constant value confirms [PITH_FULL_IMAGE:figures/full_fig_p029_14.png] view at source ↗
read the original abstract

Multifractal Detrended Fluctuation Analysis (MFDFA) is a powerful and widely used technique for characterizing the scaling properties and long-range correlations of complex time series. However, its application often involves significant practical challenges, such as the subjective identification of scaling regions (crossovers) and the disambiguation of the physical origins of multifractality. We introduce MF-toolkit, a high-performance, parallelized Python library designed to address these challenges. It integrates three key innovations: (1) fully automatic crossover detection algorithms (CDV-A and SPIC), which remove operator bias and enhance reproducibility; (2) a built-in implementation of the Iterative Amplitude Adjusted Fourier Transform (IAAFT) for generating surrogate data, enabling the robust identification of the source of multifractality; and (3) a comprehensive suite for generating synthetic time series for rigorous validation. We demonstrate the rigor and utility of MF-toolkit through its application to characterize the multifractal properties of non-stationary noise in gravitational wave (LIGO) data. The MF-toolkit library offers a robust, efficient, and user-friendly tool for advanced time series analysis, facilitating more rigorous and reproducible research across physics and other data-intensive fields.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The manuscript introduces MF-toolkit, a high-performance parallelized Python library for multifractal detrended fluctuation analysis (MFDFA). Key features include fully automatic crossover detection via CDV-A and SPIC algorithms to eliminate operator bias, a built-in IAAFT implementation for surrogate-based identification of multifractality sources, and a suite for generating synthetic time series with known scaling properties for validation. The library is demonstrated on non-stationary noise from LIGO gravitational wave data.

Significance. The library's automated methods, parallel implementation, and explicit validation on controlled synthetic series with known crossovers represent a concrete contribution to reproducibility in MFDFA. If the reported performance on synthetic data extends to real non-stationary signals, the toolkit would be a useful resource for physics and time-series analysis communities. The inclusion of IAAFT surrogates and the LIGO application section directly address practical challenges in source identification and handling of gravitational-wave noise.

major comments (1)
  1. [§4] §4 (Validation on synthetic data): The reported accuracy metrics for CDV-A and SPIC (e.g., crossover position error and scaling-region identification rates) are shown only for stationary synthetic series; quantitative results on non-stationary synthetic series matching the LIGO noise characteristics are needed to support the claim that the detectors remain unbiased on real gravitational-wave data.
minor comments (2)
  1. [Figure 5] Figure 5 (LIGO application): The scaling plots would benefit from explicit annotation of the automatically detected crossover points to allow direct visual comparison with the algorithm output.
  2. [Abstract] The abstract states that the algorithms 'remove operator bias' without referencing the specific quantitative improvement (e.g., variance reduction across users) shown in the validation tables.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive review and the recommendation for minor revision. We address the single major comment below.

read point-by-point responses

  1. Referee: [§4] §4 (Validation on synthetic data): The reported accuracy metrics for CDV-A and SPIC (e.g., crossover position error and scaling-region identification rates) are shown only for stationary synthetic series; quantitative results on non-stationary synthetic series matching the LIGO noise characteristics are needed to support the claim that the detectors remain unbiased on real gravitational-wave data.

    Authors: We agree that the quantitative validation metrics for CDV-A and SPIC were demonstrated on stationary synthetic series, while the LIGO application provides a real-world demonstration on non-stationary data. To directly address this point, the revised manuscript will include additional quantitative results generated using MF-toolkit's synthetic series module. These will consist of non-stationary time series with controlled crossovers and multifractal properties calibrated to match the statistical characteristics of LIGO noise (e.g., non-stationarity and noise spectrum). This will furnish explicit accuracy metrics (crossover position error and scaling-region identification rates) for the detectors under conditions representative of gravitational-wave data. revision: yes

Circularity Check

0 steps flagged

No significant circularity; software library with explicit validation

full rationale

The paper introduces MF-toolkit as a software library implementing MFDFA with new automatic crossover detectors (CDV-A, SPIC), IAAFT surrogates, and synthetic data generators. It supplies algorithmic descriptions, pseudocode-level implementation details, quantitative tests on controlled synthetic series with known scaling breaks, and an explicit LIGO application. No mathematical derivation chain exists that reduces predictions or uniqueness claims to fitted parameters, self-definitions, or load-bearing self-citations. Validation is performed externally via synthetic benchmarks rather than by construction from the same data. The central claims rest on reproducible code and empirical results, not circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The work is a software library paper rather than a theoretical derivation; it relies on the established MFDFA framework and standard surrogate techniques without introducing new free parameters, ad-hoc axioms, or postulated entities.

axioms (1)
  • domain assumption Standard assumptions underlying MFDFA for characterizing scaling and long-range correlations in time series
    The library automates and extends the existing MFDFA method rather than re-deriving its foundations.

pith-pipeline@v0.9.0 · 5555 in / 1276 out tokens · 32191 ms · 2026-05-10T06:54:49.764148+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

28 extracted references · 28 canonical work pages

  1. [1]

    Long correlations and truncated levy walks applied to the study latin-american market in- dices,

    S. Jaroszewicz, M. C. Mariani, and M. Ferraro, “Long correlations and truncated levy walks applied to the study latin-american market in- dices,”Physica A: Statistical Mechanics and its Applications, vol. 355, pp. 461–474, 2005

    work page 2005

  2. [2]

    Mosaic organization of dna nucleotides,

    “Mosaic organization of dna nucleotides,” tech. rep., 1994

    work page 1994

  3. [3]

    Scaling detection in extrachromosomal dna,

    S. Jaroszewicz, M. C. Mariani, O. K. Tweneboah, and M. P. Beccar- Varela, “Scaling detection in extrachromosomal dna,”Fractals, vol. 30, no. 09, p. 2250194, 2022

    work page 2022

  4. [4]

    Characterising the interplay of dynamics and artefacts: a multifractal analysis of historical humpback whale recordings,

    N. Mendez, S. Jaroszewicz, O. K. Tweneboah, M. P. Beccar-Varela, and M. C. Mariani, “Characterising the interplay of dynamics and artefacts: a multifractal analysis of historical humpback whale recordings,”Bioa- coustics, vol. 35, no. 2, pp. 220–239, 2026

    work page 2026

  5. [5]

    Wavelet-based multifractal analysis of the el niño/southern oscillation, the indian ocean dipole and the north atlantic oscillation,

    F. Maruyama, K. Kai, and H. Morimoto, “Wavelet-based multifractal analysis of the el niño/southern oscillation, the indian ocean dipole and the north atlantic oscillation,”Scientific Online Letters on the Atmo- sphere, vol. 7, no. 1, pp. 65–68, 2011

    work page 2011

  6. [6]

    Multifractal analysis of the southern oscillation index,

    S. Jaroszewicz, M. C. Mariani, O. K. Tweneboah, and M. P. Beccar- Varela, “Multifractal analysis of the southern oscillation index,”Journal of Atmospheric and Solar-Terrestrial Physics, vol. 254, p. 106161, 2024

    work page 2024

  7. [7]

    Study of the tempo- ral correlations in the magnitude time series before major earthquakes in japan,

    P. A. Varotsos, N. V. Sarlis, and E. S. Skordas, “Study of the tempo- ral correlations in the magnitude time series before major earthquakes in japan,”Journal of Geophysical Research: Space Physics, vol. 119, pp. 9192–9206, 2014. 39

    work page 2014

  8. [8]

    Investigation of the temporal correlations be- tween earthquake magnitudes before the mexico m8.2 earthquake on 7 september 2017,

    N. V. Sarlis, E. S. Skordas, P. A. Varotsos, A. Ramírez-Rojas, and E. L. Flores-Márquez, “Investigation of the temporal correlations be- tween earthquake magnitudes before the mexico m8.2 earthquake on 7 september 2017,”Physica A: Statistical Mechanics and its Applications, vol. 517, pp. 475–483, 2019

    work page 2017

  9. [9]

    Spatial analysis of oil reservoirs using detrended fluctuation analysis of geophysical data,

    R. A. Ribeiro, M. V. M. Mata, L. S. Lucena, U. L. Fulco, and G. Corso, “Spatial analysis of oil reservoirs using detrended fluctuation analysis of geophysical data,”Nonlinear Processes in Geophysics, vol. 21, pp. 1043– 1049, 2014

    work page 2014

  10. [10]

    Fractional brownian motions, fractional noises and applications,

    B. B. Mandelbrot and J. W. Van Ness, “Fractional brownian motions, fractional noises and applications,”SIAM Review, vol. 10, pp. 422–437, 1968

    work page 1968

  11. [11]

    Long-range correlations and characterization of financial and volcanic time series,

    M. C. Mariani, P. K. Asante, M. A. M. Bhuiyan, M. P. Beccar-Varela, S. Jaroszewicz, and O. K. Tweneboah, “Long-range correlations and characterization of financial and volcanic time series,”Mathematics, vol. 8, 2020

    work page 2020

  12. [12]

    Self-similar models: Relationship between the diffusion entropy analysis, detrended fluctua- tion analysis and lévy models,

    M.C.Mariani, W.Kubin, P.K.Asante, O.K.Tweneboah, M.P.Beccar- Varela, S. Jaroszewicz, and H. Gonzalez-Huizar, “Self-similar models: Relationship between the diffusion entropy analysis, detrended fluctua- tion analysis and lévy models,”Mathematics, vol. 8, 2020

    work page 2020

  13. [13]

    Multifractal detrended ductuation anal- ysis of nonstationary time series,

    J. W. Kantelhardt, S. A. Zschiegner, E. Koscielny-Bunde, S. Havlin, A. Bunde, and H. E. Stanley, “Multifractal detrended ductuation anal- ysis of nonstationary time series,” tech. rep., 2002

    work page 2002

  14. [14]

    Introduction to multifractal detrended fluctuation analysis in matlab,

    E. A. Ihlen, “Introduction to multifractal detrended fluctuation analysis in matlab,”Frontiers in Physiology, vol. 3 JUN, 2012

    work page 2012

  15. [15]

    Fractal and multifractal time series,

    J. W. Kantelhardt, “Fractal and multifractal time series,”arXiv e-prints, 2008

    work page 2008

  16. [16]

    A criterion for the determination of optimal scaling ranges in dfa and mf-dfa,

    D. Gulich and L. Zunino, “A criterion for the determination of optimal scaling ranges in dfa and mf-dfa,”Physica A: Statistical Mechanics and its Applications, vol. 397, pp. 17–30, 2014. 40

    work page 2014

  17. [17]

    Multifractal signatures of gravitational waves detected by ligo,

    D. B. de Freitas, M. M. Nepomuceno, and J. De Medeiros, “Multifractal signatures of gravitational waves detected by ligo,”Proceedings of the International Astronomical Union, vol. 14, no. S346, pp. 468–473, 2018

    work page 2018

  18. [18]

    Multifractal chaos in gravita- tional waves from binary black hole mergers,

    F. Oliveira, I. Lucena, and J. Ramos, “Multifractal chaos in gravita- tional waves from binary black hole mergers,”Chaos, Solitons & Frac- tals, vol. 200, p. 116980, 2025

    work page 2025

  19. [19]

    Multifractal analysis of the gravitational waves,

    L. F. Chernogor, O. V. Lazorenko, and A. A. Onishchenko, “Multifractal analysis of the gravitational waves,”Bulletin of V. Kharkiv National University N. Karazin, no. 26, pp. 33–39, 2017

    work page 2017

  20. [20]

    Characterization of gravitational-wave detector noise with fractals,

    M. Cavaglia, “Characterization of gravitational-wave detector noise with fractals,”Classical and Quantum Gravity, vol. 39, no. 13, p. 135012, 2022

    work page 2022

  21. [21]

    Mfdfa: Effi- cient multifractal detrended fluctuation analysis in python,

    L. R. Gorjão, G. Hassan, J. Kurths, and D. Witthaut, “Mfdfa: Effi- cient multifractal detrended fluctuation analysis in python,”Computer Physics Communications, vol. 273, p. 108254, 4 2022

    work page 2022

  22. [22]

    Crossover detection based on variances of slope differences for multi- fractal detrended fluctuation analysis (mf-dfa),

    S. Moreno-Pulido, J. C. de la Torre, P. Ruiz, and P. Pavón-Domínguez, “Crossover detection based on variances of slope differences for multi- fractal detrended fluctuation analysis (mf-dfa),”Nonlinear Dynamics, vol. 113, pp. 7425–7457, 4 2025

    work page 2025

  23. [23]

    Detection of crossover time scales in multifrac- tal detrended fluctuation analysis,

    E. Ge and Y. Leung, “Detection of crossover time scales in multifrac- tal detrended fluctuation analysis,”Journal of Geographical Systems, vol. 15, pp. 115–147, 4 2013

    work page 2013

  24. [24]

    Surrogate time series,

    T. Schreiber and A. Schmitz, “Surrogate time series,” tech. rep., 2000

    work page 2000

  25. [25]

    Using surrogate data to detect nonlinearity in time series,

    J. Theiler, B. Galdrikian, A. Longtin, S. Eubank, and D. J. Farmer, “Using surrogate data to detect nonlinearity in time series,” inNonlinear Modeling and Forecasting(M. Casdagli and S. Eubank, eds.), pp. 163– 188, Redwood City, CA: Addison-Wesley, 1992

    work page 1992

  26. [26]

    Testsforhursteffect,

    R.B.DaviesandD.S.Harte, “Testsforhursteffect,”Biometrika, vol.74, no. 1, pp. 95–101, 1987

    work page 1987

  27. [27]

    Detecting long-range correlations with detrended fluc- tuation analysis,

    A. B. Jan W. Kantelhardt; Eva Koscielny-Bundea, Henio H.A. Regoa; Shlomo Havlinb, “Detecting long-range correlations with detrended fluc- tuation analysis,”Physica A, vol. 295, p. 441–454, 2001. 41

    work page 2001

  28. [28]

    Gravitational-wave transient catalogue,

    G. W. O. S. Center, “Gravitational-wave transient catalogue,” 2023. Date of access: October 10,2023. 42

    work page 2023