Pith. sign in

REVIEW 5 major objections 3 minor 46 references

The paper proves that under generic, non-extremal subgraph-density constraints, every entropy-maximizing multi-relational graphon is a step function with finitely many blocks.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 04:56 UTC pith:MG534H3O

load-bearing objection Strong local analysis and a genuinely new setting, but the keystone global topological lemma (Thm 5.23) rests on an unjustified convergence claim, so the main theorem is not proven as stated. the 5 major comments →

arxiv 2607.22383 v1 pith:MG534H3O submitted 2026-07-24 math.CO cs.ITmath.IT

Constrained Multi-Relational Graphons with Maximum Entropy

Juan Alvarado , Jan Ramon , Yuyi Wang This is my paper
classification math.CO cs.ITmath.IT MSC 05C8060F10
keywords maximum entropygraphonsmulti-relational networksstochastic block modelssubgraph densityconstrained optimizationtopological stabilitystep functions
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper sets out to prove a conjecture that emerged from numerical experiments: among all large random networks (or their continuum limits, graphons) that match prescribed subgraph-density statistics, the one that maximizes entropy is always a step function, meaning a stochastic block model with finitely many community types. The conjecture is resolved here for multi-relational graphons — networks with several edge types — under two generic conditions: the constraint densities must be analytically independent, and the target statistics must lie in the non-extremal interior of the feasible polytope. If the proof holds, an infinite-dimensional variational problem collapses to a finite-dimensional one: the most typical network is described by a finite matrix of block parameters, which explains why low-rank approximations work on real network data and makes probabilistic inference over relational knowledge bases computationally tractable in principle. The argument proceeds by differential geometry: study the feasible surfaces cut out by density constraints, show they do not develop new connected pieces as the allowed block count grows, and certify when a candidate solution is isolated in the full graphon space.

Core claim

Central claim: for multi-relational graphons with analytically independent subgraph-density constraints and non-extremal target densities, every entropy-maximizing graphon is a step function with finitely many blocks. The proof re-expresses entropy as an h-subgraph density, shows the feasible step-function sets form smooth manifolds for almost every statistic, and proves that enlarging the block count never creates a new connected component. Since step functions are dense in L1 and entropy is L1-continuous, a global minimum appears at some finite block size m1. A Hessian decoupling lemma separates finite-dimensional 'in-manifold' directions from infinite-dimensional 'off-manifold' directions

What carries the argument

h-subgraph density — subgraph density with an analytic function applied to each edge weight — which makes the entropy rate function just another density and lets one apply the same manifold arguments to it. The refinement operator, which splits one block into two without changing the underlying function, together with its coarsening inverse, gives a controlled way to compare step functions of different sizes. The feasible sets S^{(m,r)}(F,u,h) are smooth analytic manifolds for almost every statistic u, and the topological-stability theorem (Theorem 5.23) asserts that increasing m to m+1 introduces no new connected components, the keystone of the global argument. Finally, an exact decoupling

Load-bearing premise

The argument hinges on the claim that when the block count grows from m to m+1, the feasible surface never splits into a new connected piece; the proof of that claim is sketchy, resting on an exponential-map step that is not rigorously justified — if this fails, new global minima could appear at larger block counts and the step-function conclusion collapses.

What would settle it

For two relation types with edge and triangle constraints, take a non-extremal target density u inside the feasible polytope and compute entropy-minimizing graphons numerically for block sizes 2, 3, 4, starting from many initializations. If any minimizer is not block-constant, or if two distinct minimizers are connected by a continuum of equal-entropy solutions, the main theorem is false. Mathematically, one can also check the dimension count in the proof of the 'no new components' claim: the perturbation space at a hypothetical new minimal component is asserted to have dimension m−1; verifyin

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If the main theorem is correct, entropy-maximization for constrained networks becomes finite-dimensional: solve for the block matrix at size m1, not for an arbitrary function.
  • Probabilistic inference over relational data with target densities becomes in principle tractable: query probabilities are polynomials in the found block parameters.
  • The result grounds the empirical low-rank behavior of real complex networks in a maximum-entropy principle.
  • It supplies a rigorous instance of the structural argument that an ordered, block-structured phase cannot be reached from a disordered phase by a smooth thermodynamic path.
  • With all global minima isolated, the solution set is finite, and the optimal graphon is computable by finite-dimensional search.

Where Pith is reading between the lines

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

  • Editorial extension: the keystone topological claim (Theorem 5.23) is the weakest link; its proof sketch relies on an exponential-map step whose rigor is not fully established. If that step cannot be justified, the result still yields a strong local statement: any critical point passing the margin criterion is an isolated local minimum in the infinite-dimensional space.
  • Editorial extension: the non-extremal condition is doing essential work. Near the boundary of the feasible polytope, block counts can grow without bound, so the step-function conclusion should not be expected at extremal statistics.
  • Editorial extension: the h-subgraph-density framework may be reusable for other constrained variational problems on graphons, such as minimizing free energies with interaction terms.
  • Editorial extension: the theorem is numerically testable — for generic multi-relational constraints, optimize entropy with increasing block counts and verify the minimizer stays block-constant with the same partition; a continuum of non-step minimizers would falsify the isolation hypothesis.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 3 minor

Summary. The paper claims to resolve the RRS conjecture for multi-relational graphons in the non-extremal regime: under analytic independence of the subgraph-density constraints and for almost all feasible constraint values, entropy-maximizing graphons are step functions with finitely many blocks. The proof introduces h-subgraph densities, studies finite-dimensional step-function manifolds, derives first- and second-order derivative formulas, and proposes a topological stability theorem (Theorem 5.23) saying that enlarging the step size from m to m+1 creates no new connected components of the constraint manifold. From this it derives a no-new-global-minimum result (Corollary 5.3), a principle of optimization of density functions (Theorem 5.10), and finally the entropy specialization (Theorem 5.25). The paper also gives second-order sufficient conditions for a step-function critical point to be an isolated local minimum in L1 and cut-norm topologies.

Significance. If the proof were sound, this would be a major advance: it would settle the RRS conjecture for arbitrary multi-relational constraints, reduce an infinite-dimensional variational problem to finite-dimensional optimization, and provide checkable isolation criteria that handle non-diagonal Hessians such as those arising from triangle constraints. The h-subgraph-density formalism, the explicit second variation formula (Lemma 5.6), and the exact Hessian decoupling (Lemma 5.7) are useful technical pieces. However, the central topological theorem on which the global argument depends is not established, and the main theorem as stated is stronger than what the proof supports. The advertised resolution is therefore not supported by the manuscript.

major comments (5)
  1. [§5.6.1, Theorem 5.23] Claim (C) is the keystone of the proof and is false as stated. A point in the slice S^{(m+1,r)}_R(F,u,theta(pi,lambda,k),h) has block probabilities lambda pi_k and (1-lambda)pi_k, but its matrix entries in rows/columns k and k+1 are independent coordinates; they are not required to equal the corresponding size-m entries. For any constraint graph F, density terms using an edge incident to the block of mass lambda_n carry a factor O(lambda_n), so they vanish as lambda_n -> 0 regardless of the entry values. Thus x_n need not approach T^{-(m0,m)}(F,h) in the Euclidean topology, and the contradiction u in partial T^{(m0)}(F,h) does not follow. The claimed lambda* > 0 is unsupported, and the subsequent exponential-map step cannot be set up. The dimension assertion for O_x is also inconsistent: it should be m, not m-1. Since Corollary 5.3, Theorem 5.10, and Theorem 5.25 all invoke Theorem 5.23,
  2. [§5.4/§5.7, Theorem 5.10] The first assertion W^{(m1,r)}_R(F,u,f_d,h) subset W* is proved by saying 'By Corollary 5.3', but Corollary 5.3 assumes all global minima of f_d in S^{(m1,r)}_R(F,u,h) are isolated in W^{(m1,r)}_R. No such hypothesis appears in the statement of Theorem 5.10. Therefore the theorem as stated is not proved. The 'Moreover' part adds the isolation hypothesis but still depends on the unproved Theorem 5.23. The abstract's unconditional claim that entropy-maximizing solutions are step functions is accordingly stronger than what the stated hypotheses and proof support.
  3. [§5.5.4, Theorem 5.16] The proof of the L1 isolation theorem has a gap in the final estimate. After the Taylor expansion, the paper obtains f_d(W) - f_d(V*) >= (1/2)c0 ||eta^perp||_2^2 - C ||eta||_2^2. This expression does not imply a lower bound c'_0 ||eta||_2^2: the term -C ||eta_bar||_2^2 is uncontrolled in the displayed inequality. The sentence invoking Lemma 5.4 does not supply the missing finite-dimensional contribution. Thus isolation in L1 is not established, and Corollary 5.2, which feeds into the 'Moreover' part of Theorem 5.25, rests on this gap.
  4. [§5.8, Lemma 5.9] The lemma handles only a single boundary zero. The proof says 'Without loss of generality, take (A,pi) with exactly one zero entry A_{1,1;1}=0', but the preceding argument only shows that a neighborhood of a boundary point meets the interior; it does not show that a local minimum can be assumed to have a single boundary entry. A minimum may have several entries equal to 0 or 1, across different relations and block positions. The line-integral estimate isolates one unbounded term and ignores the others. This lemma is needed to ensure that the constraints 0 <= W <= 1 are inactive for u in Omega, so Theorem 5.25 is incomplete as written.
  5. [§5.6.2, Theorem 5.24 and Corollary 5.3] The 'deformation' argument is not justified. The proof asserts that by picking v' close to v one obtains a level set N' with N' cap S^{(m,r)}_R = empty and N' cap S^{(m+1,r)}_R nonempty, and that N' is a new connected component. Small perturbations of a level set can change its topology; the required transversality and component-stability are not proved. The second case, which supposes a disconnected global minimum at value c_{m1}, also assumes without proof that the level set at c > c_{m1} intersects the larger step space but not the smaller one. Since Corollary 5.3 is the bridge from Theorem 5.23 to Theorem 5.10, this is another load-bearing gap.
minor comments (3)
  1. [§1.1.1 and References] The text attributes the symmetric-bipodal phase result to 'Neeman, Radin, and Sadun [34]', but reference [34] is Radin and Sadun 2024; the cited Neeman-Radin-Sadun paper is [29]. The reference list should be corrected.
  2. [§3.2, §5.3, Remark 5.21] The entropy-rate function I_0 is defined with inconsistent normalizations and bases: in Section 3.2 it appears with a factor 1/2 and natural log, in Definition 5.9 with log_2, and in Remark 5.21 with natural log and no 1/2. The constants do not affect the qualitative arguments but should be fixed.
  3. [§4] The statement 'partial W^{(m+1,r)}_R = W^{(m,r)}_R' is not correct as an equality of Euclidean parameter spaces. The boundary of the (m+1)-parameter space includes zero-mass blocks, and the identification only holds at the level of represented functions, not as subsets of the parameter spaces used for the topology arguments.

Circularity Check

0 steps flagged

No circularity found; the central derivation is a genuine conditional reduction, with rigor gaps (not circular steps) in the keystone topological lemma.

full rationale

The paper's derivation chain does not contain any step where a claimed prediction is identical by construction to a fitted input, nor any load-bearing conclusion obtained by renaming an input or by a self-citation that itself asserts the target. The main result, Theorem 5.10 and Theorem 5.25, is a substantial reduction: it shows that a global minimizer found at step size m1 is already global in the full graphon space, using (i) the no-new-connected-components theorem (Theorem 5.23), (ii) Corollary 5.3 excluding better minima at larger m, (iii) density of step functions in L1 to rule out lower non-step minima, and (iv) an explicit isolation hypothesis to upgrade inclusion to equality. Each of these is a distinct mathematical claim. The h-subgraph-density formalism deliberately rewrites the entropy functional as t(E_s, ·, [I_0]) (Definition 5.9 and the discussion after it), but this is a definitional embedding, not a circular reduction: the proof does not assume entropy minimizers are step functions; it derives that under stated hypotheses. The only load-bearing citation to the authors' own prior work is Theorem 3.1, taken from [1] (Alvarado, Wang, Ramon), which supplies the large-deviation/maximum-entropy characterization for multi-relational graphons. That prior theorem is the setup of the problem, not the RRS conclusion; it does not assert step-function solutions, so the self-citation does not smuggle in the target. Under the review rules, this is independent support from published prior work rather than a circularity step. The abstract is stronger than Theorem 5.25: it omits the isolation hypothesis needed for equality, although the inclusion part of Theorem 5.25 holds without it. That is an overclaim, not a circularity. More importantly, the proof of Theorem 5.23 rests on claim (C), which asserts that x_n with lambda_n -> 0 approaches T^{-(m0,m)} in the default Euclidean topology; the paper states in Section 5.4 that the Euclidean topology is the default, and claim (C) is asserted rather than proved. Because duplicated-block entries are free parameters in that topology, the claim is questionable. This is a rigor/correctness gap in the keystone lemma, not a reduction of the conclusion to its inputs. For these reasons, the circularity score is 0.

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 0 invented entities

The central claim rests on the large-deviations framework of the authors' prior work [1], on an analytic-independence condition for the constraint functions (unverified for concrete families), and — in the unconditional parts — on a silently imported isolation hypothesis. No free parameters are fitted; the paper is a pure-proof preprint.

axioms (4)
  • domain assumption Theorem 8 of [1]: the conditionally typical multi-relational graphon is the minimizer of the entropy rate I over the constraint set (Theorem 3.1 in this paper).
    The whole framework depends on the large-deviations result of the authors' earlier paper [1] being correct. This is cited, not re-derived.
  • domain assumption Analytic independence of the constraint functions {t(F∪{Es}, ·, [h|[f0]])} (Definition 5.4, assumed in Theorems 5.10 and 5.25).
    It guarantees the marginal polytope has non-empty interior and that regular values are dense (Sard), which is needed for the manifold arguments. The paper does not verify this for any concrete constraint family.
  • ad hoc to paper Isolation of the global minima of f_d in the m1-step space (assumed in the 'Moreover' parts of Theorems 5.10 and 5.25, and in Corollary 5.3).
    The equality of argmin sets requires every global minimum to be isolated; this is a checkable sufficient condition (Theorem 5.14), not a general fact, and is silently imported into the proof of the first part of Theorem 5.10.
  • domain assumption Existence of a minimizer in S^{(r)}(F,u,h) (Section 5.4, paragraph on W*(r) ≠ ∅).
    Strict convexity of f0 such as the entropy rate I0 guarantees a unique minimizer for fixed multipliers, but existence on the full constraint set is assumed.

pith-pipeline@v1.3.0-alltime-deepseek · 49020 in / 22659 out tokens · 229750 ms · 2026-08-01T04:56:05.285342+00:00 · methodology

0 comments
read the original abstract

The principle of maximum entropy provides a fundamental framework for characterizing typical structures of large random networks subject to observable constraints. In their pioneering numerical experiments \cite{radin2014asymptotics}, Radin, Ren, and Sadun conjectured that entropy-maximizing graphons satisfying subgraph density constraints are stochastic block models a conjecture we term the RRS conjecture. While several special cases have been proven for single-relation graphs with specific constraint families, the general problem has remained open, particularly for multi-relational networks. We resolve the RRS conjecture for constrained multi-relational graphons in the non-extremal regime, proving that entropy-maximizing solutions are step functions with finitely many blocks under the condition the subgraph density constraints are analytically independent and for almost all feasible combinations of sufficient statistics. Our proof employs a differential geometric technique to study solutions of constrained optimization problems in function space via functions with a finite parametrization (step functions). The two cornerstones of this work are: the generalization of subgraph density notion to $h$-subgraph density and the proof that manifolds that define the constrained region for the solutions maintain topological stability without developing new connected components under refinement. Together, these enable proving that no new global optima emerge in higher-dimensional spaces.

Figures

Figures reproduced from arXiv: 2607.22383 by Jan Ramon, Juan Alvarado, Yuyi Wang.

Figure 1
Figure 1. Figure 1: A refinement operation transforms the representation of a [PITH_FULL_IMAGE:figures/full_fig_p015_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Dependencies of the main results to prove [PITH_FULL_IMAGE:figures/full_fig_p025_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Since (𝐴, 𝜋) ∈ 𝜕(𝑚,𝑟) ∩ 𝑆 (𝑚,𝑟) (, 𝑢, ℎ), 𝑡(, (𝐴, 𝜋), ℎ) ∈ 𝜕𝑇 (𝑚,𝑟) (, ℎ) 5.8 Adaptation of the principle of optimization of density functions to entropy functions We now specialize POD to prove the RRS conjecture itself. Let ℎ be a || × 𝑟 matrix of analytic functions, and let 𝑊 (𝑟) (, 𝑢, ℎ) = arg min 𝑊 ∈𝑆(𝑟) (,𝑢,ℎ) 𝐼(𝑊 ) where 𝑆 (𝑟) (, 𝑢, ℎ) = {𝑊 ∈ ̃(𝑟) ∶ 𝑡(, 𝑊 , ℎ) = 𝑢} and 𝑊 (𝑟,𝑚) (, 𝑢, ℎ) = … view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

46 extracted references · 3 linked inside Pith

  1. [1]

    Limits of multi-relational graphs

    Juan Alvarado, Yuyi Wang, and Jan Ramon. Limits of multi-relational graphs. Machine Learning, pages 1–40, 2022

    2022

  2. [2]

    Bose-einstein condensation in complex networks.Physical review letters, 86(24):5632, 2001

    Ginestra Bianconi and Albert-László Barabási. Bose-einstein condensation in complex networks.Physical review letters, 86(24):5632, 2001

    2001

  3. [3]

    A nonparametric view of network models and newman–girvan and other modularities.Proceedings of the National Academy of Sciences, 106(50):21068–21073, 2009

    Peter J Bickel and Aiyou Chen. A nonparametric view of network models and newman–girvan and other modularities.Proceedings of the National Academy of Sciences, 106(50):21068–21073, 2009

    2009

  4. [4]

    The structure and dynamics of multilayer networks.Physics Reports, 544(1):1–122, 2014

    Stefano Boccaletti, Ginestra Bianconi, Regino Criado, Charo I Del Genio, Jesús Gómez-Gardenes,MiguelRomance,IreneSendina-Nadal,ZhenWang,andMas- similiano Zanin. The structure and dynamics of multilayer networks.Physics Reports, 544(1):1–122, 2014

    2014

  5. [5]

    Pattern formation in random networks using graphons.SIAM Journal on Mathematical Analysis, 55(3):2150–2185, 2023

    Jason Bramburger and Matt Holzer. Pattern formation in random networks using graphons.SIAM Journal on Mathematical Analysis, 55(3):2150–2185, 2023

    2023

  6. [6]

    Higher- order graphon theory: Fluctuations, degeneracies, and inference.arXiv preprint arXiv:2404.13822, 2024

    Anirban Chatterjee, Bhaswar B Bhattacharya, and Sumit Mukherjee. Higher- order graphon theory: Fluctuations, degeneracies, and inference.arXiv preprint arXiv:2404.13822, 2024

    Pith/arXiv arXiv 2024

  7. [7]

    Inference for decorated graphs and applica- tion to multiplex networks.arXiv preprint arXiv:2408.12339, 2024

    Charles Dufour and François Caron. Inference for decorated graphs and applica- tion to multiplex networks.arXiv preprint arXiv:2408.12339, 2024

    Pith/arXiv arXiv 2024

  8. [8]

    Markovgraphs.JournaloftheamericanStatistical association, 81(395):832–842, 1986

    OveFrankandDavidStrauss. Markovgraphs.JournaloftheamericanStatistical association, 81(395):832–842, 1986

    1986

  9. [9]

    Rate-optimal graphon estimation.The Annals of Statistics, 43(6):2624–2652, 2015

    Chao Gao, Yu Lu, Harrison H Zhou, et al. Rate-optimal graphon estimation.The Annals of Statistics, 43(6):2624–2652, 2015

    2015

  10. [10]

    Elusive extremal graphs.Proceedings of the London Mathematical Society, 121(6):1685–1736, 2020

    Andrzej Grzesik, Daniel Král, and László Miklós Lovász. Elusive extremal graphs.Proceedings of the London Mathematical Society, 121(6):1685–1736, 2020

    2020

  11. [11]

    Stochastic blockmodels: First steps.Social networks, 5(2):109–137, 1983

    Paul W Holland, Kathryn Blackmond Laskey, and Samuel Leinhardt. Stochastic blockmodels: First steps.Social networks, 5(2):109–137, 1983

    1983

  12. [12]

    The economic conse- quences of social-network structure.Journal of Economic Literature, 55(1):49– 95, 2017

    Matthew O Jackson, Brian W Rogers, and Yves Zenou. The economic conse- quences of social-network structure.Journal of Economic Literature, 55(1):49– 95, 2017

    2017

  13. [13]

    Spinger, 2006

    Nocedal Jorge and J Wright Stephen.Numerical optimization. Spinger, 2006. 50

    2006

  14. [14]

    Springer Science & Business Media, 2008

    Jürgen Jost.Riemannian geometry and geometric analysis. Springer Science & Business Media, 2008

    2008

  15. [15]

    Bipodal structure in oversaturated random graphs.International Mathematics Research Notices, 2018(4):1009–1044, 2016

    Richard Kenyon, Charles Radin, Kui Ren, and Lorenzo Sadun. Bipodal structure in oversaturated random graphs.International Mathematics Research Notices, 2018(4):1009–1044, 2016

    2018

  16. [16]

    Multipodal struc- ture and phase transitions in large constrained graphs.Journal of Statistical Physics, 168(2):233–258, 2017

    Richard Kenyon, Charles Radin, Kui Ren, and Lorenzo Sadun. Multipodal struc- ture and phase transitions in large constrained graphs.Journal of Statistical Physics, 168(2):233–258, 2017

    2017

  17. [17]

    RichardKenyon,CharlesRadin,KuiRen,andLorenzoSadun.Thephasesoflarge networks with edge and triangle constraints.Journal of Physics A: Mathematical and Theoretical, 50(43):435001, 2017

    2017

  18. [18]

    Oracleinequalitiesfor network models and sparse graphon estimation.Annals of Statistics, 45(1):316– 354, 2017

    OlgaKlopp,AlexandreBTsybakov,andNicolasVerzelen. Oracleinequalitiesfor network models and sparse graphon estimation.Annals of Statistics, 45(1):316– 354, 2017

    2017

  19. [19]

    Springer, 2 edition, 2020

    Eric D Kolaczyk and Gábor Csárdi.Statistical Analysis of Network Data with R. Springer, 2 edition, 2020

    2020

  20. [20]

    Emerginglandscapeof molecularinteractionnetworks: Opportunities,challengesandprospects.Journal of Biosciences, 47:24, 2022

    ManjariKumar,ShubhamSrivastava,andAdityaMishra. Emerginglandscapeof molecularinteractionnetworks: Opportunities,challengesandprospects.Journal of Biosciences, 47:24, 2022

    2022

  21. [21]

    Elsevier, 2013

    LevDavidovichLandauandEvgeniiMikhailovichLifshitz.Courseoftheoretical physics. Elsevier, 2013

    2013

  22. [22]

    Springer, 2003

    JohnMLee.Smoothmanifolds.InIntroductiontoSmoothManifolds,pages1–29. Springer, 2003

    2003

  23. [23]

    Dbpedia–a large-scale, multilingual knowledge base extracted from wikipedia.Semantic Web, 6(2):167–195, 2015

    Jens Lehmann, Robert Isele, Max Jakob, Anja Jentzsch, Dimitris Kontokostas, Pablo N Mendes, Sebastian Hellmann, Mohamed Morsey, Patrick Van Kleef, Sören Auer, et al. Dbpedia–a large-scale, multilingual knowledge base extracted from wikipedia.Semantic Web, 6(2):167–195, 2015

    2015

  24. [24]

    Limits of dense graph sequences.Journal of Combinatorial Theory, Series B, 96(6):933–957, 2006

    László Lovász and Balázs Szegedy. Limits of dense graph sequences.Journal of Combinatorial Theory, Series B, 96(6):933–957, 2006

    2006

  25. [25]

    Finitely forcible graphons.Journal of Com- binatorial Theory, Series B, 101(5):269–301, 2011

    László Lovász and Balázs Szegedy. Finitely forcible graphons.Journal of Com- binatorial Theory, Series B, 101(5):269–301, 2011

    2011

  26. [26]

    EyalLubetzkyandYufeiZhao.Onreplicasymmetryoflargedeviationsinrandom graphs.Random Structures & Algorithms, 47(1):109–146, 2015

    2015

  27. [27]

    Yago3: Aknowl- edge base from multilingual wikipedias

    FarzanehMahdisoltani,JoannaBiega,andFabianMSuchanek. Yago3: Aknowl- edge base from multilingual wikipedias. InCIDR, 2015. 51

    2015

  28. [28]

    Disease networks

    JörgMenche,AmitabhSharma,MaksimKitsak,SusanDinaGhiassian,MarcVi- dal, Joseph Loscalzo, and Albert-László Barabási. Disease networks. uncover- ing disease-disease relationships through the incomplete interactome.Science, 347(6224):1257601, 2015

    2015

  29. [29]

    Existence of a symmetric bipo- dal phase in the edge-triangle model.Journal of Physics A: Mathematical and Theoretical, 57(9):095003, 2024

    Joe Neeman, Charles Radin, and Lorenzo Sadun. Existence of a symmetric bipo- dal phase in the edge-triangle model.Journal of Physics A: Mathematical and Theoretical, 57(9):095003, 2024

    2024

  30. [30]

    Graphongames: Astatisticalframework for network games and interventions.Econometrica, 91(1):191–225, 2023

    FrancescaPariseandAsumanOzdaglar. Graphongames: Astatisticalframework for network games and interventions.Econometrica, 91(1):191–225, 2023

    2023

  31. [31]

    An empirical bayes approach to stochastic block- modelsandgraphons: shrinkageestimationandmodelselection.PeerJComputer Science, 8:e1006, 2022

    Zhanhao Peng and Qing Zhou. An empirical bayes approach to stochastic block- modelsandgraphons: shrinkageestimationandmodelselection.PeerJComputer Science, 8:e1006, 2022

    2022

  32. [32]

    Dynamic network models and graphon estimation.The Annals of Statistics, 47(4):2378–2403, 2019

    M Pensky. Dynamic network models and graphon estimation.The Annals of Statistics, 47(4):2378–2403, 2019

    2019

  33. [33]

    CharlesRadin,KuiRen,andLorenzoSadun.Theasymptoticsoflargeconstrained graphs.Journal of Physics A: Mathematical and Theoretical, 47(17):175001, 2014

    2014

  34. [34]

    Emergence in graphs with near-extreme con- straints.arXiv preprint arXiv:2411.14556, 2024

    Charles Radin and Lorenzo Sadun. Emergence in graphs with near-extreme con- straints.arXiv preprint arXiv:2411.14556, 2024

    Pith/arXiv arXiv 2024

  35. [35]

    Markov logic networks.Machine learning, 62(1-2):107–136, 2006

    Matthew Richardson and Pedro Domingos. Markov logic networks.Machine learning, 62(1-2):107–136, 2006

    2006

  36. [36]

    One knowledge graph to rule them all? an- alyzing the differences between dbpedia, yago, wikidata & co

    Daniel Ringler and Heiko Paulheim. One knowledge graph to rule them all? an- alyzing the differences between dbpedia, yago, wikidata & co. InKI 2017: Ad- vances in Artificial Intelligence, pages 366–372. Springer, 2017

    2017

  37. [37]

    Protein-protein interaction networks (ppi) and complexdiseases.GastroenterologyandHepatologyfromBedtoBench,7(1):17, 2014

    NahidSafari-Alighiarloo,MostafaTaghizadeh,MostafaRezaei-Tavirani,Bahram Goliaei, and Ali Akbar Peyvandi. Protein-protein interaction networks (ppi) and complexdiseases.GastroenterologyandHepatologyfromBedtoBench,7(1):17, 2014

    2014

  38. [38]

    Modeling relational data with graph convolutional net- works

    Michael Schlichtkrull, Thomas N Kipf, Peter Bloem, Rianne Van Den Berg, Ivan Titov, and Max Welling. Modeling relational data with graph convolutional net- works. InEuropean Semantic Web Conference, pages 593–607. Springer, 2018

    2018

  39. [39]

    Network estimation via graphon with node features.IEEE Transactions on Network Science and Engi- neering, 7(3):2078–2089, 2020

    Yi Su, Raymond KW Wong, and Thomas CM Lee. Network estimation via graphon with node features.IEEE Transactions on Network Science and Engi- neering, 7(3):2078–2089, 2020

    2078

  40. [40]

    Yago: a core of semantic knowledge unifying wordnet and wikipedia

    Fabian M Suchanek, Gjergji Kasneci, and Gerhard Weikum. Yago: a core of semantic knowledge unifying wordnet and wikipedia. InProceedings of the 16th International Conference on World Wide Web, pages 697–706, 2007. 52

    2007

  41. [41]

    Multirelational organiza- tionoflarge-scalesocialnetworksinanonlineworld.ProceedingsoftheNational Academy of Sciences, 107(31):13636–13641, 2010

    Michael Szell, Renaud Lambiotte, and Stefan Thurner. Multirelational organiza- tionoflarge-scalesocialnetworksinanonlineworld.ProceedingsoftheNational Academy of Sciences, 107(31):13636–13641, 2010

    2010

  42. [42]

    Thelow-rankhypoth- esis of complex systems.Nature Physics, pages 1–9, 2024

    VincentThibeault,AntoineAllard,andPatrickDesrosiers. Thelow-rankhypoth- esis of complex systems.Nature Physics, pages 1–9, 2024

    2024

  43. [43]

    Iowa State University, 1968

    George Eugène Uhlenbeck.Fundamental Problems in Statistical Mechanics: A Lecture Series. Iowa State University, 1968

    1968

  44. [44]

    Random walk withrestartonmultiplexandheterogeneousbiologicalnetworks.Bioinformatics, 35(3):497–505, 2019

    AlbertoValdeolivas,LaurentTichit,ClaireNavarro,SophiePerrin,GaëlleOdelin, Nicolas Levy, Pierre Cau, Elisabeth Remy, and Anaïs Baudot. Random walk withrestartonmultiplexandheterogeneousbiologicalnetworks.Bioinformatics, 35(3):497–505, 2019

    2019

  45. [45]

    Wikidata: a free collaborative knowl- edgebase.Communications of the ACM, 57(10):78–85, 2014

    Denny Vrandečić and Markus Krötzsch. Wikidata: a free collaborative knowl- edgebase.Communications of the ACM, 57(10):78–85, 2014

    2014

  46. [46]

    Ratesofconvergenceofspectralmethodsforgraphonestimation

    JiamingXu. Ratesofconvergenceofspectralmethodsforgraphonestimation. In InternationalConferenceonMachineLearning,pages5433–5442.PMLR,2018. 53

    2018