Pith. sign in

REVIEW 2 major objections 3 minor 61 references

Don't be Afraid of Cell Complexes! An Introduction from an Applied Perspective

T0 review · 2 major / 3 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper defines cell complexes purely by boundary matrices and proves equivalence with the topological notion for dimensions up to two and for every computational aspect it discusses.

desk verdict A useful, honest tutorial that packages standard algebraic topology into a matrix-only formalism for applied users; the 2D equivalence theorem is correct, but the unqualified higher-dimensional equivalence claim needs tightening. read the letter →

arxiv 2506.09726 v1 pith:W3GPXRTN submitted 2025-06-11 eess.SP cs.CGcs.SImath.AT

Josef Hoppe , Vincent P. Grande , Michael T. Schaub This is my paper · ORCID
classification eess.SPcs.CGcs.SImath.AT MSC 55U1055N31
keywords abstractcellcomplexesboundarymatricesregularhigher-ordernetworkstopologicalsignalprocessingHodgeLaplaciancomplexneuralalgebraictopologyfordata
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 is trying to make cell complexes, the higher-order network model built by gluing balls of arbitrary dimension, accessible to signal processing and network science researchers who do not want to learn point-set topology. It introduces abstract regular cell complexes (ARCCs), defined purely in terms of boundary matrices $B_k\in\{0,\pm1\}^{|C_{k-1}|\times|C_k|}$ together with the condition that every cell's iterated boundary has the homology of a point. The central claim is that for computational purposes an ARCC is interchangeable with an ordinary regular cell complex: every regular CW complex gives an ARCC with the same boundary matrices, every ARCC of dimension at most two is realizable as a regular CW complex, and in higher dimensions the two are indistinguishable for every computational tool discussed. Because the definition uses only matrices, it is easy to check and to implement, and it connects directly to homology, Hodge Laplacians, persistent homology, and cell complex neural networks.

What carries the argument

The central object is the boundary matrix $B_k$, each entry in $\{0,\pm1\}$ recording whether a $(k-1)$-cell lies in the boundary of a $k$-cell and whether the reference orientations agree or oppose. The matrices alone determine the chain spaces $\mathbb{R}^{|C_k|}$, the boundary maps, the chain-complex identity $B_{k-1}B_k=0$, the Hodge Laplacians $L_k = B_{k+1}B_{k+1}^{\top} + B_k^{\top}B_k$, the Hodge decomposition into gradient, curl, and harmonic parts, the total Dirac operator $B+B^{\top}$, and all signal-processing and message-passing operations built on them. The second condition in Definition A.1 is the algebraic substitute for the topological property that each cell is a disk: it asks that the restricted chain complex of a cell's closure has $H_0\cong\mathbb{Z}$ and zero higher homology.

What would settle it

Find an ARCC of dimension 3 whose top cell has as its boundary a triangulated homology sphere that is not the standard 3-sphere, and demonstrate that one of the paper's computational constructions, say the persistent homology of a filtration, the eigenvectors of a Hodge Laplacian, or a cell complex neural network's output, differs between this ARCC and any genuine regular CW complex with the same boundary matrices. The paper itself acknowledges such objects exist, so the falsifier tests whether the 'indistinguishable for computational aspects' claim actually holds.

Watch

Extended reading notes

Core claim

The paper's central claim is that a cell complex, for computational purposes, is nothing more than its sequence of boundary matrices. Definition A.1 formalises an abstract regular cell complex as finite ordered sets of cells $C_0,\dots,C_n$ and boundary matrices $B_k\in\{0,\pm1\}$ such that every column of $B_1$ has exactly one $+1$ and one $-1$, and such that for each cell the chain complex of its iterated boundary has trivial homology in every positive degree and $H_0\cong\mathbb{Z}$. Theorem A.5 says every regular CW complex yields such an ARCC with the same boundary matrices, every ARCC of dimension at most two is realisable as a regular CW complex with the same boundary matrices, and the equivalence extends to all computational considerations in higher dimensions. The paper is explicit that the lack of full equivalence in higher dimensions is tied to the algorithmic undecidability of sphere recognition, and it provides a simplified definition for dimension two: an oriented multigraph with no self-loops plus 2-cells attached along simple oriented cycles.

Load-bearing premise

The load-bearing premise is that a cell complex is fully and faithfully represented by its boundary-matrix chain complex for every computational use the paper considers, so topological distinctions that the chain complex cannot see, such as whether a cell's closure is a genuine disk or merely a homology ball, are irrelevant.

Editorial extensions

If this is right

  • Because ARCCs are defined by matrices, software libraries can store and manipulate cell complexes with the same ease as graphs, eliminating the need for a topological data structure.
  • Any object passing the two conditions of Definition A.1 automatically carries well-defined Hodge Laplacians and a Hodge decomposition, so signal processing on $\mathbb{R}^{|C_k|}$ is immediately available.
  • The dimension-two characterisation means a 2-dimensional cell complex can be specified simply as an oriented multigraph plus a list of simple cycles, which makes constructing cell complexes from graph data straightforward.
  • The equivalence result means that homology, persistent homology, and spectra of Hodge Laplacians computed from an ARCC coincide with those of any regular CW complex realising it, so practitioners can compute with the matrices and trust the topological meaning.
  • Verifying that a proposed matrix object is an ARCC is computationally feasible, replacing the uncheckable universal conditions of pushout squares.

Reading between the lines

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

  • The paper's interchangeability claim suggests that, within the class of chain-level invariants, two ARCCs with isomorphic boundary-matrix chain complexes are the same object; this offers a working definition of equivalence for computational topology that sidesteps undecidable geometric questions.
  • The simplified 2D definition directly opens the door to random cell complex models that sample simple cycles in a graph, which would give the field the generative null models the paper lists as an open challenge.
  • The undecidability caveat implies that no verification tool can ever certify topological realizability for a general high-dimensional ARCC; practical pipelines should therefore either restrict to dimensions at most two or treat chain-level invariants as the object of study.
  • If the equivalence holds for message-passing neural networks, then any cell-complex network architecture that uses only boundary matrices is equally valid on ARCCs and on true topological complexes, which would let liftings to cell complexes be optimised purely combinatorially.
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

2 major / 3 minor

Summary. The paper introduces abstract regular cell complexes (ARCCs), an algebraic and combinatorial model intended to stand in for regular CW complexes in computational applications. It gives an elementary definition for dimension at most 2 (Section 3), a general definition in Appendix A, develops chain spaces, boundary matrices, homology, Hodge Laplacians, weights, and an overview of applications in signal processing and neural-network-style message passing. The formal results are Lemma 3.6 (B1B2 = 0 for 2-dimensional ARCCs), Lemma A.4 (the general boundary matrices form a chain complex), and Theorem A.5 (every regular CW complex yields an ARCC, and every ARCC of dimension at most 2 is realizable as a regular CW complex with the same boundary matrices).

Significance. The pedagogical goal is valuable: the dimension-2 definition and the worked examples make the boundary-matrix perspective much more accessible than the usual point-set topology definitions, and the proved statements (Lemma 3.6, Theorem A.5 for dimension at most 2, Lemma A.4) are correct once the notation in Definition A.1 is fixed. The survey of applications, lifting procedures, and weighting schemes is useful and reasonably current. The main weakness is that the paper claims an equivalence in arbitrary dimension that its axioms do not support and that the appendix itself qualifies; this overstatement should be corrected before publication. The underlying dimension-2 machinery is sound and worth publishing after revision.

major comments (2)
  1. [Appendix A, Definition A.1(2a)] Condition 2(a) as printed reads 'ker B_k = 0', with B_k the global boundary matrix. This cannot be the intended condition: a regular CW decomposition of the 3-sphere obtained by gluing two 3-cells to a common 2-sphere has B_3 with two identical columns and hence nonzero kernel, although every cell closure is a disk and the complex should be an ARCC. The first direction of Theorem A.5 therefore fails under the literal wording. Replace 'ker B_k = 0' by 'ker \hat B_k = 0', consistently with the restricted matrices \hat B_l introduced in the same sentence.
  2. [Section 1, Contribution and outline; cf. Remark A.2] The sentence claiming that ARCCs are 'equivalent for all computational considerations in arbitrary dimensions' overstates what is shown and, on a literal reading, is false. Definition A.1 condition 2 is an acyclicity condition on the cellular chain complex of a cell closure, not a condition that the closure be a disk. A dimension-3 ARCC can be formed from the (regularly subdivided) 2-skeleton K of a homology 3-sphere with perfect fundamental group by adding one 3-cell attached along a generator of H_2(K); the closure satisfies H_0 = Z and H_i = 0 for i > 0, so condition 2 holds, but the closure is not a 3-disk and the ARCC is not realizable as a regular CW complex. The 'all computational considerations' phrase should be replaced by the qualified statement in Remark A.2, and the specific computational invariants that are determined by the boundary matrices (homology, Hodge spectra, persistence, boundary-based message passing) should be listed explicitly.
minor comments (3)
  1. [Definition 3.2] The text says that 'each row of B2 corresponds to a simple oriented cycle', but since B2 is a |C1| x |C2| matrix, it is the columns that correspond to 2-cells and hence to cycles; rows correspond to 1-cells. This should be corrected in the main definition used throughout the tutorial.
  2. [Appendix A, Lemma A.4 proof] The proof says 'It suffices to show that B_{k-1} B_k c_i^k = 0 for arbitrary cells c_i^k for k <= 2'; the intended range is k >= 2 (or 'for all k'), since the argument invokes condition 2(a) for the top dimension k.
  3. [Appendix A, text before Theorem A.5] The sentence 'A full correspondence between a computational definition (as ours) and a topological definition is not possible to to some results' contains a typo ('to to') and should read 'due to some results' or similar.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the ARCC definition and Theorem A.5 are proven from first principles; self-citations are motivational and not load-bearing.

full rationale

The paper introduces abstract regular cell complexes via explicit boundary-matrix axioms and proves Theorem A.5 from those axioms against standard topological definitions, with no fitted parameters or predicted quantities. The definition of an ARCC imposes acyclicity conditions on minimal subcomplexes, and the theorem verifies that regular CW complexes satisfy these conditions and that, up to dimension two, the conditions are sufficient to construct a regular CW complex with the same boundary matrices. This is a mathematical equivalence proof, not a reduction of the conclusion to its premises: the axioms do not presuppose the existence of the CW realization. Citations to the authors' own prior work appear only in the survey of applications and methods (e.g., [9], [18], [19], [26], [27]) and are not used to justify the central definition or theorem. The acknowledged limitation that arbitrary-dimensional ARCCs may not be topologically equivalent to CW complexes (Remark A.2; the non-decidability caveat before Theorem A.5) is a correctness/false-precision concern about the phrase 'all computational considerations', not a circularity. No self-definitional, fitted-input, or self-citation-load-bearing pattern is present.

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

No free parameters are fitted and no new physical entities are introduced. The central claim rests on standard algebraic topology (CW complexes, cellular chain complexes), standard linear algebra (Hodge decomposition), a cited undecidability result, and one domain assumption that applications only need regular cells. The ARCC axiomatization itself is the paper's contribution rather than an unsupported postulate.

assumptions (5)
  • standard math Standard definition of finite regular CW complexes, built as pushout squares whose attaching maps are homeomorphisms onto their images (Appendix B, Definition B.1).
    The paper uses this as the external benchmark in Theorem A.5 and to justify the regularity conditions.
  • standard math Cellular chain complex and homology of a CW complex are computed from boundary matrices; B_k B_{k+1} = 0 and homology is ker/im (standard algebraic topology, Hatcher 2002).
    The paper identifies boundary matrices with the chain complex and uses H_k = ker B_k / Im B_{k+1} throughout.
  • standard math Novikov's undecidability result on recognizing spheres, as cited via Volodin et al. [49].
    Used in Remark A.2 and after Theorem A.5 to argue that full ARCC-to-CW equivalence is impossible in general dimensions, so only computational equivalence is claimed.
  • standard math Hodge decomposition: each chain space splits orthogonally as Im B_{k+1} ⊕ Im B_k^T ⊕ ker L_k (Theorem 5.7, standard linear algebra).
    Stated and sketched in Section 5.2 without a proof or citation; it follows from the fundamental theorem of linear algebra applied to boundary matrices.
  • domain assumption Applied signal processing and network science only require regular cell complexes, so non-regular CW complexes can be ignored (Remark A.6).
    The paper states that both applications call for regular cell complexes; this restricts the scope of the equivalence claim. If an application required non-regular cells, the ARCC definition would not apply.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Don't be Afraid of Cell Complexes! An Introduction from an Applied Perspective." pith.science (2026). https://pith.science/paper/W3GPXRTN

@misc{pith2026250609726,
  author       = {Pith},
  title        = {Pith review of: Don't be Afraid of Cell Complexes! An Introduction from an Applied Perspective},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/W3GPXRTN}},
  note         = {Machine review of arXiv:2506.09726}
}
read the original abstract

Cell complexes (CCs) are a higher-order network model deeply rooted in algebraic topology that has gained interest in signal processing and network science recently. However, while the processing of signals supported on CCs can be described in terms of easily-accessible algebraic or combinatorial notions, the commonly presented definition of CCs is grounded in abstract concepts from topology and remains disconnected from the signal processing methods developed for CCs. In this paper, we aim to bridge this gap by providing a simplified definition of CCs that is accessible to a wider audience and can be used in practical applications. Specifically, we first introduce a simplified notion of abstract regular cell complexes (ARCCs). These ARCCs only rely on notions from algebra and can be shown to be equivalent to regular cell complexes for most practical applications. Second, using this new definition we provide an accessible introduction to (abstract) cell complexes from a perspective of network science and signal processing. Furthermore, as many practical applications work with CCs of dimension 2 and below, we provide an even simpler definition for this case that significantly simplifies understanding and working with CCs in practice.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 56 canonical work pages

  1. [1]

    Barbarossa and S

    S. Barbarossa and S. Sardellitti. ‘Topological signal processing over simplicial com- plexes’. In: IEEE Transactions on Signal Processing 68 (2020), pp. 2992–3007

    work page 2020

  2. [2]

    Battiloro, I

    C. Battiloro, I. Spinelli, L. Telyatnikov, M. M. Bronstein, S. Scardapane and P . D. Lorenzo. ‘From Latent Graph to Latent Topology Inference: Differentiable Cell Complex Module’. In: The Twelfth International Conference on Learning Representations . 2024. url: https://openreview.net/forum?id=0JsRZEGZ7L

    work page 2024

  3. [3]

    Battiston, G

    F. Battiston, G. Cencetti, I. Iacopini, V . Latora, M. Lucas, A. Patania, J.-G. Young and G. P etri. ‘Networks beyond pairwise interactions: Structure and dynamics’. In: Physics Reports 874 (2020), pp. 1–92

    work page 2020

  4. [4]

    Benjamin, L

    K. Benjamin, L. Mukta, G. Moryoussef, C. Uren, H. A. Harrington, U. Tillmann and A. Barbensi. ‘Homology of homologous knotted proteins’. In: Journal of the Royal Society Interface 20.201 (2023), p. 20220727

    work page 2023

  5. [5]

    C. Bick, E. Gross, H. A. H arrington and M. T. Schaub. ‘What are higher-order net- works?’ In: SIAM Review 65.3 (2023), pp. 686–731

    work page 2023

  6. [6]

    Bodnar, F

    C. Bodnar, F. Frasca, N. Otter, Y. Wang, P . Lio, G. F. Montufar and M. Bronstein. ‘Weisfeiler and lehman go cellular: Cw networks’. In: Advances in Neural Information Processing Systems 34 (2021), pp. 2625–2640

    work page 2021

  7. [7]

    Bredon, J

    G. Bredon, J. Ewing, F. Gehring and P . Halmos. Topology and Geometry. Graduate Texts in Mathematics. New York: Springer, 1993

    work page 1993

  8. [8]

    M. M. B ronstein, J. Bruna, Y. LeCun, A. Szlam and P . Vandergheynst. ‘Geometric deep learning: going beyond euclidean data’. In: IEEE Signal Processing Magazine 34.4 (2017), pp. 18–42

    work page 2017

Show all 61 references
  1. [9]

    Calmon, M

    L. Calmon, M. T. Schaub and G. Bianconi. ‘Dirac signal processing of higher-order topological signals’. In: New Journal of Physics 25.9 (2023), p. 093013

    2023

  2. [10]

    Carlsson

    G. Carlsson. ‘Topological methods for data modelling’. In: Nature Reviews Physics 2.12 (2020), pp. 697–708

    2020

  3. [11]

    C arlsson and M

    G. C arlsson and M. Vejdemo-Johansson. Topological data analysis with applications . Cambridge University Press, 2021

    2021

  4. [12]

    A. X. Chang, T. Funkhouser, L. Guibas, P . Hanrahan, Q. Huang, Z. Li, S. Savarese, M. Savva, S. Song, H. Su et al. ‘Shapenet: An information-rich 3d model repository’. In: arXiv preprint arXiv:1512.03012 (2015)

    2015 arXiv

  5. [13]

    Chazal and B

    F. Chazal and B. Michel. ‘An introduction to topological data analysis: fundamental and practical aspects for data scientists’. In: Frontiers in artificial intelligence 4 (2021), p. 667963

    2021

  6. [14]

    F. R. C hung. Spectral graph theory. Vol. 92. American Mathematical Soc., 1997

    1997

  7. [15]

    Derrible and C

    S. Derrible and C. Kennedy. ‘Applications of graph theory and network science to transit network design’. In: Transport reviews 31.4 (2011), pp. 495–519

    2011

  8. [16]

    tom Dieck

    T. tom Dieck. Algebraic topology. Vol. 8. Zürich: European Mathematical Society, 2008

    2008

  9. [17]

    X. Dong, D. Thanou, L. Toni, M. Bronstein and P . Frossard. ‘Graph signal processing for machine learning: A review and new perspectives’. In: IEEE Signal processing magazine 37.6 (2020), pp. 117–127. 34

    2020

  10. [18]

    Frantzen, J.-B

    F. Frantzen, J.-B. Seby and M. T. Schaub. ‘Outlier detection for trajectories via flow- embeddings’. In: 2021 55th Asilomar Conference on Signals, Systems, and Computers . IEEE. 2021, pp. 1568–1572

    2021

  11. [19]

    V . P . Grande, J. Hoppe, F. Frantzen and M. T. Schaub. ‘Topological Trajectory Classifica- tion and Landmark Inference on Simplicial Complexes’. In: 2024 58th Asilomar Conference on Signals, Systems, and Computers. 2024, pp. 44–48

    2024

  12. [20]

    Hajij, K

    M. Hajij, K. Istvan and G. Zamzmi. ‘Cell complex neural networks’. In: arXiv preprint arXiv:2010.00743 (2020)

    2020 arXiv

  13. [21]

    Hajij, M

    M. Hajij, M. Papillon, F. Frantzen, J. Agerberg, I. AlJabea, R. Ballester, C. Battiloro, G. Bernárdez, T. Birdal, A. Brent et al. ‘TopoX: a suite of Python packages for machine learning on topological domains’. In: Journal of Machine Learning Research 25.374 (2024), pp. 1–8

    2024

  14. [22]

    Hajij, G

    M. Hajij, G. Zamzmi, T. Papamarkou, N. Miolane, A. Guzmán-Sáenz, K. N. Ramamurthy, T. Birdal, T. K. D ey, S. Mukherjee, S. N. S amaga et al. ‘Topological deep learning: Going beyond graph data’. In: arXiv preprint arXiv:2206.00606 (2022)

    2022 arXiv

  15. [23]

    Hansen and R

    J. Hansen and R. Ghrist. ‘Toward a spectral theory of cellular sheaves’. In: Journal of Applied and Computational Topology 3.4 (2019), pp. 315–358

    2019

  16. [24]

    H atcher

    A. H atcher. Algebraic Topology. Cambridge University Press, 2002

    2002

  17. [25]

    Hawasly, S

    M. Hawasly, S. Ramamoorthy et al. ‘Multiscale topological trajectory classification with persistent homology’. In: Proceedings of Robotics: Science and Systems X 2014. 2014

    2014

  18. [26]

    H oppe and M

    J. H oppe and M. T. S chaub. ‘Representing Edge Flows on Graphs via Sparse Cell complexes’. In: Proceedings of the Second Learning on Graphs Conference . Ed. by S. Villar and B. Chamberlain. Vol. 231. Proceedings of Machine Learning Research. PMLR, 27–30 Nov 2024, 1:1–1:22. ur...

    2024

  19. [27]

    Hoppe and M

    J. Hoppe and M. T. Schaub. ‘Random Abstract Cell complexes’. In: arXiv Preprints (2024). arXiv:2406.019999. arXiv: 2406.01999 [cs.DS]

    2024

  20. [28]

    Isufi, G

    E. Isufi, G. Leus, B. Beferull-Lozano, S. Barbarossa and P . Di Lorenzo. ‘Topological Signal Processing and Learning: Recent Advances and Future Challenges’. In: arXiv preprint arXiv:2412.01576 (2024)

    2024 arXiv

  21. [29]

    Lawson, A

    P . Lawson, A. B. S holl, J. Q. B rown, B. T. Fasy and C. Wenk. ‘Persistent homology for the quantitative evaluation of architectural features in prostate cancer histology’. In: Scientific reports 9.1 (2019), p. 1139

    2019

  22. [30]

    Y. Liu, C. Kang, S. Gao, Y. Xiao and Y. Tian. ‘Understanding intra-urban trip patterns from taxi trajectory data’. In: Journal of geographical systems 14.4 (2012), pp. 463–483

    2012

  23. [31]

    Marinucci, C

    L. Marinucci, C. Battiloro and P . Di Lorenzo. ‘Topological adaptive learning over cell complexes’. In: 2024 32nd European Signal Processing Conference (EUSIPCO). IEEE. 2024, pp. 832–836

    2024

  24. [32]

    Morris, N

    C. Morris, N. M. Kriege, F. Bause, K. Kersting, P . Mutzel and M. Neumann. ‘Tudata- set: A collection of benchmark datasets for learning with graphs’. In: arXiv preprint arXiv:2007.08663 (2020)

    2020 arXiv

  25. [33]

    Mulder and G

    D. Mulder and G. Bianconi. ‘Network geometry and complexity’. In: Journal of Statistical Physics 173 (2018), pp. 783–805. 35

    2018

  26. [34]

    E. Munch. ‘A user’s guide to topological data analysis’. In: Journal of Learning Analytics 4.2 (2017), pp. 47–61

    2017

  27. [35]

    M. N ewman. Networks. Oxford University Press, 2018

    2018

  28. [36]

    Ortega, P

    A. Ortega, P . Frossard, J. Kovaˇcevi´c, J. M. M oura and P . Vandergheynst. ‘Graph signal processing: Overview, challenges, and applications’. In: Proceedings of the IEEE 106.5 (2018), pp. 808–828

    2018

  29. [37]

    T. M. Roddenberry, V . P . Grande, F. Frantzen, M. T. Schaub and S. Segarra. ‘Signal processing on product spaces’. In: ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE. 2023, pp. 1–5

    2023

  30. [38]

    T. M. Roddenberry, M. T. Schaub and M. Hajij. ‘Signal processing on cell complexes’. In: ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE. 2022, pp. 8852–8856

    2022

  31. [39]

    Russakovsky, J

    O. Russakovsky, J. Deng, H. Su, J. Krause, S. Satheesh, S. Ma, Z. Huang, A. Karpathy, A. Khosla, M. Bernstein et al. ‘Imagenet large scale visual recognition challenge’. In: International journal of computer vision 115 (2015), pp. 211–252

    2015

  32. [40]

    Sardellitti and S

    S. Sardellitti and S. Barbarossa. ‘Topological Signal Processing over Generalized Cell complexes’. In: IEEE Transactions on Signal Processing (2024)

    2024

  33. [41]

    Sardellitti, S

    S. Sardellitti, S. Barbarossa and L. Testa. ‘Topological signal processing over cell complexes’. In: 2021 55th Asilomar Conference on Signals, Systems, and Computers . IEEE. 2021, pp. 1558–1562

    2021

  34. [42]

    M. T. Schaub, A. R. Benson, P . Horn, G. Lippner and A. Jadbabaie. ‘Random walks on simplicial complexes and the normalized Hodge 1-Laplacian’. In: SIAM Review 62.2 (2020), pp. 353–391

    2020

  35. [43]

    M. T. Schaub and S. Segarra. ‘Flow smoothing and denoising: Graph signal processing in the edge-space’. In: 2018 IEEE Global Conference on Signal and Information Processing (GlobalSIP). IEEE. 2018, pp. 735–739

    2018

  36. [44]

    M. T. Schaub, Y. Zhu, J.-B. Seby, T. M. Roddenberry and S. Segarra. ‘Signal processing on higher-order networks: Livin’on the edge... and beyond’. In: Signal Processing 187 (2021), p. 108149

    2021

  37. [45]

    C. L. da Silva, L. M. P etry and V . Bogorny. ‘A survey and comparison of trajectory classification methods’. In: 2019 8th Brazilian conference on intelligent systems (BRACIS) . IEEE. 2019, pp. 788–793

    2019

  38. [46]

    S. H. S trogatz. ‘Exploring complex networks’. In: nature 410.6825 (2001), pp. 268–276

    2001

  39. [47]

    A. M. S ykulski, S. C. O lhede, J. M. L illy and E. Danioux. ‘Lagrangian time series models for ocean surface drifter trajectories’. In: Journal of the Royal Statistical Society Series C: Applied Statistics 65.1 (2016), pp. 29–50

    2016

  40. [48]

    M. M. S ysło. ‘On cycle bases of a graph’. In: Networks 9.2 (1979), pp. 123–132

    1979

  41. [49]

    I. A. Volodin, V . E. Kuznetsov and A. T. Fomenko. ‘The problem of discriminating algorithmically the standard three-dimensional sphere’. In: Russian Mathematical Surveys 29.5 (1974), p. 71

    1974

  42. [50]

    Wasserman

    L. Wasserman. ‘Topological data analysis’. In: Annual review of statistics and its application 5.2018 (2018), pp. 501–532. 36

    2018

  43. [51]

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang and P . S. Yu. ‘A comprehensive survey on graph neural networks’. In: IEEE transactions on neural networks and learning systems 32.1 (2020), pp. 4–24. A. A bstract Regular Cell Complexes of Arbitrary Dimension In this section, we will ...

    2020

  44. [52]

    All columns of B1 have exactly one positive and one negative entry

  45. [53]

    Remark A.2 (On the definition of abstract cell complexes)

    For every cell ci k ∈ Ck and sets of cells ˆCk−i = ∂i{ci k} for 0 ≤ i ≤ k, the associated pair ( ˆC∗, ˆB∗) with boundary matrices ˆBl : Z| ˆCl | → Z| ˆCl−1| restricted from Bl fulfills a) ker Bk = 0 and ker ˆBl−1 = Im ˆBl for all 2 ≤ l ≤ k, b) Z| ˆC0|/ Im ˆB1 ∼= Z. Remark A.2 ...

  46. [54]

    However, the definition does not allow for arbitrary attachments

    This then continues for arbitrary k-cells which are attached to (k − 1)-cells. However, the definition does not allow for arbitrary attachments. In the usual definition of topological regular cell complexes, we allow only for the attachment of k-disks, i.e. 37 Dk = {x ∈ Rk : |...

  47. [55]

    The algebraic structure of the cw complex in terms of its boundary matrices and associated Hodge Laplacians must be easily obtainable from the definition

  48. [56]

    Checking whether a given object X satisfies the definition of cw must be compu- tationally feasible

  49. [57]

    For the first point, all computational methods introduced or discussed in this paper rely on the boundary matrices or the Hodge Laplacians in some form or another

    The definition must align with the underlying geometric intuition of cells repres- enting real-world lines, areas, volumes, etc. For the first point, all computational methods introduced or discussed in this paper rely on the boundary matrices or the Hodge Laplacians in some f...

  50. [58]

    For all subspaces Xi and Xj, the intersection of Xj and the closure of Xi is nonempty, ¯Xi ∩ Xj ̸= only if Xj is a subset of ¯Xi, Xj ⊆ ¯Xi

  51. [59]

    While this definition is very concise, it provides us with little insight on how to construct the cell complex

    For every subspace Xi, there is a homeomorphism from ¯Xi to a closed disk Dni = {x ∈ Rni : ∥x∥ ≤1} restricting to a homeomorphism between Xi and the open disk ˚Dni. While this definition is very concise, it provides us with little insight on how to construct the cell complex. ...

  52. [60]

    For all vertices s ∈ S, s is a cell in X

  53. [61]

    For two cells x, y ∈ Xwith x ⊆ y, we have that rk(x) ≤ rk(y). While the above definition is very easy to verify, it does not allow for a geometric intu- ition and does not have the structural properties needed for computational applications requiring the boundary matrices. 41

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.