Pith. sign in

REVIEW 3 major objections 4 minor 27 references

Poisson approximation of large-lifetime cycles

T0 review · 3 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper proves that near-maximal-lifetime cycles in a Poisson point cloud converge to a Poisson point process, jointly for centers, lifetimes and deathtimes, under explicit conditions.

desk verdict First Poisson limits for large-lifetime cycles in arbitrary dimensions, but the claimed rates in Theorems 2.3–2.5 are not justified as written. read the letter →

arxiv 2412.17482 v1 pith:RD7UP7HE submitted 2024-12-23 math.PR

Christian Hirsch , Nikolaj Nyvold Lundbye , Moritz Otto This is my paper · ORCID
classification math.PR MSC 55U1060D0560G5560F0560G70
keywords persistenthomologyPoissonpointprocesslarge-lifetimecyclesCechcomplexVietoris-Ripsextremevaluetheoryapproximationpersistencediagram
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

Persistent homology singles out cycles—loops, cavities—that survive over a long range of scales, and topological data analysis treats these as candidates for genuine structure. This paper proves that, in a growing Poisson point cloud, the rare cycles with near-maximal lifetime are asymptotically independent and Poisson-distributed: their centers converge to a Poisson point process, and in a sparse regime the centers, scaled lifetime deviations, and deathtime deviations converge jointly to a Poisson process with an explicit intensity. On the two-dimensional flat torus the authors prove Poisson convergence of the centers for Cech complexes with multiplicative lifetimes. In dimensions d >= 2, under a deathtime bound that makes connected clusters small, they obtain joint convergence for both Cech and Vietoris-Rips filtrations, with the limiting intensity factorized as $\alpha$ * kappa(y)^m * q * u^(q-1) * H_q(v). If the Poisson description is correct, it turns extreme persistence into a null model that can be used for significance testing and for the asymptotic law of the largest lifetime.

What carries the argument

The argument is carried by the deterministic maximal lifetime ell_max (the supremum of lifetimes over all m-point clusters with deathtime at most 1) and by the integrals g(u) and h(u,v) that count, up to scaling, the volume of cluster configurations whose lifetime is within u of ell_max. The threshold u_{n,$\alpha$} is defined through $g^{{-1}}$, and the limiting density is obtained from the Taylor behavior of g and h: the exponent q and the function H_q come from the first nonvanishing derivatives. In the sparse regime the proof approximates the point process of large-lifetime cycles by a Poisson U-statistic built from cycles inside m-clusters, bounds the error by the expected number of (m+1)-clusters, and applies the functional Poisson approximation for U-statistics; Assumption U, which requires at most one near-maximal cycle per cluster, is what turns the U-statistic into a simple marked point process. In the unbounded regime the proof instead truncates to mortal cycles with a deathtime bound, stabilizes the functional to a ball around the cycle center, and controls simultaneous large cycles through the continuous disjoint-occurrence inequality and the planar separation theorem.

What would settle it

For a parameter triple (d,k,m) not covered by Proposition 5.4, search numerically for an m-point set containing two different negative simplices with deathtime at most 1 and lifetime within epsilon of ell_max(m). Such a configuration would violate Assumption U and remove the Poisson limit in Theorems 2.3-2.5, while a proof that no such configuration exists would extend the theorems to that parameter triple.

Watch

Extended reading notes

Core claim

The central claim is Theorem 2.5: for a Poisson point cloud with intensity n * kappa on a region W in R^d, under assumptions F, M, P, T, U, and H, the marked point process of large-lifetime cycles—recording each cycle's center, its scaled deviation (ell_max - ell)/u_{n,$\alpha$} from the deterministic maximal lifetime, and the scaled deathtime deviation (r_n - r_x)/u_{n,$\alpha$}—converges in Kantorovich-Rubinstein distance to a Poisson point process with intensity $\alpha$ * kappa(y)^m * q * u^(q-1) * H_q(v), with rate O(rho_{n,m+1} + rho_{n,m}^{-1/q}). Here rho_{n,m} = n (n r_n^d)^(m-1) is the expected number of m-clusters, and the threshold u_{n,$\alpha$} is set so that the expected number of large-lifetime features is $\alpha$. In the unbounded regime on the flat torus, Theorem 2.1 establishes Poisson convergence of the centers alone for Cech complexes with multiplicative lifetimes, with rate O($n^{{-1/36+epsilon}}$). A consequence is that the largest lifetime's deviation follows a Weibull law in the sparse regime, while the transformed largest lifetime is asymptotically exponential on the torus.

Load-bearing premise

Assumption U says that in any m-point cluster, at most one subset of points can form a cycle whose lifetime is close to the maximal possible lifetime; without this uniqueness, a single cluster could hold several large-lifetime cycles and the limiting process would no longer be Poisson.

Editorial extensions

If this is right

  • The scaled deviation of the largest lifetime from ell_max is asymptotically Weibull with shape q; on the torus, the transformed largest lifetime n^3 v(ell) is asymptotically Exponential(1).
  • Deathtime deviations of near-maximal cycles are asymptotically i.i.d. with density H_q, so the rightmost strip of the persistence diagram has a tractable product structure for model checking.
  • Inhomogeneous intensity functions kappa are handled: the limiting center intensity is alpha * kappa(y)^m, meaning large-lifetime cycles appear in regions where kappa is high.
  • The convergence rate O(rho_{n,m+1} + rho_{n,m}^{-1/q}) tells practitioners how sparse the deathtime bound must be for the Poisson approximation to be accurate.
  • When the model assumptions are met, extreme persistent homology ceases to need Monte Carlo or bootstrap calibration for a Poisson null.

Reading between the lines

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

  • If the uniqueness assumption U fails for some (d,k,m) combination, clusters could contribute several near-maximal cycles; the limiting point process would then have a clustered, non-Poisson structure, so locating such configurations would mark the boundary of the theorem's applicability.
  • The product form of the limiting intensity suggests a diagnostic: normalized deathtime deviations should look independent of spatial location and lifetime, so a correlation or goodness-of-fit test on that independence could be used before trusting the Poisson model.
  • The methods may extend to the flat torus in the sparse regime and to multiplicative lifetimes in d >= 3, since the limiting formulas are scale-invariant; the paper states the torus extension as a conjecture.
  • The simulation results hint that the Taylor order q for Cech complexes with m>3 may differ from the naive 2m-3 guess; if confirmed, it would change the Weibull shape used in statistical tests.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper studies Poisson approximation for large-lifetime cycles in persistent homology of Poisson point clouds. In the unbounded regime on the two-dimensional flat torus (Theorem 2.1), it proves convergence of the spatial centers of cycles with multiplicatively large lifetime to a stationary Poisson process, with an explicit rate. In the m-sparse regime on R^d (Theorems 2.3–2.5), it proves joint Poisson convergence of the centers, scaled lifetime deviations, and scaled deathtime deviations of large-lifetime cycles, under structural assumptions (F, M, P, T, U, G, H) on the filtration, sparsity, intensity, threshold, uniqueness, and regularity. The proofs combine stabilization, Malliavin–Stein methods, a U-statistic approximation, the continuous BK inequality, and detailed geometric analysis of maximal lifetime configurations. Section 5 verifies the thresholds and several model assumptions, Section 6 applies the results to largest lifetimes and persistence diagrams, and Section 7 contains simulations.

Significance. If the results hold as stated, this is a substantial contribution to extreme-value theory for persistent homology: it provides the first Poisson approximation for large-lifetime cycles in arbitrary dimensions, including marks for lifetimes and deathtimes, and it gives quantitative rates. The proofs are detailed and conditional on explicit assumptions, and the paper is honest about which assumptions remain open (notably Assumption U in general and Assumption H in d≥3). The applications to largest lifetimes and to a deathtime-based model check are attractive. However, the quantitative rates in Theorems 2.3–2.5 contain a load-bearing gap: an O(r_n) intensity error from Lemma 4.7 is absorbed without a valid justification under Assumption M. This does not necessarily invalidate the convergence statements, but it requires a correction of the stated rates or a strengthening of the assumptions.

major comments (3)
  1. [§4.3, Proof of Theorem 2.3] The proof states 'As r_n ⩽ ρ_{n,m+1} for all sufficiently large n' and uses this to absorb the O(r_n) intensity error from Lemma 4.7. This inequality is not a consequence of Assumption M. For instance, with d=10, m=3 and r_n=n^{-0.14}, one has n(nr_n^d)^2=n^{0.2}→∞ and n(nr_n^d)^3=n^{-0.2}→0, so Assumption M holds, but r_n=n^{-0.14} is much larger than n^{-0.2}=ρ_{n,m+1}. Lemma 4.7 only gives the intensity error as O(r_n), so the displayed rate O(ρ_{n,m+1}) in Theorem 2.3 is not established as stated. The same O(r_n) term is silently dropped in the proofs of Theorems 2.4 and 2.5 after invoking Lemma 4.7. The convergence assertions may remain true, but the quantitative rates need to be replaced by at least O(r_n+ρ_{n,m+1}+ρ_{n,m}^{-1/q}), or Assumption M must be strengthened so that r_n is dominated by ρ_{n,m+1}.
  2. [§6, Example 6.3] The example takes k=d−1 and m=2d with the Vietoris–Rips filtration, but Proposition 5.4 verifies Assumption U only for m=k or for k=d+1 and m=2d. The combination k=d−1 is outside the verified range, and for d=2 it would give k=1, contradicting the standing assumption k≥3. If the intended choice is k=d+1, as the surrounding formulas and the later d=2 specialization suggest, then the text should be corrected accordingly.
  3. [§5.2, Proposition 5.6] The proposition statement says that, for the Vietoris–Rips case d=2, k=3, m=4, assumption H holds with q=4. However, the proof immediately after the statement aims to prove (i) lim_{u→0} h(u,1)/u^5∈(0,∞) and (ii) lim_{u→0} \tilde h^{(0,1)}(u,v)/u^5∈(0,∞), concluding that q=5 works; this matches Remark 2.6(4), which also gives q=5 for the Vietoris–Rips case. The statement of Proposition 5.6 should be changed to q=5, and the associated asymptotic expression for ℓ_{n,α} should be corrected consistently.
minor comments (4)
  1. [Remark 2.2(4)] The limiting point process in the extension statement is described as having unit intensity on R^d × [0,∞), but in the unbounded torus setting the spatial coordinate should be on T^2; please check whether this should be T^2 × [0,∞).
  2. [§3.3, Proof of Theorem 2.1] In the final displayed chain, the first term should presumably be dKR(ξ^1_n, \hatξ^1_n) rather than dKR(\hatξ^1_n, \hatξ^1_n); as written the triangle-inequality step is not meaningful.
  3. [Theorem 2.5, Assumption H] The displayed condition for j<q is garbled: it should state that \tilde h^{(j,1)}(0,v)=0 for all j<q and all v, and that \tilde h^{(q,1)}(0,v)>0; the current line 'when j < qfor all u, v' is not grammatical and could confuse readers.
  4. [§6, Example 6.3] The intensity convention in the displayed intensity function appears to be (2π)^{-d^2} exp(-d∥y∥^2), consistent with κ(y)^{2d}; if so, it would be clearer to write it explicitly as κ(y)^m with m=2d.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: thresholds are extreme-value normalizations, and the Poisson limits are derived from the model functions via external Stein/U-statistic results.

full rationale

The derivation chain is self-contained against external tools. The thresholds in assumptions T' and T are chosen as v^{-1}(alpha/n^3) and g^{-1}(alpha/(rho_{n,m} binom(m,k)/m!)) solely to normalize the expected count of the approximating U-statistic to alpha (Remark 2.2(2) and Lemma 4.3, equation (4.4)); this is a standard extreme-value normalization, not a fitted prediction. The nontrivial Poisson content, namely independence of counts and absence of clustering, comes from the stabilizing-functional theorem [4, Theorem 4.1] and the U-statistic Poisson approximation [10, Theorem 3.1], with clustering excluded via the continuous BK inequality [14]. The limiting intensity alpha kappa^m and the mark densities q u^{q-1} and H_q(v) are obtained by Taylor-expanding the model-defined functions g and h (Lemmas 4.8 and 4.9), not assumed. Assumption U is an explicitly stated model condition, verified for certain filtrations in Propositions 5.4 to 5.6 and left open in general; it is not imported from the authors' prior work. The only self-citation used in a proof step, [9, Lemma 2] in Proposition 5.2 for continuity of the Vietoris-Rips analogue of v, is an external published lemma about a different but analogous threshold map and is not load-bearing for the Poisson approximation itself. A correctness caveat, unrelated to circularity: in the proof of Theorem 2.3, the sentence "As r_n <= rho_{n,m+1} for all sufficiently large n" is not implied by Assumption M, so the O(rho_{n,m+1}) rates in Theorems 2.3 to 2.5 omit the O(r_n) intensity error from Lemma 4.7 unless an additional argument is supplied. This affects the quantitative rate claim, not the independence of the derivation.

Assumptions & free parameters 0 free parameters · 7 assumptions · 0 invented entities

The central theorems are conditional on the stated assumptions; U and H are only partially verified, and examples contain typos regarding the admissible k for the Vietoris-Rips case. No new physical or mathematical entities are introduced.

assumptions (7)
  • domain assumption Assumption F: filtration is Cech or Vietoris-Rips; lifetime is additive or multiplicative.
    Stated in Theorem 2.3 and used throughout; defines the geometric complex and lifetime model.
  • domain assumption Assumption M (m-sparse regime): n(n r_n^d)^{m-1} → ∞ and n(n r_n^d)^m → 0.
    Defines the sparse connectivity regime used in Theorems 2.3-2.5; ensures many m-clusters but few (m+1)-clusters.
  • domain assumption Assumption P: intensity function κ is bounded, integrable, with modulus of continuity O(r_n) and boundary condition.
    Controls inhomogeneity of the Poisson point process; used in Lemma 4.4 and Lemma 4.7. Verified for convex W and C^1 densities in Proposition 5.3.
  • domain assumption Assumption U: for all P of size m at most one near-maximal large-lifetime feature per cluster.
    Load-bearing for the sparse-regime Poisson property; verified only for Cech m=k≤d+1 and VR k=d+1, m=2d (Proposition 5.4).
  • domain assumption Assumption H: regularity of h with q-th order vanishing conditions at 0.
    Needed for the joint lifetime-deathtime convergence in Theorem 2.5; verified only for d=2 and additive lifetimes (Propositions 5.5 and 5.6).
  • standard math Identification of each nontrivial p-cycle with a negative (p+1)-simplex.
    Taken from [7, Algorithm 9] and used in Section 1 to define birthtime, deathtime, and lifetime via negative simplices.
  • standard math Hausdorff stability of persistent homology.
    Used in Propositions 5.1 and 5.2 to prove continuity of the threshold functions v and g; cited as [7, Corr. 11.29].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Poisson approximation of large-lifetime cycles." pith.science (2026). https://pith.science/paper/RD7UP7HE

@misc{pith2026241217482,
  author       = {Pith},
  title        = {Pith review of: Poisson approximation of large-lifetime cycles},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RD7UP7HE}},
  note         = {Machine review of arXiv:2412.17482}
}
abstract

In topological data analysis, the notions of persistent homology, birthtime, lifetime, and deathtime are used to assign and capture relevant cycles (i.e., topological features) of a point cloud, such as loops and cavities. In particular, cycles with a large lifetime are of special interest. In this paper, we study such large-lifetime cycles when the point cloud is modeled as a Poisson point process. First, we consider the case with no bound on the deathtime, where we establish Poisson convergence of the centers of large-lifetime features on the 2-dimensional flat torus. Afterwards, by imposing a bound on the deathtime, we enter a sparse connectivity regime, and we prove joint Poisson convergence of the centers, lifetimes, and deathtimes in dimensions $d \geq 2$ under suitable model conditions.

Figures

Figures reproduced from arXiv: 2412.17482 by the authors.

Figure 1
Figure 1. Illustration of simplices and Cech versus Vietoris-Rips complexes. ˇ The flat torus Henceforth, we will replace X by a region W ⊆ R d , where d ⩾ 2. The canonical choice of metric ρ on W is the Euclidean distance, which is the metric induced by the Euclidean norm ∥· ∥. However, we can also define the flat torus T d as the unit cube [0, 1]d equipped with the toroidal metric ρ(x, y) = min q∈Zd ∥x − y + q∥, for any x, … view at source ↗
Figure 2
Figure 2. Example 1.1 with the Cech (left) and Vietoris-Rips (right) complex at ˇ time t = √ 3 2 . The red points mark the centers zx and zx′ of the negative 2-simplices [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. In both cases, the red point is the center zx of x and the dotted line is the unique edge of length bx,Pn . Left: All blue points is the cluster associated to x. Right: All blue vertices and edges is the loop associated of x. In order to control the last two error terms, E3 and EI 4 , we need to tool to deal with multiple exceedances. In particular, we will prove than when the birthtimes are bounded by b ⩾ 0, then t… view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Two loops (left) sharing 2 edges and 4 vertices and an illustration of the choice of γ0, γ1 and γ2 (right) in the proof of Lemma 3.9. 3.5. Proof of Lemmas 3.1 to 3.7 Proof of Lemma 3.1. First, to make the notation more compact, let an = log(n) log(log(n)). (3.15) Lower…
Figure 5
Figure 5. Figure 5: Left: Construction of the Qi-squares of length 2Ln inside the unit cube. Right: The smaller Qi,j -squares of length ℓn inside the shrunk version Q˜ i of Qi . Proof of Lemma 3.3. Let Tm denote the set of spanning trees on {1, . . . , m}. If y is bn-connected, there is a…
Figure 6
Figure 6. Figure 6: Placement of the points x4, . . . , xm in the proof of Proposition 5.1 such that ∥x1 − x2∥ remains the largest distance. 5.2. The m-sparse regime Proposition 5.2 (Assumption T). Under assumptions F, M and U, it holds that g in (2.9) is continuous and strictly increasin…
Figure 7
Figure 7. Figure 7: In black, the 2-simplex (0, y2, y3) and angles θ, ϕ as defined in the proof of Proposition 5.5. In gray, the center z of (0, y2, y3) in polar coordinates (r, τ ) [PITH_FULL_IMAGE:figures/full_fig_p036_7.png]
Figure 8
Figure 8. Figure 8: Illustration of the two circles ∂Bu 0 and ∂Bu 2 with intersection points p3 = p3(u) and p4 = p4(u). Additionally, the two approximations ˆp3 and ˆp4 are illustrated, and the ball B(ˆp4, 2) is (partly) displayed as the dotted semi-circle. We now find the intersection po…
Figure 9
Figure 9. Figure 9: From left to right: The construction of A˜(y3), A(y3), and Aˆ(y3) as well as an illustration of the points Q˜ 0, Q1, Q˜ 2, Q4, Q5 in the proof of Proposition 5.6. Thus, we now write h(u, v) = Z 1 1−v r 5 Z C ur 3 vol Aˆ(y3)  dy3dr+ Z 1 1−v Z C ur 3 vol Aˆ(y3)△A(y3)  …
Figure 10
Figure 10. Figure 10: Illustration of the initial configuration (red points) and final config￾uration (blue points) for the largest lifetime achieved through optimization for m = 4, 5, 6 in Experiment 7.1. The lifetimes of these configurations are 0.273, 0.374 and 0.418 and can be compared…
Figure 11
Figure 11. Figure 11: The estimated value of log(− log(S(t))) plotted against log(t) when m = 3 and a straight line is fitted (blue) with the slope recorded as q. This is done for the intensity n = 20, 50, 100, 200, 500, 1000. Next, to address the second objective, we will repeat the metho…
Figure 12
Figure 12. Figure 12: The results of checking the fit of the Weibull distribution for the devi￾ation of the largest lifetime for m = 4, 5, 6, where log(− log(S(t))) is plotted against log(t) and a straight line is fitted (blue) to determine a candidate for the value q. Each value of m is e…
Figure 13
Figure 13. Figure 13: Average correlation between deathtimes of the two largest lifetimes, which is carried out with intensity n = 100, 200, . . . , 1000. Acknowledgements The authors would like to thank Esben Trøllund Boye for his contributions to an earlier version of the manuscript as w…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 25 canonical work pages

  1. [1]

    C. A. N. Biscio, N. Chenavier, C. Hirsch, and A. M. Svane. Testing goodness of fit for point processes via topological data analysis. Electron. J. Stat. , 14(1):1024–1074, 2020

    work page 2020

  2. [2]

    Bobrowski

    O. Bobrowski. Homological connectivity in random ˇCech complexes. Probab. Theory Related Fields, 183(3-4):715– 788, 2022

    work page 2022

  3. [3]

    Bobrowski, M

    O. Bobrowski, M. Kahle, and P. Skraba. Maximally persistent cycles in random geometric complexes. Ann. Appl. Probab., 27(4):2032–2060, 2017

    work page 2017

  4. [4]

    Bobrowski, M

    O. Bobrowski, M. Schulte, and D. Yogeshwaran. Poisson process approximation under stabilization and Palm coupling. Ann. H. Lebesgue, 5:1489–1534, 2022

    work page 2022

  5. [5]

    Bobrowski and P

    O. Bobrowski and P. Skraba. A universal null-distribution for topological data analysis. Scientific Reports , 13(1):12274, 2023

    work page 2023

  6. [6]

    Bobrowski and P

    O. Bobrowski and P. Skraba. Universality in random persistent homology and scale-invariant functionals. arXiv preprint arXiv:2406.05553v3, 2024

    arXiv 2024

  7. [7]

    Boissonnat, F

    J.-D. Boissonnat, F. Chazal, and M. Yvinec. Geometric and Topological Inference. Cambridge University Press, Cambridge, 2018

    work page 2018

  8. [8]

    P. Bubenik. Statistical topological data analysis using persistence landscapes. J. Mach. Learn. Res. , 16:77–102, 2015

    work page 2015

Show all 27 references
  1. [9]

    Chenavier and C

    N. Chenavier and C. Hirsch. Extremal lifetimes of persistent cycles. Extremes, 25(2):299–330, 2022. References 52

    2022

  2. [10]

    Decreusefond, M

    L. Decreusefond, M. Schulte, and C. Th¨ ale. Functional Poisson approximation in Kantorovich-Rubinstein distance with applications to U-statistics and stochastic geometry. Ann. Probab., 44(3):2147–2197, 2016

    2016

  3. [11]

    Gidea and Y

    M. Gidea and Y. Katz. Topological data analysis of financial time series: landscapes of crashes. Phys. A, 491:820– 834, 2018

    2018

  4. [12]

    G. R. Grimmett, A. E. Holroyd, and G. Kozma. Percolation of finite clusters and infinite surfaces. Math. Proc. Cambridge Philos. Soc. , 156(2):263–279, 2014

    2014

  5. [13]

    M. A. Hern´ andez Cifre and E. Saor ´ ın. On inner parallel bodies and quermassintegrals.Israel J. Math., 177:29–47, 2010

    2010

  6. [14]

    Heydenreich, R

    M. Heydenreich, R. van der Hofstad, G. Last, and K. Matzke. Lace expansion and mean-field behavior for the random connection model. arXiv preprint arXiv:1908.11356 , 2019

    1908 arXiv

  7. [15]

    Hirsch and D

    C. Hirsch and D. Valesin. Face and cycle percolation. arXiv:2212.06243, 2022

    2022 arXiv

  8. [16]

    Janson, T

    S. Janson, T. Luczak, and A. Rucinski. Random Graphs. John Wiley & Sons Inc., New York, 2000

    2000

  9. [17]

    M. Kahle. Random geometric complexes. Discrete Comput. Geom. , 45(3):553–573, 2011

    2011

  10. [18]

    Kahle and E

    M. Kahle and E. Meckes. Limit theorems for Betti numbers of random simplicial complexes. arXiv:1009.4130, 2011

    2011 arXiv

  11. [19]

    Kanari, P

    L. Kanari, P. D lotko, M. Scolamiero, R. Levi, J. Shillcock, K. Hess, and H. Markram. A topological representation of branching neuronal morphologies. Neuroinformatics, 16(1):3–13, 2018

    2018

  12. [20]

    Krebs and C

    J. Krebs and C. Hirsch. Functional central limit theorems for persistent Betti numbers on cylindrical networks. Scand. J. Stat. , 49(1):427–454, 2022

    2022

  13. [21]

    Meester and R

    R. Meester and R. Rahul. Continuum Percolation. Cambridge University Press, Cambridge, 1996

    1996

  14. [22]

    Owada and Z

    T. Owada and Z. Wei. Functional strong law of large numbers for Betti numbers in the tail. Extremes, 25(4):655– 693, 2022

    2022

  15. [23]

    Saadatfar, H

    M. Saadatfar, H. Takeuchi, V. Robins, N. Francois, and Y. Hiraoka. Pore configuration landscape of granular crystallization. Nature Communications, 8(1):1–11, 2017

    2017

  16. [24]

    Schneider and W

    R. Schneider and W. Weil. Stochastic and Integral Geometry . Springer, Heidelberg, 2008

    2008

  17. [25]

    A. M. Thomas and T. Owada. Functional strong laws of large numbers for Euler characteristic processes of extreme sample clouds. Extremes, 24(4):699–724, 2021

    2021

  18. [26]

    van de Weygaert, G

    R. van de Weygaert, G. Vegter, H. Edelsbrunner, B. J. T. Jones, P. Pranav, C. Park, W. A. Hellwing, B. Eldering, N. Kruithof, E. G. P. Bos, J. Hidding, J. Feldbrugge, E. ten Have, M. van Engelen, M. Caroli, and M. Teillaud. Alpha, Betti and the Megaparsec Universe: on the topo...

    2011

  19. [27]

    Wasserman

    L. Wasserman. Topological data analysis. Annu. Rev. Stat. Appl. , 5(1):501–535, 2018

    2018

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.