Pith. sign in

REVIEW 3 major objections 6 minor 41 references

An $\alpha$-triangle eigenvector centrality of graphs

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read The paper introduces αTEC, a centrality that blends edge and triangle structure into a tensor eigenvector and gives every vertex of a connected graph a unique positive score.

desk verdict A correct but modest tensor-centrality extension whose empirical connectivity claim overreaches the reported experiment. read the letter →

arxiv 2506.07026 v1 pith:PMHAVAZW submitted 2025-06-08 cs.SI

classification cs.SI MSC 05C5005C8215A69
keywords centralitytensoreigenvectortrianglestructureedgePerron-FrobeniustheoremweaklyirreduciblenetworkconnectivityC.elegansmetabolic
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

Centrality measures rank vertices in networks, but most use either local counts or a single kind of global structure. This paper introduces α-triangle eigenvector centrality (αTEC), which builds one third-order nonnegative tensor from both edges and triangles and takes the unique positive eigenvector for the tensor's spectral radius as the scores. The single parameter α∈(0,1] shifts the balance from edge-driven to triangle-driven importance; larger α weights edge neighborhoods, smaller α weights triangle neighborhoods. Because connected graphs always produce a weakly irreducible tensor, every vertex receives a positive score, and the ranking responds smoothly to α. The paper reports experiments on synthetic and real networks supporting the claim that high-αTEC vertices and triangles occupy structurally critical positions and affect network connectivity.

What carries the argument

The load-bearing object is the α-triangle tensor A=αA_E+(1−α)A_△, a third-order nonnegative tensor. Its edge part has an entry 1 at (i,j,j) whenever {i,j} is an edge, and its triangle part has entry 1/2 at each ordered triple forming a triangle. The associated directed graph of this tensor contains an arc from i to j for every edge, so a connected graph makes the tensor weakly irreducible. The Perron-Frobenius theorem for weakly irreducible nonnegative tensors then supplies the unique positive spectral eigenvector that defines the centrality.

What would settle it

Take the C.elegans metabolic network, rank triangles by αTEC score, and build control triangles whose three vertices have the same degree counts as the top-ranked triangles but are not top-ranked. Delete the control vertices and count connected components. If degree-matched controls split the network into 6 or more components, the claimed connectivity advantage of αTEC triangles fails; if they split into fewer than 6, the claim survives.

Watch

Extended reading notes

Core claim

The central construction is the α-triangle tensor A=αA_E+(1−α)A_△ of a connected graph G, where A_E records edges and A_△ records triangles. Theorem 2.2 asserts that A is weakly irreducible for every α∈(0,1], since the associated directed graph contains a directed arc i→j for every edge {i,j} and G is connected; by the Perron-Frobenius theorem for nonnegative weakly irreducible tensors, the spectral radius ρ(A) has a unique positive eigenvector x up to scaling. That x is the αTEC score vector. In components, ρ(A)$x_i^{2}$=α∑_{ {i,j}∈E(G)}$x_j^{2}$+(1−α)∑_{ {i,j,k}∈V△}x_jx_k, so a vertex's score is a self-consistent blend of its edge-neighbors' squared scores and its triangle-neighbors' products. The paper also analyzes rankings as α varies and compares triangle deletion effects in the C.elegans metabolic network, where top αTEC triangles split the connected network into 6 components versus 3 or 2 for the comparison cycle ranking.

Load-bearing premise

The load-bearing premise is that deleting the three vertices of a high-scoring triangle is a fair comparison across triangles, not a proxy for deleting high-degree vertices or vertices in dense regions; the paper does not test this against degree-matched or random-triangle baselines.

Editorial extensions

If this is right

  • Every vertex of a connected graph receives a positive αTEC score, so the measure never zeros out vertices that lie outside triangles.
  • Setting α=1 recovers the classical eigenvector centrality ranking, making αTEC a continuous interpolation toward triangle-dominated rankings as α approaches 0.
  • Rank changes under decreasing α track the local triangle environment: vertices inside or adjacent to high-triangle regions rise, while vertices in edge-dense but triangle-poor regions fall.
  • Deleting the vertices of the top-ranked αTEC triangles in the C.elegans metabolic network splits the connected network into 6 components, compared with 3 or 2 for the Fiedler-vector cycle ranking used as the comparison baseline.

Reading between the lines

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

  • The same weak-irreducibility argument should carry over to any connected motif hypergraph built from edge pairs plus higher-order motifs, yielding a family of motif-tuned eigenvector centralities with positive scores on every vertex.
  • A natural control experiment would delete degree-matched random triangles rather than only the top-ranked ones; if those controls split the network into a comparable number of components, the connectivity result would be attributed to vertex deletion or local density rather than to triangle importance.
  • Because α is continuous, it can be calibrated to data, for example by maximizing agreement with an external importance ranking, turning αTEC from a descriptive index into a fitted centrality model.
  • For weighted or directed networks, the construction extends by replacing the 0/1 edge and triangle entries with observed weights, and the positive-eigenvector guarantee would survive as long as the directed support remains strongly connected.
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

3 major / 6 minor

Summary. The paper proposes an α-triangle eigenvector centrality (αTEC) for connected graphs. The centrality vector x is defined as the positive eigenvector associated with the spectral radius of the nonnegative tensor A = αA_E + (1−α)A_△, where A_E encodes edges and A_△ encodes triangles. The main theoretical result is Theorem 2.2, which states that for a connected graph and α ∈ (0,1], this tensor is weakly irreducible; combined with the Perron–Frobenius theorem for nonnegative tensors, this gives a unique positive eigenvector and hence nonzero centrality scores for every vertex. Numerical experiments on a small graph, Zachary's karate club, the Lusseau dolphin network, and the C. elegans metabolic network show how rankings vary with α. Section 3.3 defines a triangle importance score I△ and reports that deleting the vertices of top-ranked triangles under I△ splits the C. elegans network into more connected components than deleting top-ranked triangles under the Fiedler-vector-based index Ici of Jiang et al. The abstract claims that the experiments prove that vertices with higher αTEC rankings have greater impact on network connectivity.

Significance. If the theoretical claim is correct, αTEC provides a flexible, global centrality measure with a tunable parameter α that interpolates between edge-based and triangle-based influence, and it guarantees strictly positive scores for all vertices in connected graphs. The proof of weak irreducibility in Theorem 2.2 is simple and correct, and the Perron–Frobenius argument is standard. The paper also explicitly gives the componentwise recurrence (2.2), which is useful for understanding the mechanism. However, the experimental support for the connectivity claim in Section 3.3 is not adequate: the intervention is triangle-level vertex deletion, no value of α is stated for the scores used in Table 3, and no controls or statistical tests are provided. Thus the stronger empirical claim in the abstract and the conclusion is not established. The paper would be more valuable if it added a rigorous vertex-level connectivity experiment with degree-matched or randomized baselines.

major comments (3)
  1. [Abstract and Section 3.3] The claim that 'we experimentally prove that vertices with higher αTEC rankings have a greater impact on network connectivity' is not supported by the experiments. The experiment in Section 3.3 deletes all three vertices of the top-ranked triangles according to the ad hoc score I△, not individual vertices ranked by their αTEC scores. This is a triangle-level intervention, and it does not compare vertices by their αTEC rank. Furthermore, no value of α is given for the scores x used to compute I△ in Table 3, and no degree-matched triangles, random-triangle null distribution, or isolated vertex-deletion series are provided. The observed difference in connected components (6 vs. 3 or 2) could be due to the selected triangles containing higher-degree or otherwise more central vertices, rather than demonstrating a property of αTEC.
  2. [Section 3.3, Definition 3.1] Definition 3.1 is not derived from the αTEC centrality in a principled way. The score I△i = Σ x_p + x_q + x_r is simply the sum of αTEC scores of the three vertices in the triangle, but no justification is given for why this triangle-level aggregation should be considered 'the importance score' of a triangle or why it should be identified with αTEC. The notation 'Σ_{p,q,r}' is also ambiguous because it does not specify whether the sum runs over the three vertices of △i with a factor, or over all permutations. As written, the formula appears to have three separate terms x_p, x_q, x_r, which duplicates the sum. The manuscript needs a clear definition and a justification before this quantity is used to support the connectivity claim.
  3. [Conclusion] The Conclusion states: 'Furthermore, we provide conditions under which all vertices in a regular graph share identical αTEC scores.' No such conditions or theorem appear anywhere in the body of the paper. This claim should either be substantiated with a new theorem and proof, or removed from the conclusion.
minor comments (6)
  1. [Section 3.3, Table 3] Table 3 has several formatting and typographical issues: 'T riangle' should be 'Triangle', and the entries list triangles as 'c1 = [56, 153, 217]' etc., but the text refers to deleting 'c1, c2 and c3' and '△1, △2 and △3' without explaining the correspondence between the table rows and the deletions. The paragraph would benefit from a clearer description of what exactly was deleted.
  2. [Figure cross-references] The text in Section 3.2.1 says 'Figure 4 illustrates all triangles of ZKC network', but the figure is labeled as Figure 3. Later, Section 3.2.2 refers to 'Figure 4' for the degree and triangle distributions of the dolphin network, which is correct. Please renumber or fix the cross-references consistently.
  3. [Section 3.1, Table 1] The phrase 'an simple example' should be 'a simple example'. Also, the caption of Figure 1 says 'vertices’s αTEC scores', which should be 'vertices’ αTEC scores' or 'vertex αTEC scores'.
  4. [Section 3.3] The sentence 'The critical triangles identified by αTEC play a pivotal role in enhancing network Connectivity.' uses an uppercase 'C' in 'Connectivity' and is not supported by the experiment as described. The wording should be softened unless the experiment is substantially improved.
  5. [Equation (2.2)] In Equation (2.2), the notation is mostly clear, but the manuscript would benefit from explicitly stating that the sums are over edges {i,j} ∈ E(G) and triangles {i,j,k} ∈ V△, respectively, and that the equality holds componentwise for each i. This is implicit but should be stated for readability.
  6. [Reproducibility] The numerical experiments do not describe the algorithm used to compute the tensor eigenvector (e.g., the power method), the convergence tolerance, or the number of iterations. Adding these details would improve reproducibility.

Circularity Check

1 steps flagged · score 2.0 of 10

Minor definitional tautology in the α-weighting claim; the mathematical core is otherwise self-contained and no fitted-input or load-bearing self-citation circularity is present.

  1. self definitional [Section 2, Definition 2.1 and Eq. (2.2)]
    "Definition 2.1. For a connected graph G, tensor A = αAE + (1 − α)A△ is called the α-triangle tensor of G ... As evident from Equ. (2.2), the αTEC of vertex i is jointly influenced by vertices in both its edge neighborhood and triangle neighborhood, with the relative weights of these influences determined by α."

    The claimed α-dependent weighting of edge versus triangle influence is built into the definition of the centrality rather than independently derived or tested. Equation (2.2) is exactly the eigenvalue equation for the tensor A defined in Definition 2.1, namely ρ(A)x_i^2 = α times the edge-neighborhood sum plus (1−α) times the triangle-neighborhood sum, and the αTEC scores are by definition the eigenvector of A. Thus the statement that larger α strengthens edge influence and smaller α strengthens triangle influence is a restatement of the construction, not a separate prediction. This is a mild definitional tautology, not a fitted-parameter circularity, and it does not affect the uniqueness theorem.

full rationale

The load-bearing mathematical claim is self-contained: Theorem 2.2 proves weak irreducibility of the α-triangle tensor directly from connectedness of G, and the existence/uniqueness of a positive Perron vector is imported from the standard external Perron-Frobenius theorem for nonnegative weakly irreducible tensors [22]. No parameter is fitted to any data subset, and the centrality scores are computed by solving the eigenproblem, so there is no fitted input renamed as a prediction. The paper's self-citations ([38], [40], [41]) are contextual and are not load-bearing for the uniqueness or ranking results. The single mild circularity is the definitional α-weighting described in the step above. The Section 3.3 connectivity experiment has evidentiary gaps that should be weighed separately: Definition 3.1 defines a triangle-level score I△i, Table 3 deletes all three vertices of the top-ranked triangles rather than testing vertex-level αTEC rankings, the α used for the scores x is not stated, and no degree-matched or random-triangle baseline is provided, so the abstract's connectivity claim is not directly supported. The Conclusion also promises regular-graph conditions that never appear in the body. These are correctness/completeness concerns, not circularity.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The only free knob is α, which is a user input rather than a fitted value. The mathematics rests on standard tensor Perron-Frobenius theory and on the connectedness assumption. No new physical entities are introduced; the triangle importance score I_△ in Definition 3.1 is just the sum of αTEC scores and is not an independent construct.

free parameters (1)
  • α (alpha) = user-chosen within (0,1]
    Controls the weighted combination of edge and triangle tensors in Definition 2.1; not fitted, but all ranking behavior depends on it and no principled choice rule is given.
assumptions (3)
  • standard math Perron-Frobenius theorem for nonnegative weakly irreducible tensors guarantees a unique positive eigenvector for the spectral radius.
    Invoked in Section 2 after Theorem 2.2 to define centrality scores; this is an established theorem (Friedland, Gaubert, Han 2013).
  • domain assumption The graph G is connected.
    Definition 2.3 and Theorem 2.2 apply only to connected simple undirected graphs; disconnected graphs are outside the scope.
  • domain assumption Tensor eigenvector equation Ax^2 = λx^[2] is the appropriate nonlinear generalization of eigenvector centrality.
    The paper adopts the standard tensor eigenvector formulation without arguing that this semantics, rather than other centrality definitions, is the right one.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An $\alpha$-triangle eigenvector centrality of graphs." pith.science (2026). https://pith.science/paper/PMHAVAZW

@misc{pith2026250607026,
  author       = {Pith},
  title        = {Pith review of: An $\alpha$-triangle eigenvector centrality of graphs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PMHAVAZW}},
  note         = {Machine review of arXiv:2506.07026}
}
abstract

Centrality represents a fundamental research field in complex network analysis, where centrality measures identify important vertices within networks. Over the years, researchers have developed diverse centrality measures from varied perspectives. This paper proposes an $\alpha$-triangle eigenvector centrality ($\alpha$TEC), which is a global centrality measure based on both edge and triangle structures. It can dynamically adjust the influence of edges and triangles through a parameter $\alpha$ ($\alpha \in (0,1]$). The centrality scores for vertices are defined as the eigenvector corresponding to the spectral radius of a nonnegative tensor. By the Perron-Frobenius theorem, $\alpha$TEC guarantees unique positive centrality scores for all vertices in connected graphs. Numerical experiments on synthetic and real world networks demonstrate that $\alpha$TEC effectively identifies the vertex's structural positioning within graphs. As $\alpha$ increases (decreases), the centrality rankings reflect a stronger (weaker) contribution from edge structure and a weaker (stronger) contribution from triangle structure. Furthermore, we experimentally prove that vertices with higher $\alpha$TEC rankings have a greater impact on network connectivity.

Figures

Figures reproduced from arXiv: 2506.07026 by the authors.

Figure 1
Figure 1. (a) is a graph with 14 vertices and (b) is vertices’s [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Three networks and vertices’s αTEC scores distribution graphs under different values of α. 8 [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. All triangles of Zachary’s karate club network. [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: The distributions of D(i), T(i), and NT(i) in the Lusseau’s Dolphin network. 10 [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: The distribution of vertices’s centrality scores in the C. elegans metabolic [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Red(Blue) triangles are important triangles in the C.elegans metabolic [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages

  1. [1]

    Error and attack tolerance of complex networks

    R´ eka Albert, Hawoong Jeong, and Albert L´ aszl´ o Barab´ asi. Error and attack tolerance of complex networks. Nature, 406(6794):378–382, 2000. 14

  2. [2]

    A framework for second-order eigenvector centralities and clustering coefficients

    Francesca Arrigo, Desmond J Higham, and Francesco Tudisco. A framework for second-order eigenvector centralities and clustering coefficients. Proceedings of the Royal Society A , 476(2236):20190724, 2020

  3. [3]

    Betweenness centrality in large complex networks

    Marc Barthelemy. Betweenness centrality in large complex networks. The European physical journal B , 38(2):163–168, 2004

  4. [4]

    Three hypergraph eigenvector centralities

    Austin R Benson. Three hypergraph eigenvector centralities. SIAM Journal on Mathematics of Data Science , 1(2):293–312, 2019

  5. [5]

    A new centrality measure based on the negative and positive effects of clustering coefficient for identifying influential spreaders in complex networks

    Kamal Berahmand, Asgarali Bouyer, and Negin Samadi. A new centrality measure based on the negative and positive effects of clustering coefficient for identifying influential spreaders in complex networks. Chaos, Solitons & Frac- tals, 110:41–54, 2018

  6. [6]

    Factoring and weighting approaches to status scores and clique identification

    Phillip Bonacich. Factoring and weighting approaches to status scores and clique identification. Journal of mathematical sociology , 2(1):113–120, 1972

  7. [7]

    A family of centrality measures for graph data based on subgraphs

    Sebasti´ an Bugedo, Cristian Riveros, and Jorge Salas. A family of centrality measures for graph data based on subgraphs. ACM Transactions on Database Systems, 49(10):1–45, 2024

  8. [8]

    Triangle centrality

    Paul Burkhardt. Triangle centrality. ACM Transactions on Knowledge Discov- ery from Data , 18(9):1–34, 2024

Show all 41 references
  1. [9]

    Perron-frobenius theorem for nonnegative tensors

    Kung-Ching Chang, Kelly Pearson, and Tan Zhang. Perron-frobenius theorem for nonnegative tensors. Communications in Mathematical Sciences , 6(2):507– 520, 2008

  2. [10]

    Higher-order link prediction using triangle embeddings

    Neeraj Chavan and Katerina Potika. Higher-order link prediction using triangle embeddings. In 2020 IEEE International Conference on Big Data (Big Data) , pages 4535–4544. IEEE, 2020

  3. [11]

    Spectra of power hy- pergraphs and signed graphs via parity-closed walks

    Lixiang Chen, Edwin R van Dam, and Changjiang Bu. Spectra of power hy- pergraphs and signed graphs via parity-closed walks. Journal of Combinatorial Theory, Series A , 207:105909, 2024

  4. [12]

    Local fiedler vector centrality for detection of deep and overlapping communities in networks

    Pin-Yu Chen and Alfred O Hero. Local fiedler vector centrality for detection of deep and overlapping communities in networks. In 2014 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 1120–

  5. [13]

    Deep community detection

    Pin-Yu Chen and Alfred O Hero. Deep community detection. IEEE Transac- tions on Signal Processing , 63(21):5706–5719, 2015

  6. [14]

    Spectra of uniform hypergraphs

    Joshua Cooper and Aaron Dutle. Spectra of uniform hypergraphs. Linear Algebra and its applications , 436(9):3268–3292, 2012

  7. [15]

    Study on centrality mea- sures in social networks: a survey

    Kousik Das, Sovan Samanta, and Madhumangal Pal. Study on centrality mea- sures in social networks: a survey. Social network analysis and mining , 8:1–11, 2018

  8. [16]

    Communication in complex networks

    Omar De la Cruz Cabrera, Jiafeng Jin, Silvia Noschese, and Lothar Re- ichel. Communication in complex networks. Applied Numerical Mathematics , 172:186–205, 2022

  9. [17]

    Degree relations of triangles in real-world networks and graph models

    Nurcan Durak, Ali Pinar, Tamara G Kolda, and C Seshadhri. Degree relations of triangles in real-world networks and graph models. In Proceedings of the 21st ACM international conference on Information and knowledge management, pages 1712–1716, 2012

  10. [18]

    Perron communica- bility and sensitivity of multilayer networks

    Smahane El-Halouy, Silvia Noschese, and Lothar Reichel. Perron communica- bility and sensitivity of multilayer networks. Numerical Algorithms, 92(1):597– 617, 2023

  11. [19]

    Subgraph centrality in com- plex networks

    Ernesto Estrada and Juan A Rodriguez-Velazquez. Subgraph centrality in com- plex networks. Physical Review E , 71(5):056103, 2005

  12. [20]

    A property of eigenvectors of nonnegative symmetric matri- ces and its application to graph theory

    Miroslav Fiedler. A property of eigenvectors of nonnegative symmetric matri- ces and its application to graph theory. Czechoslovak mathematical journal , 25(4):619–633, 1975

  13. [21]

    Centrality in valued graphs: A measure of betweenness based on network flow

    Linton C Freeman, Stephen P Borgatti, and Douglas R White. Centrality in valued graphs: A measure of betweenness based on network flow. Social networks, 13(2):141–154, 1991

  14. [22]

    Perron–frobenius theo- rem for nonnegative multilinear forms and extensions

    Shmuel Friedland, St´ ephane Gaubert, and Lixing Han. Perron–frobenius theo- rem for nonnegative multilinear forms and extensions. Linear Algebra and its Applications, 438(2):738–749, 2013

  15. [23]

    Graph clustering using triangle- aware measures in large networks

    Yang Gao, Xiangzhan Yu, and Hongli Zhang. Graph clustering using triangle- aware measures in large networks. Information Sciences, 584:618–632, 2022. 16

  16. [24]

    Triangular stability maximization by influence spread over social networks

    Zheng Hu, Weiguo Zheng, and Xiang Lian. Triangular stability maximization by influence spread over social networks. Proceedings of the VLDB Endowment, 16(11):2818–2831, 2023

  17. [25]

    Searching for key cycles in a complex network

    Siyang Jiang, Jin Zhou, Michael Small, Jun-an Lu, and Yanqi Zhang. Searching for key cycles in a complex network. Physical Review Letters, 130(18):187402, 2023

  18. [26]

    A new measure of centrality for brain networks

    Karen E Joyce, Paul J Laurienti, Jonathan H Burdette, and Satoru Hayasaka. A new measure of centrality for brain networks. PloS one , 5(8):e12200, 2010

  19. [27]

    A critical review of centrality measures in social networks

    Andrea Landherr, Bettina Friedl, and Julia Heidemann. A critical review of centrality measures in social networks. Business & Information Systems Engi- neering, 2:371–385, 2010

  20. [28]

    Singular values and eigenvalues of tensors: a variational ap- proach

    Lek-Heng Lim. Singular values and eigenvalues of tensors: a variational ap- proach. In 1st IEEE International Workshop on Computational Advances in Multi-Sensor Adaptive Processing, pages 129–132. IEEE, 2005

  21. [29]

    On a generalization of the spectral mantel’s theorem

    Chunmeng Liu and Changjiang Bu. On a generalization of the spectral mantel’s theorem. Journal of Combinatorial Optimization , 46(2):14, 2023

  22. [30]

    The high order spectrum of a graph and its applications in graph colouring and clique counting

    Chunmeng Liu, Jiang Zhou, and Changjiang Bu. The high order spectrum of a graph and its applications in graph colouring and clique counting. Linear and Multilinear Algebra, 71(14):2354–2365, 2023

  23. [31]

    Eigenvector centrality mapping for analyzing con- nectivity patterns in fMRI Data of the human brain

    Gabriele Lohmann, Daniel S Margulies, Annette Horstmann, Burkhard Pleger, Joeran Lepsien, Dirk Goldhahn, Haiko Schloegl, Michael Stumvoll, Arno Vill- ringer, and Robert Turner. Eigenvector centrality mapping for analyzing con- nectivity patterns in fMRI Data of the human brain...

  24. [32]

    David Lusseau, Karsten Schneider, Oliver J Boisseau, Patti Haase, Elisabeth Slooten, and Steve M Dawson. The bottlenose dolphin community of doubtful sound features a large proportion of long-lasting associations: can geographic isolation explain this unique trait? Behavioral ...

  25. [33]

    The local triangle structure centrality method to rank nodes in networks

    Xiaojian Ma and Yinghong Ma. The local triangle structure centrality method to rank nodes in networks. Complexity, 2019(1):9057194, 2019. 17

  26. [34]

    Eigenvalues of a real supersymmetric tensor

    Liqun Qi. Eigenvalues of a real supersymmetric tensor. Journal of Symbolic Computation, 40(6):1302–1324, 2005

  27. [35]

    Tensor Analysis: Spectral Theory and Special Tensors

    Liqun Qi and Ziyan Luo. Tensor Analysis: Spectral Theory and Special Tensors. SIAM, Philadelphia, 2017

  28. [36]

    The network data repository with interactive graph analytics and visualization

    Ryan Rossi and Nesreen Ahmed. The network data repository with interactive graph analytics and visualization. In Proceedings of the AAAI conference on artificial intelligence , volume 29, 2015

  29. [37]

    Node and edge nonlinear eigenvec- tor centrality for hypergraphs

    Francesco Tudisco and Desmond J Higham. Node and edge nonlinear eigenvec- tor centrality for hypergraphs. Communications Physics, 4(1):201, 2021

  30. [38]

    The two-steps eigenvector centrality in complex networks

    Qing Xu, Lizhu Sun, and Changjiang Bu. The two-steps eigenvector centrality in complex networks. Chaos, Solitons and Fractals , 173:113753, 2023

  31. [39]

    An information flow model for conflict and fission in small groups

    Wayne W Zachary. An information flow model for conflict and fission in small groups. Journal of anthropological research, 33(4):452–473, 1977

  32. [40]

    Community detection via a tri- angle and edge combination conductance partitioning

    Teng Zhang, Lizhu Sun, and Changjiang Bu. Community detection via a tri- angle and edge combination conductance partitioning. Journal of Statistical Mechanics: Theory and Experiment , 2020(7):073405, 2020

  33. [41]

    Estrada index and subgraph cen- trality of hypergraphs via tensors

    Hong Zhou, Lizhu Sun, and Changjiang Bu. Estrada index and subgraph cen- trality of hypergraphs via tensors. Discrete Applied Mathematics, 341:120–129, 2023. 18

Pith tools

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