Pith. sign in

REVIEW 3 major objections 8 minor 24 references

Density-Robust Spherical Coordinates from Persistent Cohomology

T0 review · 3 major / 8 minor · reviewed 2026-07-31 · grok-4.5

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

Pith's one-line read The first density-robust spherical coordinates from persistent cohomology are built by computing classical sphere maps on rejection-sampled uniform subsamples and aligning them with a spherical Procrustes problem.

desk verdict Solid first density-robust S² pipeline via subsample-and-Procrustes; math is clean, robustness still empirical and only synthetic. read the letter →

arxiv 2607.24650 v2 pith:WENJU54Z submitted 2026-07-27 math.AT cs.CG

Nick Nordwald , Inés García-Redondo , Anthea Monod This is my paper · ORCID
classification math.ATcs.CG MSC 55N3162R4068U05
keywords persistentcohomologysphericalcoordinatesdensityrobustnessProcrustesalignmentrejectionsamplingtopologicaldimensionalityreductionVietoris–Rips
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 cohomology can turn a degree-two feature into a map from data to the sphere, giving intrinsic nonlinear coordinates when the latent space is spherical rather than Euclidean. Those maps are badly warped by uneven sampling density, and the usual density fixes for circular coordinates do not transfer because spherical coordinates come from a nonlinear energy on sphere-valued maps, not from weighted harmonic cocycles. This paper shows that density robustness can still be recovered without changing that optimization: thin the data by rejection sampling into approximately uniform subsamples, run the classical spherical construction on each, extend the maps, and average them after solving a spherical Procrustes alignment problem. A Euclidean relaxation of that alignment is proved close enough to the geodesic optimum to be usable in practice, and experiments recover accurate azimuth and elevation under severe sampling bias while matching the classical method on uniform data and scaling to ten thousand points by never building the full high-dimensional complex.

What carries the argument

The spherical Procrustes problem (jointly rotating several S²-valued maps and choosing a consensus map to minimize summed squared geodesic discrepancy), together with Theorem 3.4, which bounds the distance between the Euclidean O(3) Procrustes solution (projected back to the sphere) and the true spherical optimum by (1+π) times the square root of the minimal spherical loss.

What would settle it

Take a sphere sampled from a strongly biased density (for example von Mises–Fisher with large concentration), run the full subsample-align-average pipeline with the paper’s density estimator, and check whether recovered azimuth and elevation still correlate tightly with ground truth; a clear systematic warp that survives many subsamples and alignments would refute the claim.

Watch

Extended reading notes

Core claim

Density-robust spherical topological coordinates exist and can be constructed by extending the subsampling-and-alignment paradigm from the circle to the sphere: compute classical spherical coordinates on approximately uniform rejection samples, extend them by kernel-weighted nearest neighbors, and combine them by solving a spherical Procrustes problem whose Euclidean relaxation approximates the intrinsic optimum within a factor controlled by the optimal spherical loss.

Load-bearing premise

That local-count rejection sampling really produces subsamples on which the classical sphere-map optimizer is already unbiased, so that extending and averaging those maps yields a faithful global coordinate.

Editorial extensions

If this is right

  • Sphere-valued latent structure can be recovered from data whose observation density is badly non-uniform without rewriting the underlying energy.
  • Degree-two persistent cohomology becomes practical on datasets of several thousand to ten thousand points because cohomology is run only on fixed-size subsamples.
  • The same alignment idea supplies a template for density-robust coordinates targeting other spaces once a Procrustes-type alignment on that space is available.
  • Under uniform sampling the method matches classical spherical coordinates, so it can replace the baseline without quality loss when density is unknown.

Reading between the lines

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

  • Because S² is CP¹, the same subsample-and-align pattern is a natural first step toward density-robust projective coordinates from higher cohomology.
  • If the density estimator fails on outliers, a more stable density model (or explicit outlier rejection) would be the highest-leverage practical upgrade before chasing tighter Procrustes bounds.
  • The (1+π)√L* guarantee is useful mainly when the spherical loss is already small; experiments that report L* alongside angle error would make the theory–practice link sharper.
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

3 major / 8 minor

Summary. The manuscript constructs the first density-robust spherical (degree-two) coordinates from persistent cohomology. The strategy ports the subsampling-and-alignment framework of Blumberg et al. [4] from S^1 to S^2: rejection sampling against a local-count density estimator produces approximately uniform subsamples, the Schonsheck–Schonsheck spherical coordinate is computed on each, kernel-extended to the full dataset, and the maps are aligned and averaged. The new mathematical content is a spherical Procrustes formulation (Def. 3.2) and an approximation theorem (Thm. 3.4, via the great-circle reduction Lem. 3.3) showing the Euclidean O(3) Procrustes solution, projected back to S^2, lies within (1+π)√L* of the geodesic optimum. Experiments show: two implementation accelerations give >100× speedup of the baseline; subsampling removes the O(n^4) memory ceiling and runs at near-constant cost to 10,000 points; under uniform sampling the method matches the baseline, and under von Mises–Fisher and other biased sampling it recovers azimuth/elevation where the baseline fails. Code and reproduction scripts are public.

Significance. If the empirical robustness holds up, this closes a genuine gap: density-robust topological coordinates existed only in degree one, and the authors correctly explain why the weighted-harmonic-cocycle mechanism of Paik–Park does not transfer to the sphere-valued variational problem. The paper has real strengths that should be credited: Lemma 3.3 and Theorem 3.4 are clean, correctly proved, and the great-circle reduction is a neat observation; the spherical Procrustes formulation is a natural contribution in its own right; the scalability results (removal of the quartic memory bottleneck, near-constant runtime to 10^4 points) are practically important for the degree-two pipeline; and the authors ship a public implementation with fixed seeds, five-trial runtime means, and scripts reproducing all figures. The limitations section (§5) is unusually candid about what is not proved. The main caveats concern the evidentiary basis of the headline robustness claim, which is currently qualitative, and the regime in which Theorem 3.4 is informative.

major comments (3)
  1. [§4.3, Figs. 5–7] §4.3: the central empirical claim — accurate azimuth/elevation recovery under severe sampling bias — is supported only by scatter-plot figures (Figs. 5–7) at a single configuration each. No numerical error metric is reported: the protocol says coordinates are evaluated 'by correlating' angles against ground truth, but no correlation values or angular errors appear anywhere. The baseline's failure in Fig. 6 is likewise asserted visually. Since robustness is the paper's headline result, I ask for a quantitative table (e.g., circular correlation and/or mean absolute angular error for azimuth and elevation) for baseline vs. subsampling across the uniform, vMF κ=1.3, and the three severe-bias regimes, with variability over seeds and over the subsampling parameters (M, subsample size). This is feasible with the released code and would substantially strengthen the claim.
  2. [§3.1, §4.3] §4.3 ('Robustness to non-uniform sampling') and §3.1: the mechanism argument — that on each rejection-sampled subsample the Möbius-centering constraint is 'essentially satisfied already,' so per-subsample coordinates are unbiased — is heuristic, and nothing quantifies the residual non-uniformity of the subsamples or relates it to per-subsample coordinate bias. Theorem 3.1 is asymptotic (ε_n→0, ε_n n^{1/(2m)}→∞) and gives no guidance at the n=250–1500 used; the text never states how ε and the acceptance scale C are chosen in the experiments. At minimum, report a subsample-uniformity diagnostic (e.g., the distribution of local counts or estimated density after acceptance) alongside the coordinate errors, and state the estimator parameters used per experiment. If the causal story ('rejection sampling restores uniformity, hence coordinates are unbiased') is to be kept, it needs this quantita
  3. [§3.2, Thm. 3.4 and proof] Theorem 3.4 is correct as proved, but its informativeness regime should be stated. On S^2 the trivial diameter bound d ≤ π always holds, so (1+π)√L* is non-vacuous only when L* < (π/(1+π))^2 ≈ 0.58 rad², i.e., mean squared geodesic misalignment below about 0.76 rad RMS. Nothing in the paper bounds L* a priori, and in the severe-bias regimes of Fig. 7 — where per-subsample coordinates are computed on 75–120 points and kernel-extended — L* could plausibly approach this threshold. This does not undermine the theorem's purpose (justifying the Euclidean initialization for the subsequent SO(3) refinement), but the paper should (i) state the non-vacuity condition explicitly, and (ii) report the empirical L* (or the bound value) in the experiments so readers can see the guarantee is operative where invoked. Relatedly, the metric d_{(S^2)^n} and d_{(R^3)^n} on configuration space are never define
minor comments (8)
  1. [§3.2] Proof of Thm. 3.4: 'by the triangle inequality and Theorem 3.3' should read Lemma 3.3 (there is no Theorem 3.3).
  2. [§3.2 / §4.1] Def. 3.2 restricts rotations to SO(3), but the Euclidean relaxation is solved over O(3) (ten Berge), which includes improper rotations. The theorem legitimately compares against the O(3) optimum, but the initialization passed to the SO(3) refinement may be a reflection; a sentence on how this is handled in practice would help.
  3. [§3.1] Notation inconsistency in §3.1: the kernel-average formula defines Φ_i, but the following sentence refers to 'unit vectors ϕ_i(y)'. Also 'an Euclidean relaxation' (p. 14) and mixed ϵ/ε throughout.
  4. [§3.2] Lem. 3.3 cites 'Lemma 4.3 of [4]', an arXiv preprint whose numbering may shift between versions; give the version number or restate the one-dimensional estimate (it is a line or two) for self-containedness.
  5. [§4.2] Figure 4 appears in the float sequence but is never discussed in the text of §4.2 beyond the caption; either integrate it (which configurations dominate runtime?) or cut it.
  6. [§4.3] §4.3 opening: 'Our main goal is to prove the robustness of the construction' overstates what experiments can do; 'demonstrate' is the right verb, consistent with §5.
  7. [§4.1 / §4.3] The azimuth/elevation evaluation is ill-conditioned near the poles (azimuth is undefined there), and two of the severe-bias distributions concentrate mass at the pole(s). The requested quantitative error table should note how polar points are treated, or use a pole-invariant metric (e.g., geodesic error on S² after a best-fit rotation).
  8. [§4] All coordinate-quality experiments are on (noisy) S² itself. Since the motivating applications are data near manifolds with spherical topology rather than literally on S², one non-sphere example (e.g., a genus-0 surface with a prominent H² class) would broaden the evidence; at minimum the scope should be stated.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: density-robust spherical coordinates are a genuine composition of independent prior constructions plus a new Procrustes approximation theorem, not a claim forced by its own inputs.

full rationale

The paper’s derivation chain is: (i) classical spherical coordinates via degree-two persistent cohomology and Möbius-centered harmonic-energy minimization (Schonsheck–Schonsheck [19], external); (ii) rejection sampling against a local-count density estimator to obtain approximately uniform subsamples (consistency from Blumberg et al. [4, Prop. 3.3], external authors); (iii) kernel extension of subsample maps; (iv) alignment by a newly formulated spherical Procrustes problem whose Euclidean O(3) relaxation is proved to approximate the geodesic optimum within (1+π)√L* (Theorem 3.4, proved from the great-circle reduction Lemma 3.3 and triangle inequalities). None of these steps defines the target coordinate in terms of a fitted quantity that is then reported as a prediction; experiments compare recovered azimuth/elevation to synthetic ground truth rather than recycling fitted parameters. Author lists do not overlap with the load-bearing citations [4, 19, 16, 21], so there is no self-citation chain. Empirical caveats in §5 (robustness presently empirical rather than topological; density estimator outlier-sensitive; bound can be loose) are correctness/limitation issues, not circular reductions. The construction is therefore self-contained against its stated inputs.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The central claim rests on standard algebraic topology and on the cited circular subsampling framework, plus several operational choices (subsample size, kernel width, density radius, VR scale) that are set by hand for experiments. No new physical entities are postulated; the main invented mathematical object is the spherical Procrustes problem as formulated for coordinate alignment.

free parameters (4)
  • subsample size and number of subsamples M = experiment-specific (e.g. 30 subsamples of size 100–120)
    Chosen per experiment (e.g. 20×75, 30×100, 30×120); runtime and quality depend on these fixed sizes rather than n.
  • density-estimator radius ε and acceptance scale C
    Control rejection sampling; must be set so subsamples are large enough for persistent H² yet approximately uniform.
  • Gaussian extension bandwidth σ
    Kernel width for extending subsample coordinates to the full cloud; affects smoothness of the consensus map.
  • Vietoris–Rips scale for cohomology
    Chosen inside the lifetime of a prominent degree-two bar on each subsample; standard but still a free experimental choice.
assumptions (5)
  • standard math Brown representability: [X, K(Z,n)] ≅ H^n(X;Z), with S¹ ≃ K(Z,1) and S² ↪ CP∞ ≃ K(Z,2) used to realize degree-one/two coordinates.
    Invoked in Sections 2.2–2.4 as the existence principle for circular and spherical coordinates.
  • standard math Structure theorem for persistence modules over a field: barcodes uniquely decompose interval summands; homology and cohomology barcodes coincide.
    Section 2.1; justifies selecting a prominent degree-two interval as the coordinate class.
  • domain assumption Consistency of the local-count density estimator under ε_n → 0 and ε_n n^{1/(2m)} → ∞ (Blumberg et al., Prop. 3.3).
    Theorem 3.1; load-bearing for the claim that rejection sampling approximately restores uniformity.
  • domain assumption Classical spherical coordinates exist via harmonic-energy minimization with Möbius centering on the 2-skeleton and extension over the 3-skeleton (Schonsheck–Schonsheck).
    Theorem 2.2 / Section 2.4; the subsample maps are black-box calls to this construction.
  • ad hoc to paper On approximately uniform subsamples, Möbius centering does not introduce large non-isometric distortion, so per-subsample coordinates remain faithful.
    Implicit in Section 3 and the experimental narrative of Section 4.3; Section 5 admits this lacks an analytical (Laplacian-style) guarantee.
invented entities (1)
  • Spherical Procrustes problem for coordinate alignment independent evidence
    purpose: Jointly rotate independently computed Φ_i : X → S² and estimate a centroid map Θ minimizing mean squared geodesic discrepancy.
    Definition 3.2; the principal new mathematical device enabling consensus spherical coordinates. It is a natural geometric optimization problem rather than a physical postulate.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Density-Robust Spherical Coordinates from Persistent Cohomology." pith.science (2026). https://pith.science/paper/WENJU54Z

@misc{pith2026260724650,
  author       = {Pith},
  title        = {Pith review of: Density-Robust Spherical Coordinates from Persistent Cohomology},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WENJU54Z}},
  note         = {Machine review of arXiv:2607.24650}
}
abstract

Persistent cohomology provides a principled framework for constructing nonlinear coordinates that reflect the topology of data. However, these topological coordinates can be severely distorted by non-uniform sampling density, limiting their applicability to real-world data. While density-robust circular coordinates have recently been developed, the extension to spherical coordinates remains an open challenge: unlike the circular case, spherical coordinates are obtained through a nonlinear variational problem for sphere-valued maps, to which existing density-correction mechanisms are not directly applicable. In this paper, we introduce the first density-robust construction of spherical coordinates from persistent cohomology. Rather than modifying the coordinate optimization itself, we extend a subsampling-and-alignment framework for circular coordinates to $S^2$, which first computes spherical coordinates on approximately uniform subsamples obtained by rejection sampling and then combines them into a global consensus map. The principal mathematical difficulty is the alignment of independently computed sphere-valued coordinates. We formulate this challenge as a spherical Procrustes problem and establish approximation guarantees for a computationally tractable Euclidean relaxation. Our resulting construction is robust to non-uniform sampling and retains the accuracy of classical spherical coordinates under uniform sampling. Moreover, by computing persistent cohomology only on fixed-size subsamples, our approach avoids the quartic memory bottleneck of the classical spherical coordinate pipeline and scales to substantially larger datasets. We conduct experiments on synthetic data and demonstrate accurate coordinate recovery under severe sampling bias and scalability to datasets of 10,000 points.

Figures

Figures reproduced from arXiv: 2607.24650 by the authors.

Figure 1
Figure 1. Effect of the two accelerations in isolation, comparing the baseline with the optimized baseline. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_1.png] view at source ↗
Figure 2
Figure 2. Overall runtime of the baseline and the optimized baseline on the full spherical coordinate pipeline. Scalability of the subsampling method. The subsampling method removes this limitation. Figure 3a compares the two directly on inputs of 100 to 800 points: the optimized baseline is faster on small clouds, but its runtime climbs with the point count while the subsampling method stays nearly flat, so the two cross at … view at source ↗
Figure 3
Figure 3. Runtime scaling of the subsampling method: a direct comparison against the optimised baseline on moderate inputs, and its near-constant runtime on large inputs [PITH_FULL_IMAGE:figures/full_fig_p019_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Runtime of the subsampling method as a function of subsample size and number of subsamples. 19 [PITH_FULL_IMAGE:figures/full_fig_p019_4.png]
Figure 5
Figure 5. Figure 5: Uniform sampling control experiment. Top: uniform sample over S 2 . Middle and bottom rows: recovered azimuth and elevation angles (respectively), each compared against the ground truth. Columns in these rows correspond to the baseline method and the two subsampling co…
Figure 6
Figure 6. Figure 6: Embeddings and angle recovery on the non-uniform dataset. Top: input sampled from a von Mises–Fisher distribution centered at (0, 0, 1). Middle: baseline spherical coordinates. Bottom: the proposed subsampling method. The baseline suffers from strong distortion caused …
Figure 7
Figure 7. Figure 7: Performance of the subsampling method under three strongly non-uniform sampling distributions. In all cases both azimuth and elevation are recovered with high accuracy despite the uneven density. 23 [PITH_FULL_IMAGE:figures/full_fig_p023_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

24 extracted references · 8 canonical work pages

  1. [4]

    Subsampling, aligning, and averaging to find circular coordinates in recurrent time series

    Andrew J Blumberg et al. “Subsampling, aligning, and averaging to find circular coordinates in recurrent time series”. In:arXiv preprint arXiv:2412.18515(2024) (cit. on pp. 2, 3, 7, 9, 11–14, 16, 25)

    arXiv 2024

  2. [1]

    Corrections and Supplementaries to My Paper Concerning Krull-Remak- Schmidt’s Theorem

    Gorˆ o Azumaya. “Corrections and Supplementaries to My Paper Concerning Krull-Remak- Schmidt’s Theorem”. In:Nagoya Mathematical Journal1 (June 1950), pp. 117–124.issn: 0027-7630, 2152-6842.doi:10.1017/S002776300002290X(cit. on p. 4)

    work page doi:10.1017/s002776300002290x(cit 1950

  3. [2]

    Ripser: efficient computation of Vietoris–Rips persistence barcodes

    Ulrich Bauer. “Ripser: efficient computation of Vietoris–Rips persistence barcodes”. In:Pro- ceedings of the 37th International Symposium on Computational Geometry (SoCG 2021). Vol. 189. Leibniz International Proceedings in Informatics (LIPIcs). Schloss Dagstuhl–Leibniz- Zentrum f¨ ur Informatik, 2021, 11:1–11:18.doi:10 . 4230 / LIPIcs . SoCG . 2021 . 11(...

    2021

  4. [3]

    A General Solution to Orthogonal Procrustes Problems

    Jos M. F. ten Berge. “A General Solution to Orthogonal Procrustes Problems”. In:Psy- chometrika51.2 (1986), pp. 267–276.doi:10.1007/BF02293996(cit. on p. 16)

    work page doi:10.1007/bf02293996(cit 1986

  5. [5]

    Decomposition of Persistence Modules

    Magnus Botnan and William Crawley-Boevey. “Decomposition of Persistence Modules”. In: Proceedings of the American Mathematical Society148.11 (Aug. 2020), pp. 4581–4596.issn: 0002-9939, 1088-6826.doi:10.1090/proc/14790(cit. on p. 4)

    work page doi:10.1090/proc/14790(cit 2020

  6. [6]

    Topology and data

    Gunnar Carlsson. “Topology and data”. In:Bulletin of the American Mathematical Society 46.2 (2009), pp. 255–308 (cit. on pp. 2, 3)

    2009

  7. [7]

    Decomposition of Pointwise Finite-Dimensional Persistence Mod- ules

    William Crawley-Boevey. “Decomposition of Pointwise Finite-Dimensional Persistence Mod- ules”. In:Journal of Algebra and Its Applications14.05 (June 2015), p. 1550066.issn: 0219- 4988.doi:10.1142/S0219498815500668(cit. on p. 4)

    work page doi:10.1142/s0219498815500668(cit 2015

  8. [8]

    Persistent homology-a survey

    Herbert Edelsbrunner and John Harer. “Persistent homology-a survey”. In:Contemporary mathematics453.26 (2008), pp. 257–282 (cit. on p. 3)

    2008

Show all 24 references
  1. [9]

    Barcodes: The Persistent Topology of Data

    Robert Ghrist. “Barcodes: The Persistent Topology of Data”. In:Bulletin of the American Mathematical Society45.1 (2008), pp. 61–75 (cit. on p. 2). 25

    2008

  2. [10]

    John C Gower and Garmt B Dijksterhuis.Procrustes problems. Vol. 30. Oxford university press, 2004 (cit. on p. 10)

    2004

  3. [11]

    Cambridge, UK: Cambridge University Press, 2002 (cit

    Allen Hatcher.Algebraic Topology. Cambridge, UK: Cambridge University Press, 2002 (cit. on p. 2)

    2002

  4. [12]

    Numba: A LL VM-based Python JIT Compiler

    Siu Kwan Lam, Antoine Pitrou, and Stanley Seibert. “Numba: A LL VM-based Python JIT Compiler”. In:Proceedings of the Second Workshop on the LL VM Compiler Infrastructure in HPC. LL VM ’15. Austin, TX, USA: Association for Computing Machinery, 2015, pp. 1–6. doi:10.1145/2833157...

    2015

  5. [13]

    Visualizing Data Using t-SNE

    Laurens van der Maaten and Geoffrey Hinton. “Visualizing Data Using t-SNE”. In:Journal of Machine Learning Research9 (2008), pp. 2579–2605 (cit. on p. 1)

    2008

  6. [14]

    UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction

    Leland McInnes, John Healy, and James Melville. “UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction”. In:arXiv preprint(2018). arXiv:1802 . 03426 [stat.ML](cit. on p. 1)

    2018

  7. [15]

    Geomstats: A Python Package for Riemannian Geometry in Machine Learning

    Nina Miolane et al. “Geomstats: A Python Package for Riemannian Geometry in Machine Learning”. In:Journal of Machine Learning Research21.223 (2020), pp. 1–9.url:http : //jmlr.org/papers/v21/19-027.html(cit. on p. 16)

    2020

  8. [16]

    Circular Coordinates for Density-Robust Analysis

    Taejin Paik and Jaemin Park. “Circular Coordinates for Density-Robust Analysis”. In:arXiv preprint(2023). arXiv:2301.12742 [cs.LG](cit. on pp. 2, 11, 24)

    2023 arXiv

  9. [17]

    Peter Petersen.Riemannian Geometry. 3rd ed. Vol. 171. Graduate Texts in Mathematics. Springer, 2016.doi:10.1007/978-3-319-26654-1(cit. on p. 16)

    2016 doi

  10. [18]

    Nonlinear Dimensionality Reduction by Locally Linear Embedding

    Sam T. Roweis and Lawrence K. Saul. “Nonlinear Dimensionality Reduction by Locally Linear Embedding”. In:Science290.5500 (2000), pp. 2323–2326 (cit. on p. 1)

    2000

  11. [19]

    Spherical coordinates from persistent coho- mology

    Nikolas C Schonsheck and Stefan C Schonsheck. “Spherical coordinates from persistent coho- mology”. In:Journal of Applied and Computational Topology8.1 (2024), pp. 149–173 (cit. on pp. 2, 3, 5, 7, 9, 16, 24)

    2024

  12. [20]

    Dualities in Persistent (Co)Homology

    Vin de Silva, Dmitriy Morozov, and Mikael Vejdemo-Johansson. “Dualities in Persistent (Co)Homology”. In:Inverse Problems27.12 (Nov. 2011), p. 124003.issn: 0266-5611.doi: 10.1088/0266-5611/27/12/124003(cit. on p. 4)

    2011 doi

  13. [21]

    Persistent cohomology and circular coordinates

    Vin de Silva, Dmitriy Morozov, and Mikael Vejdemo-Johansson. “Persistent cohomology and circular coordinates”. In:Discrete & Computational Geometry45.4 (2011), pp. 737–759.doi: 10.1007/s00454-011-9344-x(cit. on pp. 2–6)

    2011 doi

  14. [22]

    A Global Geometric Framework for Nonlinear Dimensionality Reduction

    Joshua B. Tenenbaum, Vin De Silva, and John C. Langford. “A Global Geometric Framework for Nonlinear Dimensionality Reduction”. In:Science290.5500 (2000), pp. 2319–2323 (cit. on p. 1)

    2000

  15. [23]

    SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python

    Pauli Virtanen et al. “SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python”. In:Nature Methods17 (2020), pp. 261–272.doi:10.1038/s41592-019-0686-2(cit. on p. 16)

    2020 doi

  16. [24]

    Computing persistent homology

    Afra Zomorodian and Gunnar Carlsson. “Computing persistent homology”. In:Proceedings of the twentieth annual symposium on Computational geometry. Association for Computing Machinery, 2004, pp. 347–356 (cit. on p. 3). 26

    2004

Pith tools

Reviewed July 31, 2026 · model on record in the stance chip above.