Pith. sign in

REVIEW 2 major objections 6 minor 53 references

Online Change-Point Detection with Persistent Laplacian Features

T0 review · 2 major / 6 minor · reviewed 2026-08-02 · deepseek-v4-flash

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

Pith's one-line read Persistent Laplacian spectra catch changes that Betti counts miss, with delay of order log A over squared separation.

desk verdict New topological detector with a clean oracle theory; the plug-in weak-dependence guarantee is the main gap. read the letter →

arxiv 2607.08635 v2 pith:PHUPYOIF submitted 2026-07-09 stat.ME

Shiheng Nie , Yunguang Yue This is my paper · ORCID
classification stat.ME MSC 62L1055N3162M10
keywords change-pointdetectionpersistentLaplacianpositivespectrumCUSUMBettinumberstopologicaldataanalysisridge-whitenedseparationaveragerunlength
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 PL-CUSUM, an online change-point detector that turns sliding windows of a high-dimensional time series into point clouds and then into two topological features: persistent Betti numbers and the positive eigenvalues of persistent Laplacians. The authors argue the spectral part records within-scale connectivity and geometric information that Betti numbers drop, so changes that leave Betti counts identical can still be detected. They prove that, in a finite-support local model, the oracle CUSUM using the full feature has detection delay at most (2 log A)/γ²(1+o(1)) while the local minimax lower bound is the same order, so the logarithmic delay law is optimal up to constants. For the plug-in version with estimated parameters, they give finite-horizon false-alarm and delay bounds under a sub-Gaussian dependence condition. If the claims hold, the method offers a nonparametric way to monitor nonlinear, high-dimensional processes with the same asymptotic performance as a likelihood-ratio CUSUM.

What carries the argument

The engine is the persistent combinatorial Laplacian at a scale pair (a,b): L_q^{a,b} = δ(δ*) + (∂)*∂, whose zero-eigenvalue multiplicity equals the persistent Betti number β_q^{a,b}. The positive eigenvalues of L_q^{a,b}, truncated to J per homological level, form the spectrum vector that carries geometric information beyond homological counts. These features feed a ridge-whitened projection: W_ρ = (Σ + ρI)^{-1/2} defines the separation γ_ρ = ||W_ρ(μ_1 - μ_0)||, and the scalar score s_n = u^T W_ρ X_n with midpoint reference c gives drift ±γ_ρ/2 under the two states; Page's CUSUM recursion then accumulates this score. The theory ties the detection-delay bound precisely to γ_ρ, making the whi

What would settle it

Numerically test Eq. (3) on a geometrically beta-mixing process with strong serial dependence: compute the moment-generating function of centered partial sums of the plug-in scores and check the exponential bound with a fixed C_dep. If the MGF exceeds exp(C_dep m σ² λ²/2), or if the finite-horizon false-alarm probability at the Prop. 2 threshold exceeds α, the plug-in guarantee is falsified. A second decisive check: reproduce the tree-family experiment with a Betti-only detector; any nonzero detection probability would contradict Proposition 1.

Watch

Extended reading notes

Core claim

The paper's central claim is that the positive spectrum of the persistent combinatorial Laplacian carries change-point information absent from persistent Betti numbers, and that a CUSUM built on it is near-optimal. For any tree on m vertices, the paper constructs point clouds with identical persistent Betti vectors for every scale pair but different positive Laplacian spectra whenever the sum of squared degrees differs; the path and star trees realize this. In the finite-support exponential-tilting model, the oracle likelihood-ratio CUSUM using the full PL feature attains ARL0 ≥ A and worst-case delay ≤ 2 log A / γ² (1+o(1)), while the local minimax lower bound is (2/C)(log A / γ²)(1+o(1)):

Load-bearing premise

The finite-horizon false-alarm and delay bounds rest on Eq. (3), the assumption that, after Phase I estimation, centered partial sums of the plug-in scores are sub-Gaussian with a constant C_dep; the paper derives this condition only for independent or q-dependent evaluation scores, not for the geometrically beta-mixing case it advertises.

Editorial extensions

If this is right

  • Detection delay inherits the classical logarithmic law: to raise the average run length by a factor, delay grows only as log A / γ_ρ^2, so PL-CUSUM is asymptotically as fast as an oracle that knows the post-change distribution.
  • Persistent Betti vectors alone are provably blind to some changes in this model; including the positive PL spectrum is necessary and sufficient for those changes, so feature sets for topological monitoring should include spectral information.
  • With finite Phase I training, the plug-in score preserves the order as long as the estimation error e is below γ_ρ/2; the effective separation γ_ρ - 2e and the dependence constant C_dep enter exactly where one would expect.
  • Spectral truncation acts as a regularizer: when the feature dimension exceeds the Phase I sample size, keeping the first 5–20 positive eigenvalues retains detection probability, whereas using the full spectrum degrades.
  • The two-phase calibration with moving-block bootstrap keeps the finite-horizon false-alarm probability near its budget in experiments, supporting use in online industrial monitoring.

Reading between the lines

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

  • The ridge-whitened CUSUM shell is feature-agnostic; a testable extension is to substitute other topological summaries (persistence landscapes, persistence images, or spectral variants of zigzag persistence) and check whether the same log A/γ² law holds.
  • Because the proof of Proposition 1 only uses graph Laplacians of trees, the separation result likely extends to any pair of finite metric spaces with same persistent Betti vectors but different spectral invariants; this suggests a family of synthetic benchmarks for non-homological change detection.
  • The finite-horizon theory explicitly assumes a sub-Gaussian concentration condition (Eq. 3) that the paper proves only for independent or q-dependent scores; closing the gap for general geometrically beta-mixing processes is a natural next step.
  • The paper itself notes that reordering delay vectors inside a window is invisible to the PL features; a transition-sensitive variant with directed or weighted PL operators would extend detection to changes in dynamics with unchanged marginals.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 6 minor

Summary. The paper proposes PL-CUSUM, a two-phase online change-point detection method for high-dimensional nonlinear time series. Each monitoring window is delay-embedded into a point cloud; PL-CUSUM extracts persistent Betti vectors and positive persistent Laplacian spectra from the same Vietoris–Rips filtration, projects these features through a ridge-whitened linear score, and feeds the score into Page's CUSUM recursion. The theoretical part proves, under a finite-support exponential-tilting local model, that an oracle LLR-CUSUM based on the full PL feature has detection delay of order log A / γ_ρ^2, and that a local minimax lower bound has the same order. It also gives finite-horizon false-alarm and expected-delay bounds for the plug-in whitened score under a sub-Gaussian partial-sum condition (Eq. (3)), and constructs tree point clouds with identical persistent Betti vectors but different positive PL spectra. The empirical section reports synthetic tree-family experiments, sensitivity analyses, and real-data comparisons on SWaT and Electric Motor Vibrations.

Significance. If the results hold, the paper makes a useful contribution by introducing persistent Laplacian spectra into online change-point detection and by providing a rigorous oracle/minimax benchmark of order log A / γ_ρ^2. The tree-family construction in Proposition 1 is an elegant and concrete separation result: it shows that positive PL spectra can carry change information invisible to persistent Betti vectors. The experiments are extensive and appear carefully designed, with disjoint training/calibration/evaluation partitions, MBB calibration, and reproducibility code in a public repository. The paper is also candid about the limitations of the finite-support local model and the marginal-distribution scope of the features. The main weakness is that the advertised plug-in finite-horizon guarantees under general weak dependence rest on an unverified concentration assumption, Eq. (3).

major comments (2)
  1. [§4.4, Eq. (3)] The finite-horizon plug-in guarantees in Proposition 2 and Corollary 2 are conditional on Eq. (3), a sub-Gaussian partial-sum condition for the centered one-step scores. The paper promises geometric beta-mixing in §3.1 and Proposition 6 shows that beta-mixing transfers to the score sequence, but Proposition 6 does not imply Eq. (3). Proposition 10 supplies explicit values of C_dep only for independent or q-dependent scores. Thus, as written, the plug-in false-alarm and delay bounds are not established for the geometric beta-mixing case advertised in the abstract and §3.1. This is a load-bearing gap in the paper's finite-sample weak-dependence claims. A Bernstein-type concentration inequality for bounded geometrically beta-mixing sequences would close the gap; otherwise the claims should be restricted to the independent/q-dependent cases.
  2. [§4.2, Theorem 1] The oracle upper bound in Theorem 1 is proved for the LLR-CUSUM that knows P_0 and P_θ, not for the ridge-whitened PL-CUSUM recursion itself. The abstract phrase 'oracle upper bound' is technically accurate, and Proposition 2 provides a separate plug-in guarantee for PL-CUSUM, but the distinction should be made more prominent. In particular, Theorem 1 and Corollary 1 do not by themselves establish that PL-CUSUM attains the local minimax rate; they establish that the rate is of order log A / γ_ρ^2 for the oracle benchmark. The current framing risks overstating what is proved about the proposed method.
minor comments (6)
  1. [§4.4, Proposition 2 statement] The threshold formula contains ambiguous parentheses. It should read η ≥ (C_dep σ^2 / γ_eff) log( N(N+1)/(2α) ).
  2. [§3.4, Algorithm 1] The expression for the empirical false-alarm probability, [FAR_N(η) = ..., has a stray bracket. Also, line 14 'return' is unclear: it could specify that the flag is returned to the selection step.
  3. [§4.4, Proposition 9] The proof invokes 'vector Hoeffding concentration' and the matrix Bernstein inequality but does not spell out how the same δ is allocated to the mean and covariance events. Since only the rate with constants depending on R, ρ, ρ_mult, γ_ρ is claimed, this is easy to fix but should be made explicit.
  4. [§6, Discussion] The last paragraph correctly notes that PL features are invariant to reordering of delay vectors within a window and that only marginal-distribution changes are covered. This is an important scope restriction; it would help to state it earlier, when the local model is introduced in §4.1.
  5. [Figure 3] The heatmap cells are readable, but the caption should describe the color scale and the meaning of the reference value γ_ρ,ref in the right column.
  6. [Table D.7] For SWaT, the selected scale pair (14.051, 2906.380) spans a very wide range. Reporting the scale grid used in the experiment would make this selection interpretable.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the oracle upper/lower bounds are standard LLR-CUSUM information bounds with explicit separation constants, the tree-family spectral separation is an independent construction, and the plug-in results are explicitly conditioned on an assumed concentration condition rather than fitted.

full rationale

The paper's core claims are self-contained and not circular. Theorem 1's oracle upper bound follows from the standard CUSUM renewal/LLR argument: Lemma 4 gives ARL0 >= e^eta, Proposition 3 gives delay <= (eta+o(1))/I, and Corollary 3 converts the KL information to the ridge-whitened separation via I = (kappa/2) gamma^2, with 1 <= kappa <= 1+rho/lambda_min. The delay bound log A/gamma^2 is therefore a genuine theorem relating a feature-space separation to a stopping-time performance measure; gamma is not defined in terms of delay. The minimax lower bound (Proposition 4) is an independent change-of-measure argument that applies to all stopping times in C_A, not just to the proposed procedure. Proposition 1 is also independent: it constructs tree metric realizations, proves Betti-vector invariance through a levelwise dominated-vertex collapse (Lemma 1 and Remark 1), and proves spectrum difference via tr(L_T^2) = sum deg^2 + 2(m-1), so the claim that positive PL spectra can carry information invisible to Betti counts is not assumed but exhibited. The plug-in finite-horizon results (Proposition 2, Corollary 2) are conditional theorems: they explicitly assume Eq. (3), a sub-Gaussian partial-sum condition with constant Cdep. Proposition 10 derives Cdep for independent and q-dependent scores, but for the geometric beta-mixing case promised in Section 3.1 no derivation of Eq. (3) is supplied. That is a support gap in the advertised weak-dependence claim, not a circular reduction: the paper does not assume the conclusion, and the effect is confined to the plug-in dependence result, not to Theorem 1 or Corollary 1. There are no author self-citations used as load-bearing evidence, no fitted parameter renamed as a prediction, and no uniqueness theorem imported from the authors' prior work.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central theoretical claim rests on the finite-support exponential-tilting model, the beta-mixing domain assumption, and the unproven-for-mixing concentration condition Eq. (3). The method itself has several user-selected free parameters (L, J, ρ_mult, scale pairs, control limit) that influence γρ and therefore the delay rate. No new physical or hidden entities are introduced.

free parameters (5)
  • Delay-embedding dimension L = 1 (SWaT), 3–6 (Motor)
    User-chosen per task; changes the point-cloud geometry and the information captured by PL features.
  • Spectral truncation level J = 31–61 (real data); 5–20 recommended in high dimensions
    Controls feature dimension; experiments show full spectra degrade when p ≫ Ntr, so J is a tuned regularizer.
  • Ridge multiplier ρ_mult = 2 (SWaT), 0.125 (Motor)
    Selected from {0.125,0.5,2,8,32}; directly sets ρ and thus γρ and the theoretical rates.
  • Scale grid and scale-pair set S = 4 or 8 scale pairs; selected pairs reported in Table D.7
    The finite scale grid makes features finite-range; the scale pair is selected by largest γ̂, so the detection rate depends on this choice.
  • Control limit η = Calibrated by moving-block bootstrap to meet α
    Threshold controlling the ARL/delay tradeoff; not a model parameter but essential for false-alarm control.
assumptions (4)
  • ad hoc to paper Post-change feature distribution is an exponential tilt of P0: Pθ(i) ∝ P0(i) exp(θᵀxᵢ).
    Section 4.1; the entire oracle minimax theory is restricted to this finite-support local model, which real changes need not follow.
  • domain assumption Raw observations are strictly stationary and geometrically β-mixing under each state.
    Section 3.1; used to transfer stationarity and weak dependence to point clouds, features, and plug-in increments. Not verified on real data.
  • ad hoc to paper Equation (3): conditional on Phase I, centered partial sums of scores are sub-Gaussian with constant Cdep.
    Section 4.4, Eq. (3); assumed in Proposition 2 but proven only for independent/q-dependent scores in Proposition 10, not for general geometric β-mixing.
  • standard math Standard CUSUM/martingale, exponential-family, Hoeffding, Bernstein, and matrix Lipschitz results.
    Used in the appendix proofs; standard results from Wald, Lorden, Hoeffding, Tropp, Bhatia, and van der Vaart.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Online Change-Point Detection with Persistent Laplacian Features." pith.science (2026). https://pith.science/paper/PHUPYOIF

@misc{pith2026260708635,
  author       = {Pith},
  title        = {Pith review of: Online Change-Point Detection with Persistent Laplacian Features},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PHUPYOIF}},
  note         = {Machine review of arXiv:2607.08635}
}
read the original abstract

Online change-point detection in high-dimensional nonlinear time series faces two challenges. The underlying distributions are difficult to model, and state changes are difficult to characterize. We propose persistent Laplacian cumulative sum (PL-CUSUM) to address these challenges. PL-CUSUM maps delay-embedded sliding windows to point clouds. It extracts persistent Betti vectors and the positive spectra of persistent Laplacians from the same Vietoris-Rips filtration. A ridge-whitened projection converts these features into a scalar score. Page's CUSUM recursion then accumulates this score over time. The positive spectra capture within-scale connectivity and geometric information that persistent Betti vectors do not record. Under a finite-support local model, we prove that the oracle upper bound on detection delay and a local minimax lower bound have the same order. This common order is the logarithm of the average run length constraint divided by the squared ridge-whitened separation. We also establish finite-horizon false-alarm and expected-delay bounds for plug-in whitened scores under weak dependence within each state. The method has two phases. Phase I estimates the projection parameters, selects the feature configuration, and calibrates the control limit. Phase II updates the resulting CUSUM statistic online. Experiments on simulated and real monitoring data show stable false-alarm control and competitive detection performance.

Figures

Figures reproduced from arXiv: 2607.08635 by the authors.

Figure 1
Figure 1. Training, calibration, and online monitoring workflow of PL-CUSUM. (a) Window segmentation, (b) [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 1
Figure 1. Results from the tree-family experiments. (a) Detection probabilities for persistent Betti features alone [PITH_FULL_IMAGE:figures/full_fig_p015_1.png] view at source ↗
Figure 2
Figure 2. Positive PL spectral separation in the tree family. The horizontal axis is the tree order [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗
Figures from the paper (10 more)
Figure 2
Figure 2. Figure 2: Plug-in PL-CUSUM as the Phase I sample size increases. Panels (a)–(d) show the estimation error, [PITH_FULL_IMAGE:figures/full_fig_p016_2.png]
Figure 3
Figure 3. Figure 3: Separation-scaling experiment in a finite-support PL model. The left panel compares the local [PITH_FULL_IMAGE:figures/full_fig_p014_3.png]
Figure 3
Figure 3. Figure 3: Sensitivity to the positive-spectrum truncation level [PITH_FULL_IMAGE:figures/full_fig_p017_3.png]
Figure 4
Figure 4. Figure 4: Effect of the Phase I training sample size on plug-in PL-CUSUM. The left panel shows the single-step [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]
Figure 4
Figure 4. Figure 4: Detection results for SWaT and Electric Motor under combined load and background vibration at [PITH_FULL_IMAGE:figures/full_fig_p020_4.png]
Figure 5
Figure 5. Figure 5: Nonlinear geometric stress test with standardized windows. The left panel compares the ridge-whitened [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: Hyperparameter sensitivity of PL-CUSUM. The top row varies the window length [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]
Figure 7
Figure 7. Figure 7: Task-family coverage heatmap in the real-data experiments. Color indicates the pass fraction of each [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]
Figure 8
Figure 8. Figure 8: Normalized online PL-CUSUM trajectories in real-data case studies. The vertical axis is [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: Pre- and post-change PL projection scores in real-data case studies. Black horizontal lines indicate [PITH_FULL_IMAGE:figures/full_fig_p020_9.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references

  1. [1]

    and Harer, John , title =

    Perea, Jose A. and Harer, John , title =. Foundations of Computational Mathematics , year =

  2. [2]

    Edelsbrunner, Herbert and Harer, John , title =

  3. [3]

    Wang, Rui and Nguyen, D. D. and Wei, Guo-Wei , title =. International Journal for Numerical Methods in Biomedical Engineering , year =

  4. [4]

    Persistent Laplacians: Properties, algorithms and implications , journal =

    M\'. Persistent Laplacians: Properties, algorithms and implications , journal =. 2022 , volume =

    2022

  5. [5]

    Page, E. S. , title =. Biometrika , year =

  6. [6]

    The Annals of Mathematical Statistics , year =

    Lorden, Gary , title =. The Annals of Mathematical Statistics , year =

  7. [7]

    Siegmund, David , title =

  8. [8]

    , title =

    Moustakides, George V. , title =. The Annals of Statistics , year =

Show all 53 references
  1. [9]

    Wald, Abraham , title =

  2. [10]

    IEEE Transactions on Information Theory , year =

    Lai, Tze Leung , title =. IEEE Transactions on Information Theory , year =

  3. [11]

    Sequential Analysis: Hypothesis Testing and Changepoint Detection , publisher =

    Tartakovsky, Alexander and Nikiforov, Igor and Basseville, Mich. Sequential Analysis: Hypothesis Testing and Changepoint Detection , publisher =

  4. [12]

    van der Vaart, A. W. , title =

  5. [13]

    I-divergence geometry of probability distributions and minimization problems , journal =

    Csisz\'. I-divergence geometry of probability distributions and minimization problems , journal =. 1975 , volume =

    1975

  6. [14]

    Barndorff-Nielsen, O. E. , title =

  7. [15]

    Discrete & Computational Geometry , year =

    Barmak, Jonathan Ariel and Minian, Elias Gabriel , title =. Discrete & Computational Geometry , year =

  8. [16]

    Bhatia, Rajendra , title =

  9. [17]

    and Thomas, Joy A

    Cover, Thomas M. and Thomas, Joy A. , title =

  10. [18]

    Colosimo, B. M. and Jones-Farmer, L. A. and Megahed, F. M. and Paynabar, K. and Ranjan, C. and Woodall, W. H. , title =. Technometrics , year =

  11. [19]

    and Kirch, C

    Aue, A. and Kirch, C. , title =. Biometrika , year =

  12. [20]

    and Xie, X

    Qiu, P. and Xie, X. , title =. Technometrics , year =

  13. [21]

    , title =

    Carlsson, G. , title =. Bulletin of the American Mathematical Society , year =

  14. [22]

    Fasy, B. T. and Lecci, F. and Rinaldo, A. and Wasserman, L. and Balakrishnan, S. and Singh, A. , title =. The Annals of Statistics , year =

  15. [23]

    , title =

    Bubenik, P. , title =. Journal of Machine Learning Research , year =

  16. [24]

    and Bubenik, P

    Patrangenaru, V. and Bubenik, P. and Paige, R. and Osborne, D. , title =. Sankhya A , year =

  17. [25]

    and Emerson, T

    Adams, H. and Emerson, T. and Kirby, M. and Neville, R. and Peterson, C. and Shipman, P. and Chepushtanova, S. and Hanson, E. and Motta, F. and Ziegelmeier, L. , title =. Journal of Machine Learning Research , year =

  18. [26]

    Stolz, B. J. and Harrington, H. A. and Porter, M. A. , title =. Chaos , year =

  19. [27]

    and Katz, Y

    Gidea, M. and Katz, Y. , title =. Physica A: Statistical Mechanics and its Applications , year =

  20. [28]

    and Yuvaraj, M

    Islambekov, U. and Yuvaraj, M. and Gel, Y. R. , title =. Environmetrics , year =

  21. [29]

    and Mak, S

    Zheng, X. and Mak, S. and Xie, L. and Xie, Y. , title =. Technometrics , year =

  22. [30]

    Thomas, A. M. and Jauch, M. and Matteson, D. S. , title =. Technometrics , year =

  23. [31]

    and Munch, E

    Myers, A. and Munch, E. and Khasawneh, F. A. , title =. Physical Review E , year =

  24. [32]

    and Algar, S

    Tan, E. and Algar, S. D. and Corr. Network representations of attractors for change point detection , journal =. 2023 , volume =

    2023

  25. [33]

    Myers, A. and Mu\. Temporal network analysis using zigzag persistence , journal =. 2023 , volume =

    2023

  26. [34]

    and Wan, Z

    Davies, T. and Wan, Z. and Sanchez-Garcia, R. J. , title =. Proceedings of the 40th International Conference on Machine Learning , year =

  27. [35]

    and Wei, G.-W

    Wei, X. and Wei, G.-W. , title =. 2023 , eprint =

    2023

  28. [36]

    , title =

    Pollak, M. , title =. The Annals of Statistics , year =

  29. [37]

    and Moustakides, G

    Xie, L. and Moustakides, G. V. and Xie, Y. , title =. IEEE Transactions on Information Theory , year =

  30. [38]

    and Xie, Y

    Wei, S. and Xie, Y. , title =. Journal of the Royal Statistical Society Series B: Statistical Methodology , year =

  31. [39]

    Mahalanobis, P. C. , title =. Proceedings of the National Institute of Sciences of India , year =

  32. [40]

    The jackknife and the bootstrap for general stationary observations , journal =

    K. The jackknife and the bootstrap for general stationary observations , journal =. 1989 , volume =

    1989

  33. [41]

    Bradley, R. C. , title =. Probability Surveys , year =

  34. [42]

    and Celisse, A

    Arlot, S. and Celisse, A. and Harchaoui, Z. , title =. Journal of Machine Learning Research , year =

  35. [43]

    and Cheong, K

    Bao, Z. and Cheong, K. M. and Li, Y. and Qiu, J. , title =. 2026 , eprint =

    2026

  36. [44]

    and Chen, H

    Chu, L. and Chen, H. , title =. IEEE Transactions on Signal Processing , year =

  37. [45]

    , title =

    Hoeffding, W. , title =. Journal of the American Statistical Association , year =

  38. [46]

    Tropp, J. A. , title =. Foundations of Computational Mathematics , year =

  39. [47]

    and Edelsbrunner, H

    Cohen-Steiner, D. and Edelsbrunner, H. and Harer, J. , title =. Discrete & Computational Geometry , year =

  40. [48]

    and Fasy, B

    Chazal, F. and Fasy, B. T. and Lecci, F. and Rinaldo, A. and Singh, A. and Wasserman, L. , title =. Modeling and Analysis of Information Systems , year =

  41. [49]

    and Mileyko, Y

    Turner, K. and Mileyko, Y. and Mukherjee, S. and Harer, J. , title =. Discrete & Computational Geometry , year =

  42. [50]

    Sliced Wasserstein kernel for persistence diagrams , booktitle =

    Carri. Sliced Wasserstein kernel for persistence diagrams , booktitle =. 2017 , series =

    2017

  43. [51]

    and Li, J

    Liu, J. and Li, J. and Wu, J. , title =. Homology, Homotopy and Applications , year =

  44. [52]

    and Wang, R

    Cottrell, M. and Wang, R. and Wei, G.-W. , title =. Journal of Chemical Information and Modeling , year =

  45. [53]

    and Hozumi, Y

    Cottrell, M. and Hozumi, Y. and Wei, G.-W. , title =. Computers in Biology and Medicine , year =

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.