arxiv: 2604.17775 · v1 · submitted 2026-04-20 · 🧮 math.CO · math.GT

Recognition: unknown

The Cactus Criterion: When Nonlinear Hodge Theory Reduces to Linear on Graphs

Sebastian Pardo-Guerra , Anil Thapa , Jonathan Washburn

Authors on Pith 1 claimed

Jonathan Washburn ORCID verified

Pith reviewed 2026-05-10 04:49 UTC · model grok-4.3

classification 🧮 math.CO math.GT

keywords cactus graphsnonlinear Hodge theoryHodge projectoredge potentialscoclosed equationsgraph selectorsnonlinear minimization

0 comments

The pith

Nonlinear Hodge theory on graphs reduces to the linear Hodge projector exactly when the graph is a cactus.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper defines a nonlinear selector Picc on the space of 1-cochains of a finite connected simple graph by minimizing a sum of even, strictly convex edge potentials over each affine class. This selector satisfies a nonlinear coclosed condition and is shown to be real analytic. The central result states that Picc coincides with the ordinary linear Hodge projector on the entire space if and only if the graph is a cactus. A reader would care because this supplies an exact combinatorial condition under which all nonlinear corrections vanish globally, turning a nonlinear variational problem back into ordinary linear Hodge theory.

Core claim

For every admissible edge potential—even, C², strictly convex, and non-quadratic—the associated nonlinear selector coincides with the ordinary Hodge projector ΠH on all of C¹(G) if and only if G is a cactus graph.

What carries the argument

The nonlinear selector Picc obtained by minimizing ∑ψ(z_e) over each affine class ω + dC⁰(G), which satisfies the nonlinear coclosed equation δ sinh z = 0 and whose image is arsinh(ker δ).

If this is right

Picc is real analytic with image equal to arsinh(ker δ).
Picc agrees with ΠH to first order at the origin, with the first nonlinear correction being cubic.
The two-triangle graph is the smallest connected simple graph on which the selectors differ for some admissible potential.
A self-concordant Newton method computes Picc efficiently on any fixed graph.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

On cactus graphs one may safely replace the nonlinear problem by its linearization for any admissible potential without losing exactness.
The criterion isolates the combinatorial feature (cycles sharing edges) that forces nonlinear corrections to survive.
Similar reductions may hold for other strictly convex potentials or for higher-degree forms if the same cycle-sharing obstruction is absent.

Load-bearing premise

The edge potentials must belong to the class of even, C², strictly convex, non-quadratic functions, and the graph must be finite, connected, and simple.

What would settle it

Compute Picc and ΠH explicitly on the two-triangle graph for the potential cosh t − 1 and check whether they differ on some 1-cochain that is not harmonic.

Figures

Figures reproduced from arXiv: 2604.17775 by Anil Thapa, Jonathan Washburn, Sebastian Pardo-Guerra.

read the original abstract

Let $G$ be a finite connected simple graph with a chosen orientation of its edges. For the edge potential $\psi(t)=\cosh t-1,$ we minimize $\sum_{e\in E^\to}\psi(z_e)$ over each affine class $\omega+dC^0(G)\subset C^1(G)$. The minimizer is the unique representative satisfying the nonlinear coclosed equation $\delta\sinh z=0,$ and hence defines a nonlinear selector $\Picc:C^1(G)\to C^1(G).$ We show that $\Picc$ is real analytic, identify its image as $\imop \Picc=\Mcc=\operatorname{arsinh}(\ker\delta),$ and compute its differential as a weighted Hodge projector. In particular, $\Picc$ agrees with the ordinary Hodge projector $\PiH$ to first order at the origin, and the first nonlinear correction is cubic. Our main global theorem is a graph-theoretic criterion: for every admissible edge potential -- even, $C^2$, strictly convex, and non-quadratic -- the associated nonlinear selector coincides with $\PiH$ on all of $C^1(G)$ if and only if $G$ is a cactus graph. Finally, we work out the two-triangle graph, the smallest connected simple obstruction, and record a self-concordant Newton method for computing $\Picc$.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives a clean graph-theoretic criterion for when a nonlinear Hodge selector on graphs equals the linear projector everywhere: exactly on cactus graphs.

read the letter

The main thing to know is that this paper pins down a precise condition under which nonlinear Hodge theory on graphs reduces exactly to the linear case. For any even, C2, strictly convex, non-quadratic edge potential, the nonlinear selector that minimizes the sum of potentials in each affine class coincides with the ordinary Hodge projector on the whole space if and only if the graph is a cactus. That iff statement is the central new result, and it looks like it holds up from the outline and the stress-test reasoning on cycle spaces.

Referee Report

2 major / 3 minor

Summary. The paper defines, for an even C² strictly convex non-quadratic edge potential ψ, a nonlinear selector Picc: C¹(G) → C¹(G) obtained by minimizing ∑ ψ(z_e) over each affine coset ω + dC⁰(G) in the cochain space of a finite connected oriented graph G. It proves that Picc is real analytic, identifies its image as arsinh(ker δ), computes its differential at the origin as a weighted Hodge projector that agrees with the linear projector ΠH to first order (with cubic correction), and establishes the global theorem that Picc coincides with ΠH on all of C¹(G) for every admissible ψ if and only if G is a cactus graph. The two-triangle graph is treated explicitly as the minimal obstruction, and a self-concordant Newton method is supplied for computation.

Significance. If the central theorem holds, the work supplies a clean combinatorial criterion separating graphs on which a broad class of nonlinear variational problems reduce exactly to linear Hodge theory from those on which they do not. The argument exploits the edge-disjoint cycle basis of cactus graphs versus the shared-edge configurations that force the Cauchy functional equation f(a) + f(b) = f(a+b) on the nonlinearity f = ψ', thereby contradicting the non-quadratic hypothesis. The explicit two-triangle calculation and the self-concordant solver constitute concrete strengths that make the result immediately usable in discrete optimization and geometric graph theory.

major comments (2)

[§4, Theorem 4.2] §4, Theorem 4.2 (global iff statement): the 'only if' direction reduces non-cactus graphs to the two-triangle subgraph and invokes the Cauchy equation on the 2-dimensional cycle space spanned by the two cycle indicators; the manuscript must explicitly confirm that the global minimizer Picc, when restricted to the edges of this subgraph, satisfies the same first-order stationarity condition δ sinh z = 0 as the local nonlinear selector on the two-triangle graph itself.
[§3.3, Proposition 3.5] §3.3, Proposition 3.5 (image identification im Picc = arsinh(ker δ)): the argument that every element of arsinh(ker δ) is attained by some minimizer relies on strict convexity and coercivity of ψ; the proof should record the precise lower bound on the Hessian of the objective that guarantees uniqueness of the critical point inside each coset.

minor comments (3)

[Abstract] The abstract introduces the symbol Picc before it is defined; add a parenthetical gloss or move the definition sentence earlier in the introduction.
[§3.2, Eq. (12)] Notation: the weighted Hodge projector appearing in the differential computation (Eq. (12)) uses edge weights derived from ψ''(z_e); clarify whether these weights are evaluated at the origin or at the nonlinear solution z itself.
[Figure 2] Figure 2 (two-triangle graph): label the shared edge explicitly and indicate the two independent cycle indicators v1, v2 used in the proof of the obstruction.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading, positive evaluation, and constructive suggestions. The comments help clarify key steps in the proofs. We address each major comment below and will revise the manuscript accordingly.

read point-by-point responses

Referee: [§4, Theorem 4.2] §4, Theorem 4.2 (global iff statement): the 'only if' direction reduces non-cactus graphs to the two-triangle subgraph and invokes the Cauchy equation on the 2-dimensional cycle space spanned by the two cycle indicators; the manuscript must explicitly confirm that the global minimizer Picc, when restricted to the edges of this subgraph, satisfies the same first-order stationarity condition δ sinh z = 0 as the local nonlinear selector on the two-triangle graph itself.

Authors: We agree that an explicit confirmation strengthens the reduction argument. In the 'only if' direction of Theorem 4.2, the global stationarity condition δ ψ'(Picc(ω)) = 0 holds at every vertex of G. Restricting to the two-triangle subgraph H, any test function f supported on the vertices of H yields the pairing ⟨ψ'(z), df⟩ = 0, which is precisely the first-order stationarity condition for the local nonlinear selector on H. Thus the restriction of Picc(ω) to the edges of H satisfies δ_H ψ'(z|_H) = 0. We will insert a short paragraph in §4 making this restriction explicit, including the observation that the relevant cycle indicators remain linearly independent on H. revision: yes
Referee: [§3.3, Proposition 3.5] §3.3, Proposition 3.5 (image identification im Picc = arsinh(ker δ)): the argument that every element of arsinh(ker δ) is attained by some minimizer relies on strict convexity and coercivity of ψ; the proof should record the precise lower bound on the Hessian of the objective that guarantees uniqueness of the critical point inside each coset.

Authors: We accept this suggestion for added precision. The objective ∑_e ψ(z_e) is strictly convex on each affine coset ω + dC^0(G) because ψ is strictly convex. Its Hessian at any point z is the diagonal operator with entries ψ''(z_e). On the tangent space im d, the quadratic form satisfies ⟨H(z)v, v⟩ = ∑_e ψ''(z_e) v_e^2. Since ψ is C² and strictly convex, ψ''(t) > 0 for all t (or, more generally, the form is positive definite on nonzero coboundaries by connectedness of G). This yields a positive lower bound given by the smallest eigenvalue of the weighted graph Laplacian with weights ψ''(z_e) > 0, guaranteeing that any critical point is unique. We will record this bound explicitly in the proof of Proposition 3.5. revision: yes

Circularity Check

0 steps flagged

No significant circularity; self-contained iff proof via kernel preservation

full rationale

The central theorem is proved by direct analysis of when componentwise application of f=ψ' (for arbitrary admissible even C² strictly convex non-quadratic ψ) maps ker δ into itself. On cactus graphs the edge-disjoint cycle basis ensures invariance. On non-cactus graphs, intersecting cycles force the Cauchy equation f(a)+f(b)=f(a+b) on the relevant 2D subspace, implying f linear by monotonicity and hence ψ quadratic, a contradiction. This uses only the cycle-space structure of the graph and standard properties of convex functions; no equations reduce to fitted inputs, no load-bearing self-citations appear, and the nonlinear selector Picc is defined independently of the linear projector ΠH. The two-triangle case is exhibited explicitly as the minimal obstruction, confirming the argument is non-circular.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 2 invented entities

The paper relies on standard properties of finite graphs and convex functions; it introduces the selector Picc and the set Mcc as new objects defined by minimization.

axioms (2)

domain assumption Finite connected simple graph with chosen edge orientation
Stated at the beginning of the setup.
domain assumption Edge potential is even, C^2, strictly convex and non-quadratic
Required for the admissible class in the main theorem.

invented entities (2)

Nonlinear selector Picc no independent evidence
purpose: Maps each cochain to its nonlinear minimizer representative
Defined via minimization of sum psi(z_e) over affine classes.
Image Mcc = arsinh(ker delta) no independent evidence
purpose: Characterizes the range of the nonlinear selector
Identified in the analyticity and image theorem.

pith-pipeline@v0.9.0 · 5552 in / 1342 out tokens · 51327 ms · 2026-05-10T04:49:01.026326+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

22 extracted references · 1 canonical work pages

[1]

+ cosh( 25 8 −t ) + 2cosh( 5 4 − 2t) > 0, the functionF is strictly increasing, so it has at most one root. The sign changesF (5/4) < 0and F (3/2) > 0show that a root exists in (5/4,3/2), and numerically ϕ1 ≈1.35690123027,(65) which gives Πcc(ω)≈(1.76809877,1.76809877,−1.46380246,0.73190123,0.73190123).(66) The nonlinear representative differs from the li...
[2]

W. V. D. Hodge,The Theory and Applications of Harmonic Integrals(Cambridge University Press, Cambridge, 1941)

1941
[3]

Eckmann, Comment

B. Eckmann, Comment. Math. Helv.17, 240 (1944/45)

1944
[4]

Dodziuk and V

J. Dodziuk and V. K. Patodi, J. Indian Math. Soc. (N.S.)40, 1 (1976)

1976
[5]

Dodziuk, Amer

J. Dodziuk, Amer. J. Math.98, 79 (1976). 27

1976
[6]

Friedman, Algorithmica21, 331 (1998)

J. Friedman, Algorithmica21, 331 (1998)

1998
[7]

F. R. K. Chung,Spectral Graph Theory, CBMS Regional Conference Series in Mathematics No. 92 (American Mathematical Society, Providence, RI, 1997)

1997
[8]

Horak and J

D. Horak and J. Jost, Adv. Math.244, 303 (2013)

2013
[9]

L. J. Grady and J. R. Polimeni,Discrete Calculus: Applied Analysis on Graphs for Computational Science(Springer, London, 2010)

2010
[10]

Kozlov,Combinatorial Algebraic Topology, Algorithms and Computation in Mathematics, Vol

D. Kozlov,Combinatorial Algebraic Topology, Algorithms and Computation in Mathematics, Vol. 21 (Springer, Berlin, 2008)

2008
[11]

Lim, SIAM Rev.62, 685 (2020)

L.-H. Lim, SIAM Rev.62, 685 (2020)

2020
[12]

Jiang, L.-H

X. Jiang, L.-H. Lim, Y. Yao, and Y. Ye, Math. Program.127, 203 (2011)

2011
[13]

M. T. Schaub, A. R. Benson, P. Horn, G. Lippner, and A. Jadbabaie, SIAM Rev.62, 353 (2020)

2020
[14]

Washburn and M

J. Washburn and M. Zlatanović, Mathematics14, Article 935 (2026)

2026
[15]

Pardo-Guerra, M

S. Pardo-Guerra, M. Simons, A. Thapa, and J. Washburn, arXiv preprint (2026), recognition Physics Institute preprint, arXiv:2601.12194

work page arXiv 2026
[16]

Husimi, J

K. Husimi, J. Chem. Phys.18, 682 (1950)

1950
[17]

Harary and G

F. Harary and G. E. Uhlenbeck, Proc. Natl. Acad. Sci. USA39, 315 (1953)

1953
[18]

Diestel,Graph Theory, 5th ed., Graduate Texts in Mathematics, Vol

R. Diestel,Graph Theory, 5th ed., Graduate Texts in Mathematics, Vol. 173 (Springer, Berlin, 2017)

2017
[19]

Nesterov and A

Y. Nesterov and A. Nemirovski,Interior-Point Polynomial Algorithms in Convex Programming, SIAM Studies in Applied Mathematics, Vol. 13 (SIAM, Philadelphia, PA, 1994)

1994
[20]

Nesterov,Lectures on Convex Optimization, 2nd ed., Springer Optimization and Its Applications, Vol

Y. Nesterov,Lectures on Convex Optimization, 2nd ed., Springer Optimization and Its Applications, Vol. 137 (Springer, Cham, 2018)

2018
[21]

Boyd and L

S. Boyd and L. Vandenberghe,Convex Optimization(Cambridge University Press, Cambridge, 2004)

2004
[22]

S. G. Krantz and H. R. Parks,The Implicit Function Theorem: History, Theory, and Applications (Birkhäuser, Boston, MA, 2003). 28

2003