Inductive construction of path homology chains

Matthew Burfitt; Tyrone Cutler

REVIEW 3 major objections 4 minor 1 cited by

Inductive construction of path homology chains

T0 review · 3 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Inductive elements generate every path homology chain module over finite fields.

desk verdict A genuinely new inductive construction of path homology chains with a real open gap in the main generation theorem. read the letter →

arxiv 2411.09501 v1 pith:I3LINS77 submitted 2024-11-14 math.AT math.CO

Matthew Burfitt , Tyrone Cutler This is my paper · ORCID

classification math.ATmath.CO MSC 55U1005C2055N35

keywords pathhomologydirectedgraphschaincomplexesinductiveelementsfacemultihypergraphsbasisconstructionEulercharacteristicfinitefields

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 introduces a way to build elements of the path homology chain complex of any directed graph, one dimension at a time, by extending elements from the two previous dimensions. The construction is governed by labeled multihypergraphs the authors call face multihypergraphs, whose hyperedges record how boundary pieces cancel. The paper's central claim is that these inductive elements generate every path chain module over any finite field, in every dimension, and that in low dimensions they provide bases over the integers and over characteristic-zero fields. This matters because path homology has had no general chain-level basis description for arbitrary digraphs, and computing one was the bottleneck for algorithms. As a demonstration, the construction yields digraphs whose path Euler characteristic changes with the coefficient field, resolving an open question.

What carries the argument

Face multihypergraphs are labeled multihypergraphs whose vertices are path chains in dimension n and whose hyperedges record how boundary pieces of those chains cancel. An upper or lower extension appends a new vertex to every path in a chain, and a complete extension over a face multihypergraph packages exactly the cancellations needed for the extended element to lie in Ω_{n+1}. Strong connectedness, defined through mutation equivalence of face multihypergraphs, prevents disconnected redundancies and makes the extended elements suitable as generators and basis candidates.

What would settle it

Compute dim Ω_4(E_t;F_p) for a prime p that does not divide t using the accompanying implementation: the theorem predicts 0, so any nonzero dimension, or any path chain in Ω_n(G;F_p) not expressible from inductive elements, would refute the generation claim.

Watch

Extended reading notes

Core claim

On the paper's own terms: for every digraph G, the n-dimensional inductive elements—obtained by iterated strongly connected complete extensions over face multihypergraphs starting from the vertex basis—generate Ω_n(G;F_p) for each prime p and every n≥0, and contain bases in all dimensions over F_p. With integer or rational coefficients the same inductive elements contain bases of Ω_i(G;Z) and Ω_i(G;K) for i=0,1,2, and generate Ω_3(G;Z); over a characteristic-zero field K they contain a basis of Ω_3(G;K). The same machinery constructs digraphs E_t with dim Ω_4(E_t;K)=1 exactly when K=F_p for a prime p dividing t and 0 otherwise, while all other dimensions agree, so the path Euler characteristic can differ according to the coefficient field.

Load-bearing premise

The proof relies on the claim that the face multihypergraph assembled from a chosen basis element can be split into strongly connected pieces whose upper extensions sum back to the original element; that split is asserted rather than proved.

Editorial extensions

If this is right

Over every finite field F_p, a basis of Ω_n(G;F_p) can be chosen from n-dimensional inductive elements for every n and every digraph G.
In dimensions 0, 1, and 2, inductive elements coincide up to sign with the natural generators, so the existing low-dimensional basis descriptions are special cases of one construction.
The integral and characteristic-zero results give a basis-level description of Ω_3(G;R) with no restriction on double edges or multisquares.
The digraphs E_t make the path Euler characteristic depend on the coefficient field for odd primes, answering the open question.

Reading between the lines

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

A natural next test is to compute Ω_4(E_t;Z) directly: if it has Z/t-torsion, the p-dependent dimensions in Example 6.2 would be explained by prime-sized hyperedge matchings, a mechanism the paper does not state.
If the subdivision step in Theorem 5.1 can be made algorithmic, the same construction would give a direct chain-basis algorithm in all dimensions, bypassing the Hermite normal form reduction the paper mentions as a fallback.
The strong-connectedness condition is a purely combinatorial property of mutation equivalence classes; understanding its decision problem could turn inductive generation into a practical computational tool for arbitrary digraphs.

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.

Desk Editor's Note

read the letter

The face multihypergraph machinery is the real thing: it gives a concrete, inductive way to build path chains without the usual restrictions on multisquares or double edges, and it unifies several earlier basis descriptions. The low-dimensional coincidence with known generators is a good sanity check, and the construction is self-contained, with no sign of circularity. The authors also ship code, which makes the examples checkable independently.

The soft spot is exactly where the reader put it. In the proof of Theorem 5.1, after constructing the complete face multihypergraph from a basis element, the paper says a minimal subdivision into strongly connected face multihypergraphs provides inductive elements whose sum is the original element. That is asserted, not proved. It is load-bearing: completeness is a global condition, and splitting the hypergraph does not obviously preserve it. Without that step, the generation statement for Ω_n(G;Z_p) is unsupported. This is not a minor gap; it is the bridge between the construction and the main theorem. I do not think it is fatal—the claim sounds plausible and may well be fixable—but it needs a real argument.

The secondary issue is in Example 6.2, where uniqueness claims like “no other face multihypergraphs can be constructed up to sign” are stated without proof and underpin the dimension computation that resolves the Fu–Ivanov question. Those claims are more checkable, especially with the code, but they are still asserted rather than demonstrated.

Citation practice looks sound. The paper builds on Fu–Ivanov, Grigor’yan, and Asao appropriately, and the new examples are cross-validated against known algorithms. No data fitting or circular reasoning is apparent.

Who is this for? Anyone working in GLMY theory or path homology computation. The inductive elements are a promising tool even if the main theorem currently has a hole.

Recommendation: send to peer review, but the referee should be explicitly asked to prove or fix the subdivision step in Theorem 5.1 before acceptance. Conditional accept is the right posture.

Referee Report

3 major / 4 minor

Summary. The paper develops a method for building elements of the path chain complex Ω_n(G;R) of a digraph G from chains in dimensions n−1 and n−2 via 'upper' and 'lower extensions' over labeled multihypergraphs ('face multihypergraphs'). It defines 'inductive elements' by iterating strongly connected complete extensions and claims (Theorem 5.1, Corollaries 5.2 and 5.3) that these generate Ω_n(G;Z_p) for every prime p, generate Ω_3(G;Z), and contain bases in low dimensions. The paper also constructs two families of digraphs: M_t, whose boundary matrix with respect to an inductive basis contains an entry of multiplicity t, and E_t, whose 4-dimensional chain group has dimension 1 over Z_p for primes p dividing t and 0 over Q, resolving a question of Fu and Ivanov.

Significance. If the main theorem is correct, this is the first chain-level generating set for path homology with no restrictions on digraph structure, and it gives a new computational route and settles the Fu–Ivanov question on coefficient dependence of the path Euler characteristic. The paper is also accompanied by implementation code and states that the examples can be checked by direct computation, which is a strength. However, the proof of the central generation theorem contains an unproved subdivision step and a questionable algebraic identity, and the examples rely on unproven uniqueness assertions. The contribution is potentially significant but not yet fully established.

major comments (3)

[§5.1, proof of Theorem 5.1] The step 'a minimal subdivision of F^{n−1}_G(x_1,...,x_m) into strongly connected face multihypergraph provides (B_{n−1},B_{n−2})-inductive elements whose sum is x' is asserted without proof. Completeness (Definition 4.4) is a global property: for each u and each x^{u,k}_i, the required hyperedge may run between different pieces of the subdivision, and no argument shows that the pieces inherit completeness. Strong connectivity (Definitions 4.6 and 4.8) is also a property of the entire mutation class and is not automatically inherited by connected components. Since this is the only step linking the constructed face multihypergraph to the definition of inductive elements, the generation claim of Theorem 5.1, and therefore Corollaries 5.2 and 5.3, is not established.
[§5.1, after Eq. (5.2)] The displayed equality 0 = ∂M_{n−1,n−1}δh_{n,v}(x) = ∑_{v∈Vx}∑_k x^{v,i}_k is not justified. The right-hand side is the sum of the pieces in the decompositions of δh_{n−1,v}(x_i), whereas the left-hand side is the magnitude boundary of δh_{n,v}(x); the equality would require ∂M x_i = δh_{n−1,v}(x_i), which is not true in general. The subsequent conclusion that every x^{v,i}_k can be paired with a negative copy, as in Eq. (5.4), depends on the sum of these pieces being zero. Without a correct proof of this cancellation, the hyperedges of the constructed face multihypergraph may not exist as specified.
[§6.2, Example 6.2 (also §6.1, Example 6.1)] The assertions 'as no other face multihypergraphs can be constructed up to sign' and 'the only face multihypergraph up to sign and mutation that can be constructed on the elements E_i' are unproven. These uniqueness claims are load-bearing: they are what allow the computation of dim Ω_4(E_t;Z_t)=1 and dim=0 over other fields, which is the content of Theorem 1.4. The analogous uniqueness claim in Example 6.1 ('the inductive structure on I_t^4 ... is the only strongly connected H-complete face multigraph ... up to mutations') is used to conclude that I_t^4 is the unique generator and that its boundary contains an entry of multiplicity t. If these enumerations are not supplied, the examples should be described as computer-verified rather than as proved.

minor comments (4)

[§5.1, Theorem 5.1 statement] The statement uses '(B1, B2)-inductive elements' where the proof and context require '(B_{n−1}, B_{n−2})-inductive elements'; part (2) should likewise refer to (B_{n−1}, B_{n−2}).
[Abstract] The phrase 'from elements in the proceeding two dimensions' should read 'preceding two dimensions'.
[§6.2, Example 6.2] After defining E_i for i=1,...,t, the text refers to 'the elements Ei for i = 1, . . . ,2t'; this should be t.
[Definition 4.2 and §5.1] The phrase 'no sub-sequence ... sums to zero' is used without specifying whether proper subsequences are meant; this ambiguity matters for the pairing argument in Eq. (5.4).

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the chain-level construction is self-contained and the main theorem is not a restatement of its inputs.

full rationale

I walked the claimed derivation chain. Definition 5.1 and Theorem 5.1 do not reduce to each other by construction: an inductive element is defined as a particular upper/lower extension over a strongly connected complete face multihypergraph, while Theorem 5.1 starts with an arbitrary basis element x, decomposes the face maps δh_{n,v}(x) using independently given bases of Ω_{n-1} and Ω_{n-2}, and then asserts that the resulting face multihypergraph can be subdivided into strongly connected complete pieces. That subdivision is a missing proof, not a circular reduction: the pieces are not constructed so that the conclusion holds automatically. Proposition 4.1 is a sufficiency condition and is not used to define Ω_n. The background results cited (Asao's Lemma 2.1, GLMY Proposition 2.2, Grigor'yan's basis work, and Fu-Ivanov's no-multisquare basis) are external support, not premises containing the target theorem. The only self-reference is the accompanying code [2], which is used for independent verification of examples and is explicitly said to be checkable without the theory developed in the paper. The unproved subdivision assertion in Section 5.1 and the unproved uniqueness assertions in Example 6.2 are correctness risks, not circularity, and under the pass rules they do not raise the circularity score.

Assumptions & free parameters 0 free parameters · 4 assumptions · 3 invented entities

The central claim rests on the new combinatorial framework of face multihypergraphs, upper and lower extensions, and inductive elements. These are defined within the paper without fitted parameters. Background results from the path homology literature (Asao, GLMY) are used as axioms. No independent physical or computational entities are postulated; the provided code is an implementation, not an entity.

assumptions (4)

domain assumption The path chain complex (Ω_*(G;R), ∂^P_*) is defined as the largest submodule of allowed paths on which the path differential is a differential.
This is the standard definition from Grigor'yan et al. [16], used throughout Section 2.3.
domain assumption Lemma 2.1 (Asao): Ω_n(G;R) is isomorphic to the diagonal magnitude homology H^M_{n,n}(G;R), and membership in Ω_n is equivalent to vanishing of all ∂^M_{n,n,i} for i=1,...,n-1.
Imported from Asao [1]; used in Lemma 3.1 and Proposition 4.1 to characterize path chains.
domain assumption Proposition 2.2 (GLMY): double edges, directed triangles, and directed squares generate Ω_2(G;R), and bases are obtained by choosing bases of directed squares within each multisquare.
Imported from Grigor'yan et al. [16,17,20]; used in Corollary 5.3 to identify dimension-2 inductive elements with known generators.
standard math Axiom of choice is required to choose bases of Ω_{n-1}(G;R) and Ω_{n-2}(G;R) as subsets of infinite spanning sets when G is not finite.
Explicitly noted in Section 5.1 before Theorem 5.1.

invented entities (3)

Face multihypergraphs (including face multigraphs)
purpose: Labeled multihypergraphs whose hyperedges record cancellation patterns among boundary pieces, used to certify that an upper or lower extension lands in the path chain module.
New combinatorial objects defined in Definitions 4.2 and 4.3; their existence and completeness properties are established internally by Proposition 4.1, with no external confirmation yet.
Upper and lower extensions [x]_v
purpose: Operations that build an (n+1)-chain from an n-chain by appending a vertex at the head or tail; central to the inductive construction.
Definition 4.1; validity of extensions over complete face multihypergraphs is shown in Proposition 4.1.
Inductive elements
purpose: Elements of Ω_n(G;R) obtained as strongly connected complete extensions over face multihypergraphs from elements in the previous two dimensions; they are proved to generate Ω_n over finite fields.
Defined in Section 5.2; the generation theorem is Theorem 5.1/Corollary 5.2, and the accompanying code can check dimensions in specific examples, but the concept itself is an internal definition.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Inductive construction of path homology chains." pith.science (2026). https://pith.science/paper/I3LINS77

@misc{pith2026241109501,
  author       = {Pith},
  title        = {Pith review of: Inductive construction of path homology chains},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/I3LINS77}},
  note         = {Machine review of arXiv:2411.09501}
}

abstract

Path homology plays a central role in digraph topology and GLMY theory more general. Unfortunately, the computation of the path homology of a digraph $G$ is a two-step process, and until now no complete description of even the underlying chain complex has appeared in the literature. In this paper we introduce an inductive method of constructing elements of the path homology chain modules $\Omega_n(G;R)$ from elements in the proceeding two dimensions. This proceeds via the formation of what we call upper and lower \emph{extensions}, that are parametrised by certain labeled multihypergraphs which we introduce and call \emph{face multihypergraphs}. When the coefficient ring $R$ is a finite field the inductive elements we construct generate $\Omega_*(G;R)$. With integral or rational coefficients, the inductive elements generate at least $\Omega_i(G;R)$ for $i=0,1,2,3$. Since in low dimensions the inductive elements extended over labeled multigraphs coincide with naturally occurring generating sets up to sign, they are excellent candidates to reduce to a basis. Inductive elements provide a new concrete structure on the path chain complex that can be directly applied to understand path homology, under no restriction on the digraph $G$. We employ inductive elements to construct a sequence of digraphs whose path Euler characteristic can differ arbitrarily depending on the choice of field coefficients. In particular, answering an open question posed by Fu and Ivanov.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Inductive construction of path homology chains and the structure of $\Omega_3(G;R)$
math.AT 2026-07 conditional novelty 7.0 of 10

Path homology chains of any digraph are generated by inductive extensions over face multigraphs: completely in characteristic 2 and in dimensions 0–3 in characteristic 0.

Reference graph

Works this paper leans on

26 extracted references · 23 canonical work pages · cited by 1 Pith paper

[1]

Asao, Magnitude homology and path homology , Bull

Y. Asao, Magnitude homology and path homology , Bull. Lond. Math. Soc. 55 (2023), no. 1, 375–398

work page 2023
[2]

Burﬁtt, Path homology with ﬁeld coeﬃcients , November 2024, https://github.com/MatthewBurfitt/Path-homology-with-field-coefficients.git

M. Burﬁtt, Path homology with ﬁeld coeﬃcients , November 2024, https://github.com/MatthewBurfitt/Path-homology-with-field-coefficients.git

work page 2024
[3]

Carlsson, Topology and data , Bull

G. Carlsson, Topology and data , Bull. Amer. Math. Soc. 46 (2009), no. 2, 255–308

work page 2009
[4]

Carranza et al., Python script for computing path homology of digraphs , September 2022, https://github.com/sheaves/path_homology

D. Carranza et al., Python script for computing path homology of digraphs , September 2022, https://github.com/sheaves/path_homology. [5] , Coﬁbration category of digraphs for path homology , Algebr. Comb. 7 (2024), no. 2, 475–514

work page 2024
[6]

Chowdhury, T

S. Chowdhury, T. Gebhart, S. Huntsman, and M. Yutin, Path homologies of deep feed- forward networks , ICMLA, 2019, pp. 1077–1082

work page 2019
[7]

Chowdhury, S

S. Chowdhury, S. Huntsman, and M. Yutin, Path homologies of motifs and temporal network representations, Appl. Netw. Sci. 7 (2022), no. 4

work page 2022
[8]

Chowdhury and F

S. Chowdhury and F. M´ emoli, Persistent path homology of directed networks , SODA’ 18, SIAM, 2018, p. 1152–1169. 36

work page 2018
[9]

T. Dey, T. Li, and Y. Wang, An eﬃcient algorithm for 1-dimensional (persistent) path homology, Discrete Comput. Geom. 68 (2022), 1102–1132

work page 2022

Show all 26 references

[10]

Dimakis and F

A. Dimakis and F. M¨ uller-Hoissen, Discrete diﬀerential calculus: Graphs, topologies, and gauge theory , J. Math. Phys. 35 (1994), no. 12, 6703–6735

1994
[11]

Edelsbrunner and D

H. Edelsbrunner and D. Morozov, Persistent homology: theory and practice , ECM (2014), 31–50

2014
[12]

Fu and S

X. Fu and S. Ivanov, Path homology of digraphs without multisquares and its comp arison with homology of spaces , (2024), arXiv:2407.17001

2024 arXiv
[13]

Grbi´ c, J

J. Grbi´ c, J. Wu, K. Xia, and G. Wei, Aspects of topological approaches for data science , FoDS 4 (2022), no. 2, 165–216

2022
[14]

Grigor’yan, R

A. Grigor’yan, R. Jimenez, Y. Muranov, and S.-T. Yau, On the path homology theory of digraphs and eilenberg–steenrod axioms , Homology Homotopy Appl. 20 (2018), no. 2

2018
[15]

Grigor’yan, Y

A. Grigor’yan, Y. Lin, Y. Muranov, and R. Jimenez, Homology of digraphs, Math. Notes 109 (2021), 712–726

2021
[16]

Grigor’yan, Y

A. Grigor’yan, Y. Lin, Y. Muranov, and S.-T. Yau, Homologies of path complexes and digraphs, 2013, arXiv:1207.2834

2013 arXiv
[17]

, Homotopy theory for digraphs , Pure Appl. Math. Q. 10 (2014), no. 4, 919–674

2014
[18]

Grigor’yan and Y

A. Grigor’yan and Y. Muranov, On homology theories of cubical digraphs , Paciﬁc J. Math. 322 (2023), no. 1, 39–58

2023
[19]

Grigor’yan, Y

A. Grigor’yan, Y. Muranov, and S.-T. Yau, Graphs associated with simplicial complexes , Homology Homotopy Appl. 16 (2014), no. 1, 295–311

2014
[20]

Grigor’yan, Advances in path homology theory of digraphs , ICCM 10 (2022), no

A. Grigor’yan, Advances in path homology theory of digraphs , ICCM 10 (2022), no. 2, 61–124

2022
[21]

Grigor’yan, Y

A. Grigor’yan, Y. Muranov, and S.-T. Yau, Homologies of digraphs and k¨ unneth formu- las, Comm. Anal. Geom. 25 (2017), 969–1018

2017
[22]

Hepworth and E

R. Hepworth and E. Roﬀ, Bigraded path homology and the magnitude-path spectral se- quence, (2024), arXiv:2404.06689

2024 arXiv
[23]

Hepworth and S

R. Hepworth and S. Willerton, Categorifying the magnitude of a graph , Homology Ho- motopy Appl. 19 (2017), no. 2, 31–60

2017
[24]

Ivanov and F

S. Ivanov and F. Pavutnitskiy, Simplicial approach to path homology of quivers, marked categories, groups and algebras , J. Lond. Math. Soc. 109 (2024), no. 1, e12812

2024
[25]

Leinster and M

T. Leinster and M. Shulman, Magnitude homology of enriched categories and metric spaces, Algebr. Geom. Topol. 21 (2021), 2175–2221. 37

2021
[26]

Masulli and P

P. Masulli and P. Villa, The topology of the directed clique complex as a network inva ri- ant, SpringerPlus 5 (2016), no. 388

2016
[27]

Reimann et al., Cliques of neurons bound into cavities provide a missing lin k between structure and function , Front

M. Reimann et al., Cliques of neurons bound into cavities provide a missing lin k between structure and function , Front. Comput. Neurosci. 11 (2017). 38

2017

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Asao, Magnitude homology and path homology , Bull

Y. Asao, Magnitude homology and path homology , Bull. Lond. Math. Soc. 55 (2023), no. 1, 375–398

work page 2023

[2] [2]

Burﬁtt, Path homology with ﬁeld coeﬃcients , November 2024, https://github.com/MatthewBurfitt/Path-homology-with-field-coefficients.git

M. Burﬁtt, Path homology with ﬁeld coeﬃcients , November 2024, https://github.com/MatthewBurfitt/Path-homology-with-field-coefficients.git

work page 2024

[3] [3]

Carlsson, Topology and data , Bull

G. Carlsson, Topology and data , Bull. Amer. Math. Soc. 46 (2009), no. 2, 255–308

work page 2009

[4] [4]

Carranza et al., Python script for computing path homology of digraphs , September 2022, https://github.com/sheaves/path_homology

D. Carranza et al., Python script for computing path homology of digraphs , September 2022, https://github.com/sheaves/path_homology. [5] , Coﬁbration category of digraphs for path homology , Algebr. Comb. 7 (2024), no. 2, 475–514

work page 2024

[5] [6]

Chowdhury, T

S. Chowdhury, T. Gebhart, S. Huntsman, and M. Yutin, Path homologies of deep feed- forward networks , ICMLA, 2019, pp. 1077–1082

work page 2019

[6] [7]

Chowdhury, S

S. Chowdhury, S. Huntsman, and M. Yutin, Path homologies of motifs and temporal network representations, Appl. Netw. Sci. 7 (2022), no. 4

work page 2022

[7] [8]

Chowdhury and F

S. Chowdhury and F. M´ emoli, Persistent path homology of directed networks , SODA’ 18, SIAM, 2018, p. 1152–1169. 36

work page 2018

[8] [9]

T. Dey, T. Li, and Y. Wang, An eﬃcient algorithm for 1-dimensional (persistent) path homology, Discrete Comput. Geom. 68 (2022), 1102–1132

work page 2022