pith. machine review for the scientific record. sign in

arxiv: 2605.01624 · v1 · submitted 2026-05-02 · 🧮 math.AT · stat.AP· stat.ML

Recognition: 3 theorem links

· Lean Theorem

Persistent Homology of Time Series through Complex Networks

\.Ismail G\"uzel

Pith reviewed 2026-05-08 19:28 UTC · model grok-4.3

classification 🧮 math.AT stat.APstat.ML
keywords persistent homologytime series classificationcomplex networksvisibility graphsVietoris-Rips filtrationpersistence landscapesUCR benchmarksdiffusion distance
0
0 comments X

The pith

Converting time series to graphs and extracting persistent homology features yields classification performance that varies with graph construction and distance metric.

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

The paper develops a standardized pipeline that maps a univariate time series to a graph using one of five constructions from visibility, transition, or proximity families, derives a dissimilarity matrix, builds a Vietoris-Rips filtration to obtain persistence diagrams, and vectorizes those diagrams into fixed-length features via persistence landscapes and topological summary statistics. By holding the vectorization steps fixed across all variants, the work isolates the impact of the upstream graph construction and choice of distance on the graph. Experiments across twelve UCR benchmark datasets establish three concrete results: the best-performing construction changes with the discriminative structure of each signal, diffusion distance on the graph produces higher accuracy than shortest-path distance in every case examined, and the resulting features maintain useful performance as additive noise increases in the input series.

Core claim

The central claim is that a time series can be represented as a complex network via visibility, horizontal visibility, transition, or proximity constructions; a dissimilarity matrix can then be formed from the graph; a Vietoris-Rips filtration of that matrix produces persistence diagrams; and those diagrams can be turned into classification features by persistence landscapes together with summary statistics. When this pipeline is applied uniformly, classification results on twelve standard time-series collections show that no single graph construction is universally superior, that diffusion distance uniformly outperforms shortest-path distance, and that the extracted features degrade only in

What carries the argument

The standardized pipeline that converts a time series to a graph, extracts a dissimilarity matrix, computes its Vietoris-Rips persistence diagrams, and vectorizes the diagrams with persistence landscapes and summary statistics.

If this is right

  • Classification accuracy on a given time-series task will depend on selecting the graph construction whose structure best matches the discriminative patterns in that signal.
  • Replacing shortest-path distance with diffusion distance on the constructed graph will raise accuracy for every construction and every dataset tested.
  • The topological features extracted from the graphs will continue to support classification even after moderate amounts of noise are added to the original time series.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Users facing a new collection of time series may need to evaluate several graph constructions on a validation subset rather than adopting one default construction.
  • The observed noise robustness suggests the pipeline could be applied directly to sensor recordings that contain typical measurement fluctuations.
  • The same standardization approach could be used to compare additional graph constructions or alternative filtrations on the same benchmark collections.

Load-bearing premise

That holding the persistence-landscape and summary-statistic steps constant removes all interactions between graph construction and downstream feature quality, so performance gaps can be attributed only to the choice of graph and distance metric.

What would settle it

Repeating the twelve-dataset experiments while replacing persistence landscapes with a different vectorization such as persistence images and checking whether the relative ordering of the five graph constructions stays the same.

Figures

Figures reproduced from arXiv: 2605.01624 by \.Ismail G\"uzel.

Figure 1
Figure 1. Figure 1: Schematic for Section 2. Once a network has been constructed from the time series, the remaining view at source ↗
Figure 2
Figure 2. Figure 2: A single toy example summarizing the Section 1.2 route from graph to persistence. Panel (a) view at source ↗
Figure 3
Figure 3. Figure 3: Visibility-graph comparison for x = [1, 2, 1, 5, 2]. Panel (a): signal; node 4 (x4 = 5) is the global maximum. Panel (b): NVG in linear layout; long-range edges (1, 4) and (2, 4) appear as nested arcs. Panel (c): NVG redrawn in a graph layout with no spatial constraints; node 4 is the hub of degree 4. Panel (d): HVG in the same graph layout; the dashed grey edge marks the absent edge (1, 4), blocked becaus… view at source ↗
Figure 5
Figure 5. Figure 5: Coarse-grained state-space network for x = [1, 4, 7, 2, 5, 8, 3, 6, 1] with n = 3, τ = 1, b = 2. Panel (a): signal; the shaded bands mark bin 0 = [1, 4.5) (light) and bin 1 = [4.5, 8] (medium), with the dashed boundary line at 4.5. Panel (b): CGSSN with five state-nodes. The bold edge {3, 6} carries weight 2 because both directions appear once; all other edges carry weight 1. State 6 is the hub of degree 3… view at source ↗
Figure 6
Figure 6. Figure 6: k-nearest neighbor graph for x = [1, 2, 1, 5, 2, 3, 1] with n = 3, τ = 1, k = 2. Panel (a): signal; the shaded region marks the first sliding window of width n = 3. Panel (b): 2-NN graph; nodes placed at the first two delay-vector components (ξ1, ξ2) to reveal the geometric structure that drives edge formation. Edges connect pairs whose 3-D Euclidean distance is small enough to fall within each other’s 2-n… view at source ↗
Figure 8
Figure 8. Figure 8: Distance-matrix ablation: mean F1 averaged over the twelve datasets, for every (graph, distance) combination. Rows are graph constructions; columns are distance metrics. Colour encodes F1 (cividis; darker is lower). Each row’s winning distance is outlined in a bold black border. Hatched cells (n/a) denote combinations that are not applicable—unweighted graphs (HVG, NVG, k-NN) do not admit weight-based dist… view at source ↗
Figure 9
Figure 9. Figure 9: Noise robustness on Coffee: F1 versus SNR for each pipeline. Error bars show the standard deviation across five folds. OPN loses its clean-data advantage at moderate noise, while CGSSN degrades most gracefully. problems. 6 Conclusion and Future Directions The framework developed in this study provides a unified route from univariate time series to persistence-based features through complex networks. By sep… view at source ↗
Figure 10
Figure 10. Figure 10: Noise robustness on PowerCons: F1 versus SNR. CGSSN maintains the highest F1 at every noise level. HVG collapses to the majority-class baseline even on clean data. requires further work on both network design and topological interpretation. On the theoretical side, the Stability Theorem (Theorem 1) controls the persistence stage once the filtration is specified, but the stability of the full chain from ra… view at source ↗
read the original abstract

We present a unified pipeline for univariate time series classification via complex networks and persistent homology. A time series is mapped to a graph through one of five constructions across three families (visibility (natural and horizontal visibility graphs), transition, and proximity) and the graph is converted to a dissimilarity matrix from which a Vietoris-Rips filtration yields persistence diagrams. These diagrams are vectorized into fixed-length features through persistence landscapes and topological summary statistics. By standardizing the downstream processing, differences in classification performance are attributable to the network construction and distance metric alone. Experiments on twelve UCR benchmarks show that (i) no single construction dominates: the optimal graph type depends on the signal's discriminative structure; (ii) the graph distance metric is a first-order design choice, with diffusion distance uniformly outperforming shortest-path alternatives; and (iii) persistence-based features degrade gracefully under noise, consistent with the classical stability theorem of persistent homology.

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

2 major / 2 minor

Summary. The manuscript proposes a unified pipeline for univariate time series classification that maps each series to one of five complex-network constructions (natural/horizontal visibility graphs, transition graphs, proximity graphs), converts the graph to a dissimilarity matrix, computes a Vietoris-Rips persistence diagram, and vectorizes the diagram via persistence landscapes plus topological summary statistics. By fixing the downstream vectorization, performance differences across constructions and distance metrics (diffusion vs. shortest-path) are attributed to the upstream choices. Experiments on twelve UCR benchmarks show that no construction dominates, diffusion distance consistently outperforms shortest-path alternatives, and the resulting features degrade gracefully under additive noise.

Significance. If the empirical comparisons hold after addressing scale normalization, the work supplies a systematic, benchmark-driven evaluation of how graph-construction choices and distance metrics shape persistent-homology features for time series. The finding that diffusion distance is uniformly superior and that topological summaries remain stable under noise aligns with classical stability theorems and offers concrete guidance for practitioners choosing among visibility, transition, and proximity representations.

major comments (2)
  1. [Abstract and Methods] Abstract and Methods (pipeline description): the claim that 'standardizing the downstream processing' makes classification differences attributable solely to network construction and distance metric is not yet supported. The five constructions produce dissimilarity matrices with systematically different scales and sparsity (visibility graphs yield sparse, locally determined distances; proximity and transition graphs yield denser or probabilistic weights). Vietoris-Rips filtrations are sensitive to absolute distance values; without explicit per-matrix normalization (e.g., diameter scaling or min-max), birth/death times embed scale-dependent information that the fixed persistence landscapes and summary statistics can exploit, creating a hidden interaction between upstream construction and feature quality.
  2. [Results] Results section (benchmark tables): it is unclear whether the reported accuracy differences are statistically significant after correction for multiple comparisons across twelve datasets, five constructions, and two distance metrics, or whether hyperparameters and data splits were pre-specified. Without these details the claim that 'no single construction dominates' and that 'diffusion distance uniformly outperforms' cannot be evaluated for robustness.
minor comments (2)
  1. [Methods] The manuscript would benefit from an explicit table or subsection listing the precise parameter settings (e.g., visibility thresholds, proximity radii, transition-matrix construction) used for each of the five graph families.
  2. [Figures] Figure captions should state the exact noise model and SNR levels used in the robustness experiments to allow direct replication.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for these insightful comments on scale sensitivity and statistical robustness. We agree that both points identify areas where the current manuscript can be strengthened without altering its core contributions. We respond to each major comment below and will incorporate the suggested revisions.

read point-by-point responses

  1. Referee: [Abstract and Methods] Abstract and Methods (pipeline description): the claim that 'standardizing the downstream processing' makes classification differences attributable solely to network construction and distance metric is not yet supported. The five constructions produce dissimilarity matrices with systematically different scales and sparsity (visibility graphs yield sparse, locally determined distances; proximity and transition graphs yield denser or probabilistic weights). Vietoris-Rips filtrations are sensitive to absolute distance values; without explicit per-matrix normalization (e.g., diameter scaling or min-max), birth/death times embed scale-dependent information that the fixed persistence landscapes and summary statistics can exploit, creating a hidden interaction between upstream construction and feature quality.

    Authors: We agree that the absence of explicit per-matrix normalization leaves open the possibility that scale differences in the dissimilarity matrices influence the resulting persistence diagrams and downstream features. Although the vectorization step is held fixed, the Vietoris-Rips filtration itself depends on absolute distances. To eliminate this confounding factor, we will revise the Methods section to apply diameter normalization to every dissimilarity matrix (scaling all entries by the matrix diameter so that distances lie in [0,1]) before filtration. This change will be documented with pseudocode and will ensure that performance differences can be attributed more cleanly to the choice of network construction and distance metric. revision: yes

  2. Referee: [Results] Results section (benchmark tables): it is unclear whether the reported accuracy differences are statistically significant after correction for multiple comparisons across twelve datasets, five constructions, and two distance metrics, or whether hyperparameters and data splits were pre-specified. Without these details the claim that 'no single construction dominates' and that 'diffusion distance uniformly outperforms' cannot be evaluated for robustness.

    Authors: The train/test splits follow the pre-specified UCR protocols, and graph-construction hyperparameters were set to the literature defaults cited in the manuscript. However, we did not conduct formal significance testing or apply multiple-comparison corrections. In the revised Results section we will add a statistical analysis subsection that reports Wilcoxon signed-rank tests (or paired t-tests where appropriate) on the accuracy differences, together with Bonferroni correction across the 12 datasets × 5 constructions × 2 metrics comparisons. We will also state the hyperparameter choices explicitly. These additions will allow readers to assess the robustness of the claims that no construction dominates and that diffusion distance is uniformly superior. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical pipeline with standardized downstream processing

full rationale

The paper describes a fixed pipeline (five graph constructions from time series, dissimilarity matrices, Vietoris-Rips filtrations, persistence landscapes plus summary statistics) and reports comparative classification accuracies on twelve public UCR benchmarks. No mathematical derivation, uniqueness theorem, or prediction is claimed; performance differences are presented as experimental outcomes under the explicit methodological choice of standardized vectorization. No self-citations appear as load-bearing premises, no parameters are fitted then relabeled as predictions, and no ansatz or renaming reduces the central claims to inputs by construction. The standardization step is a design decision whose validity is testable against the benchmarks rather than tautological.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The approach rests on standard persistent homology theory and graph constructions from prior work; no new free parameters, axioms, or invented entities are introduced in the abstract.

axioms (2)
  • standard math Vietoris-Rips filtration and persistence diagrams are well-defined on the dissimilarity matrices derived from the graphs.
    Invoked when converting graphs to persistence diagrams.
  • domain assumption Persistence landscapes and topological summary statistics provide stable, fixed-length features.
    Used for vectorization step.

pith-pipeline@v0.9.0 · 5450 in / 1301 out tokens · 44042 ms · 2026-05-08T19:28:37.704441+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

44 extracted references · 36 canonical work pages

  1. [1]

    The ring of algebraic functions on persistence bar codes.Homology, Homotopy and Applications, 18(1):381–402, 2016

    Aaron Adcock, Erik Carlsson, and Gunnar Carlsson. The ring of algebraic functions on persistence bar codes.Homology, Homotopy and Applications, 18(1):381–402, 2016. doi: 10.4310/HHA.2016.v18.n1.a21

    work page doi:10.4310/hha.2016.v18.n1.a21 2016

  2. [2]

    Akingbade, Marian Gidea, Matteo Manzi, and Vahid Nateghi

    Samuel W. Akingbade, Marian Gidea, Matteo Manzi, and Vahid Nateghi. Why topological data analysis detects financial bubbles?Communications in Nonlinear Science and Numerical Simulation, 128:107665, 2024. doi: 10.1016/j.cnsns.2023.107665

    work page doi:10.1016/j.cnsns.2023.107665 2024

  3. [3]

    Multiscale permutation entropy of physiological time series

    Wajid Aziz and Muhammad Arif. Multiscale permutation entropy of physiological time series. In9th International Multitopic Conference (INMIC). IEEE, 2005. doi: 10.1109/INMIC.2005. 334494

    work page doi:10.1109/inmic.2005 2005

  4. [4]

    The great time series classification bake off: a review and experimental evaluation of recent algorithmic advances.Data Mining and Knowledge Discovery, 31(3):606–660, 2017

    Anthony Bagnall, Jason Lines, Aaron Bostrom, James Large, and Eamonn Keogh. The great time series classification bake off: a review and experimental evaluation of recent algorithmic advances.Data Mining and Knowledge Discovery, 31(3):606–660, 2017. doi: 10.1007/s10618-016-0483-9

    work page doi:10.1007/s10618-016-0483-9 2017

  5. [5]

    Ollivier and W

    Christoph Bandt and Bernd Pompe. Permutation entropy: a natural complexity measure for time series.Physical Review Letters, 88(17):174102, 2002. doi: 10.1103/PhysRevLett.88. 174102

    work page doi:10.1103/physrevlett.88 2002

  6. [6]

    Nicolas Bonneel, Julien Rabin, Gabriel Peyr ´e, and Hanspeter Pfister

    Ulrich Bauer. Ripser: efficient computation of Vietoris–Rips persistence barcodes.Journal of Applied and Computational Topology, 5(3):391–423, 2021. doi: 10.1007/s41468-021-00071-5

    work page doi:10.1007/s41468-021-00071-5 2021

  7. [7]

    Statistical topological data analysis using persistence landscapes.Jour- nal of Machine Learning Research, 16:77–102, 2015

    Peter Bubenik. Statistical topological data analysis using persistence landscapes.Jour- nal of Machine Learning Research, 16:77–102, 2015. URLhttp://jmlr.org/papers/v16/ bubenik15a.html

    2015

  8. [8]

    An introduction to topological data analysis: funda- mental and practical aspects for data scientists.Frontiers in Artificial Intelligence, 4:667963,

    Frédéric Chazal and Bertrand Michel. An introduction to topological data analysis: funda- mental and practical aspects for data scientists.Frontiers in Artificial Intelligence, 4:667963,

  9. [9]

    doi: 10.3389/frai.2021.667963

    work page doi:10.3389/frai.2021.667963 2021

  10. [10]

    David Cohen-Steiner, Herbert Edelsbrunner, John Harer, and Yuriy Mileyko

    DavidCohen-Steiner, HerbertEdelsbrunner, andJohnHarer. Stabilityofpersistencediagrams. Discrete & Computational Geometry, 37(1):103–120, 2007. doi: 10.1007/s00454-006-1276-5

    work page doi:10.1007/s00454-006-1276-5 2007

  11. [11]

    Applied and computational harmonic analysis21(1), 5–30 (2006) https://doi.org/10.1016/j.acha.2006.04.006

    RonaldR.CoifmanandStéphaneLafon. Diffusionmaps.Applied and Computational Harmonic Analysis, 21(1):5–30, 2006. doi: 10.1016/j.acha.2006.04.006

    work page doi:10.1016/j.acha.2006.04.006 2006

  12. [12]

    The UCR time series archive.IEEE/CAA Journal of Automatica Sinica, 6(6):1293–1305, 2019

    Hoang Anh Dau, Anthony Bagnall, Kaveh Kamgar, Chin-Chia Michael Yeh, Yan Zhu, Shaghayegh Gharghabi, Chotirat Ann Ratanamahatana, and Eamonn Keogh. The UCR time series archive.IEEE/CAA Journal of Automatica Sinica, 6(6):1293–1305, 2019. doi: 10.1109/JAS.2019.1911747. 30

    work page doi:10.1109/jas.2019.1911747 2019

  13. [13]

    Enhanc- ing financial time series forecasting through topological data analysis.Neural Computing and Applications, 2025

    Luiz Carlos de Jesus Jr., Francisco Fernández-Navarro, and Mariano Carbonero-Ruz. Enhanc- ing financial time series forecasting through topological data analysis.Neural Computing and Applications, 2025. doi: 10.1007/s00521-024-10787-x

    work page doi:10.1007/s00521-024-10787-x 2025

  14. [14]

    Cambridge University Press, 2022

    Tamal Krishna Dey and Yusu Wang.Computational Topology for Data Analysis. Cambridge University Press, 2022. doi: 10.1017/9781009099950

    work page doi:10.1017/9781009099950 2022

  15. [15]

    Donner, Yong Zou, Jonathan F

    Reik V. Donner, Yong Zou, Jonathan F. Donges, Norbert Marwan, and Jürgen Kurths. Recur- rence networks—a novel paradigm for nonlinear time series analysis.New Journal of Physics, 12:033025, 2010. doi: 10.1088/1367-2630/12/3/033025

    work page doi:10.1088/1367-2630/12/3/033025 2010

  16. [16]

    Topological persistence and simplification.Discrete & Computational Geometry, 28(4):511–533, 2002

    Herbert Edelsbrunner, David Letscher, and Afra Zomorodian. Topological persistence and simplification.Discrete & Computational Geometry, 28(4):511–533, 2002. doi: 10.1007/ s00454-002-2885-2

    2002

  17. [17]

    Complex network from time series based on phase space reconstruction.Chaos, 19(3):033137, 2009

    Zhong-Ke Gao and Ning-De Jin. Complex network from time series based on phase space reconstruction.Chaos, 19(3):033137, 2009. doi: 10.1063/1.3227736

    work page doi:10.1063/1.3227736 2009

  18. [18]

    Classification of stochastic processes with topological data analysis.Concurrency and Computation: Practice and Experience, 35(24):e7732, 2023

    İsmail Güzel and Atabey Kaygun. Classification of stochastic processes with topological data analysis.Concurrency and Computation: Practice and Experience, 35(24):e7732, 2023. doi: 10.1002/cpe.7732

    work page doi:10.1002/cpe.7732 2023

  19. [19]

    Persistent homology of featured time series data and its applications.AIMS Mathematics, 9(10):27028–27057, 2024

    Eunwoo Heo and Jae-Hun Jung. Persistent homology of featured time series data and its applications.AIMS Mathematics, 9(10):27028–27057, 2024. doi: 10.3934/math.20241315

    work page doi:10.3934/math.20241315 2024

  20. [20]

    Persistent homology of complex net- works.Journal of Statistical Mechanics: Theory and Experiment, 2009(3):P03034, 2009

    Danijela Horak, Slobodan Maletić, and Milan Rajković. Persistent homology of complex net- works.Journal of Statistical Mechanics: Theory and Experiment, 2009(3):P03034, 2009. doi: 10.1088/1742-5468/2009/03/P03034

    work page doi:10.1088/1742-5468/2009/03/p03034 2009

  21. [21]

    Time series analysis using persistent homology of distance matrix.Non- linear Theory and Its Applications, IEICE, 14(2):79–91, 2023

    Takashi Ichinomiya. Time series analysis using persistent homology of distance matrix.Non- linear Theory and Its Applications, IEICE, 14(2):79–91, 2023. doi: 10.1587/nolta.14.79

    work page doi:10.1587/nolta.14.79 2023

  22. [22]

    Machine learning of time series data using persistent homology.Scientific Reports, 15(15), 2025

    Takashi Ichinomiya. Machine learning of time series data using persistent homology.Scientific Reports, 15(15), 2025. doi: 10.1038/s41598-025-06551-3

    work page doi:10.1038/s41598-025-06551-3 2025

  23. [23]

    Time series classification via topological data analysis

    Alperen Karan and Atabey Kaygun. Time series classification via topological data analysis. Expert Systems with Applications, 183:115326, 2021. doi: 10.1016/j.eswa.2021.115326

    work page doi:10.1016/j.eswa.2021.115326 2021

  24. [24]

    Kennel, Reggie Brown, and Henry D

    Matthew B. Kennel, Reggie Brown, and Henry D. I. Abarbanel. Determining embedding dimension for phase-space reconstruction using a geometrical construction.Physical Review A, 45(6):3403–3411, 1992. doi: 10.1103/PhysRevA.45.3403

    work page doi:10.1103/physreva.45.3403 1992

  25. [25]

    Khasawneh, Elizabeth Munch, Danielle Barnes, Max M

    Firas A. Khasawneh, Elizabeth Munch, Danielle Barnes, Max M. Chumley, İsmail Güzel, Audun D. Myers, Sunia Tanweer, Sarah Tymochko, and Melih Yesilli. Teaspoon: A Python package for topological signal processing.Journal of Open Source Software, 10(107):7243,

  26. [26]

    doi: 10.21105/joss.07243

    work page doi:10.21105/joss.07243

  27. [27]

    From time series to complex networks: the visibility graph.Proceedings of the National Academy of Sciences, 105(13):4972–4975, 2008

    LucasLacasa, BartoloLuque, FernandoBallesteros, JordiLuque, andJuanCarlosNuno. From time series to complex networks: the visibility graph.Proceedings of the National Academy of Sciences, 105(13):4972–4975, 2008. doi: 10.1073/pnas.0709247105. 31

    work page doi:10.1073/pnas.0709247105 2008

  28. [28]

    Zixin Lin, Nur Fariha Syaqina Zulkepli, Mohd Shareduwan Mohd Kasihmuddin, and R. U. Gobithaasan. Time-series forecasting via topological information supervised framework with efficient topological feature learning.Expert Systems with Applications, 299:130032, 2026. doi: 10.1016/j.eswa.2025.130032

    work page doi:10.1016/j.eswa.2025.130032 2026

  29. [29]

    Horizontal visibility graphs: exact results for random time series.Physical Review E, 80(4):046103, 2009

    Bartolo Luque, Lucas Lacasa, Fernando Ballesteros, and Jordi Luque. Horizontal visibility graphs: exact results for random time series.Physical Review E, 80(4):046103, 2009. doi: 10.1103/PhysRevE.80.046103

    work page doi:10.1103/physreve.80.046103 2009

  30. [30]

    Clustering and classification of time series using topological data analysis with applications to finance.Expert Systems with Applications, 162: 113868, 2020

    Sourav Majumdar and Arnab Kumar Laha. Clustering and classification of time series using topological data analysis with applications to finance.Expert Systems with Applications, 162: 113868, 2020. doi: 10.1016/j.eswa.2020.113868

    work page doi:10.1016/j.eswa.2020.113868 2020

  31. [31]

    Time lagged ordinal partition networks for capturing dynamics of continuous dynamical systems.Chaos, 25(5):053101, 2015

    Michael McCullough, Michael Small, Thomas Stemler, and Herbert Ho-Ching Iu. Time lagged ordinal partition networks for capturing dynamics of continuous dynamical systems.Chaos, 25(5):053101, 2015. doi: 10.1063/1.4919075

    work page doi:10.1063/1.4919075 2015

  32. [32]

    Khasawneh

    Audun Myers, Elizabeth Munch, and Firas A. Khasawneh. Persistent homology of complex networks for dynamic state detection.Physical Review E, 100(2):022314, 2019. doi: 10.1103/ PhysRevE.100.022314

    2019

  33. [33]

    Khasawneh, and Elizabeth Munch

    Audun Myers, Firas A. Khasawneh, and Elizabeth Munch. Persistence of weighted ordinal partition networks for dynamic state detection.SIAM Journal on Applied Dynamical Systems, 22(1):65–89, 2023. doi: 10.1137/22M1476848

    work page doi:10.1137/22m1476848 2023

  34. [34]

    Myers, Firas A

    Audun D. Myers, Firas A. Khasawneh, and Brittany T. Fasy. ANAPT: additive noise analysis for persistence thresholding.Foundations of Data Science, 4(2):243–269, 2022. doi: 10.3934/ fods.2022005

    2022

  35. [35]

    Myers, Max M

    Audun D. Myers, Max M. Chumley, Firas A. Khasawneh, and Elizabeth Munch. Persistent homology of coarse-grained state-space networks.Physical Review E, 107(3):034303, 2023. doi: 10.1103/PhysRevE.107.034303

    work page doi:10.1103/physreve.107.034303 2023

  36. [36]

    Myers, Max M

    Audun D. Myers, Max M. Chumley, and Firas A. Khasawneh. Delay parameter selection in permutation entropy using topological data analysis.La Matematica, 3(3):1103–1136, 2024. doi: 10.1007/s44007-024-00110-4

    work page doi:10.1007/s44007-024-00110-4 2024

  37. [37]

    Perea and John Harer

    Jose A. Perea and John Harer. Sliding windows and persistence: an application of topological methods to signal analysis.Foundations of Computational Mathematics, 15(3):799–838, 2015. doi: 10.1007/s10208-014-9206-z

    work page doi:10.1007/s10208-014-9206-z 2015

  38. [38]

    CatBoost: unbiased boosting with categorical features

    Liudmila Prokhorenkova, Gleb Gusev, Aleksandr Vorobev, Anna Veronika Dorogush, and Andrey Gulin. CatBoost: unbiased boosting with categorical features. InAdvances in Neural Information Processing Systems, volume 31, 2018

    2018

  39. [39]

    An introduction to persistent homology for time series

    Nalini Ravishanker and Renjie Chen. An introduction to persistent homology for time series. WIREs Computational Statistics, 13(3):e1548, 2021. doi: 10.1002/wics.1548

    work page doi:10.1002/wics.1548 2021

  40. [40]

    Time series analysis via network science: concepts and algorithms.WIREs Data Mining and Knowledge Discovery, 11(3):e1404, 2021

    Vanessa Freitas Silva, Maria Eduarda Silva, Pedro Ribeiro, and Fernando Silva. Time series analysis via network science: concepts and algorithms.WIREs Data Mining and Knowledge Discovery, 11(3):e1404, 2021. doi: 10.1002/widm.1404. 32

    work page doi:10.1002/widm.1404 2021

  41. [41]

    Detecting strange attractors in turbulence

    Floris Takens. Detecting strange attractors in turbulence. InDynamical Systems and Turbu- lence, Warwick 1980, volume 898 ofLecture Notes in Mathematics, pages 366–381. Springer,

    1980

  42. [42]

    doi: 10.1007/BFb0091924

    work page doi:10.1007/bfb0091924

  43. [43]

    A., Someya, T

    Joshua B. Tenenbaum, Vin de Silva, and John C. Langford. A global geometric framework for nonlinear dimensionality reduction.Science, 290(5500):2319–2323, 2000. doi: 10.1126/science. 290.5500.2319

    work page doi:10.1126/science 2000

  44. [44]

    From time series to complex networks: the phase space coarse graining.Physica A: Statistical Mechanics and its Applications, 461:456–468, 2016

    Minggang Wang and Lixin Tian. From time series to complex networks: the phase space coarse graining.Physica A: Statistical Mechanics and its Applications, 461:456–468, 2016. doi: 10.1016/j.physa.2016.06.028. 33

    work page doi:10.1016/j.physa.2016.06.028 2016