Pith. sign in

REVIEW 5 major objections 5 minor 126 references

Multi-Granularity Position Embedding of Graphs via Granular-Ball for Link Prediction

T0 review · 5 major / 5 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read The paper claims that link prediction improves when node positions are encoded relative to landmarks at multiple granularities, organized as a tree of homophilic subgraphs.

desk verdict A reasonable, clearly-written extension of HPLC to multi-granularity landmarks, but the 'consistent SOTA' claim is contradicted by its own Table 1 and the evaluation reporting needs tightening. read the letter →

arxiv 2607.29115 v1 pith:AHEMVFXK submitted 2026-07-31 cs.SI cs.AI

Sen Zhao , Cheng Liu , Shuyin Xia , Zhiyuan Liu , Yi Liu , Yi Wang , Wei Wang This is my paper
classification cs.SIcs.AI
keywords linkpredictionpositionembeddinggranular-ballcomputingmulti-granularityhierarchicalgraphhomophilyneuralnetworkslandmarks
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

The paper argues that node position embeddings for link prediction should capture homophilic structures at multiple levels of granularity, not just a single level. It proposes MGLP, which recursively splits the graph into connected subgraphs ('granular balls') scored by a simple internal edge-density measure, treats the central node of each ball as a landmark, and organizes these landmarks into a hierarchical central graph. Node embeddings combine spectral center vectors from this graph with distance vectors whose edge weights are adjusted by tree depth. Experiments on six networks report that MGLP outperforms all 17 baselines, including a prior landmark-based method, on AUC or Hits@K. If correct, the finding suggests that the scale at which homophily is measured matters for link prediction, and that one adaptive splitting rule can supply that scale across diverse graphs.

What carries the argument

Granular-ball graph refinement — an adaptive greedy binary split of the graph into connected subgraphs scored by Q = 2|E|/|V|, kept when a child improves Q — produces a hierarchical tree and a Hierarchical Central Graph (HCG). The HCG's Laplacian eigenvectors provide center vectors; a Multi-Granularity Hierarchical Distance (MGHD) weights shortest-path distances by tree depth to produce distance vectors. The two vectors concatenate into the position embedding.

What would settle it

Run MGLP on a stochastic block model with planted communities and compare the granular-ball hierarchy to the planted partition: if splits reduce Q by cutting true communities, or if a random/fixed-depth tree matches MGLP's accuracy, the core claim that adaptive refinement captures optimal granularity would be contradicted.

Watch

Extended reading notes

Core claim

On its own terms, MGLP establishes that a position embedding built from two complementary vectors — a center vector from Laplacian eigenvectors of a hierarchical graph of granular-ball central nodes, and a distance vector whose edge weights are modified by tree depth — outperforms single-granularity landmark embeddings and all other baselines on six link-prediction benchmarks (Cora, Citeseer, PubMed, Facebook, DDI, COLLAB). The refinement rule retains a split only when at least one child granular ball has higher quality Q = 2|E|/|V| than its parent, deepening the hierarchy exactly where internal connectivity is densest, and the resulting landmarks appear at an adaptive granularity that varie

Load-bearing premise

The load-bearing premise is that average internal edge density Q = 2|E|/|V| is a faithful proxy for homophily, so that a greedy split retained only when a child improves Q yields the optimal granularity hierarchy; if Q misidentifies meaningful subdomains, the landmarks and distances built on them inherit that error.

Editorial extensions

If this is right

  • Position embeddings with multi-granularity landmarks improve link prediction accuracy across both dense and sparse graphs.
  • The adaptive split rule automatically picks a per-region granularity, removing the need to tune landmark scale per dataset.
  • The method scales to a 235k-node graph using about √N initial landmarks.
  • Ablation shows both center vectors and distance vectors contribute; distance encoding adds signal beyond landmark positions alone.
  • The embeddings integrate with a standard GNN and binary cross-entropy training, so gains come from the position representation, not a bespoke decoder.

Reading between the lines

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

  • One could test whether the quality metric Q is the right homophily proxy by comparing it to conductance or feature similarity as the split criterion; if another proxy yields a different hierarchy with better link prediction, Q is not the load-bearing choice.
  • The greedy split condition explores only one hierarchical partition; a global or randomized search over granular-ball trees might produce better landmarks, especially on graphs with overlapping communities.
  • The depth-weighted distance in Eq. 12 blends tree depth with ordinary shortest-path length; an ablation that separates these two factors would clarify exactly what the hierarchy contributes beyond standard distances.
  • The position embeddings are architecture-agnostic, so they could be plugged into transformers or edge-focused GNNs for link prediction, and possibly adapted to node classification or graph-level tasks.
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

5 major / 5 minor

Summary. The paper proposes MGLP, a link-prediction method that uses granular-ball decomposition to partition a graph into homophilic subdomains at multiple granularities, treats the central nodes of these subdomains as landmarks, builds a Hierarchical Central Graph (HCG) to encode landmark positions, and computes Multi-Granularity Hierarchical Distances (MGHD) as node-to-landmark distances. Node embeddings are formed by concatenating center vectors and distance vectors and are fed into a GNN. The method is evaluated on six datasets against 17 baselines, with ablations and a parameter study, and the authors claim consistent state-of-the-art results.

Significance. If the empirical claims survive scrutiny, the multi-granularity landmark idea is a plausible and useful extension of single-granularity landmark methods such as HPLC. The paper provides an anonymous code link, reports means and standard deviations over 10 runs, and includes ablations showing that each component contributes to performance. However, the central claim of consistent superiority is contradicted by the paper's own Table 1, and the parameter analysis contains an inconsistency that undermines the choice of the key hyperparameter α. As written, the evidence supports competitiveness on most datasets, not consistent state-of-the-art.

major comments (5)
  1. [Sec. 4.2, Table 1] The sentence 'MGLP consistently surpasses all 17 baselines across datasets' is directly contradicted by the Facebook row: HPLC achieves AUC 99.69±0.00 while MGLP achieves 99.66±0.00. Thus MGLP does not surpass HPLC on Facebook. The abstract's 'consistently outperforms state-of-the-art methods' and Observation 1 in Section 4.2 must be revised to reflect the actual result (e.g., 'competitive on 5 of 6 datasets') or supported by a statistical test showing the 0.03 difference is not significant. As written, this is an internal inconsistency in the central empirical claim.
  2. [Sec. 4.1, Table 1] For DDI and COLLAB, the evaluation metric is described only as 'the proportion of positive edges in the test data that ranked within the top-k positions,' but the value of k (K) is never specified. The reported numbers (e.g., DDI 71.01 vs 70.03; COLLAB 56.59 vs 56.04) are therefore uninterpretable and not reproducible. Please state the exact K used (e.g., Hits@100) and, ideally, report Hits@10/50/100 as well.
  3. [Sec. 4.4 vs Sec. 3.2, Fig. 3] The text says the number of initial central nodes is α=√N (Eq. 3) and that Fig. 3 demonstrates this choice is best. However, Fig. 3's x-axis is labeled 'N/4, N/2, N, 2N,' which are O(N) values, not O(√N). For Cora (N=2,708), √N≈52 while N/4≈677, so the tested values never include α=√N. Thus Figure 3 cannot support the stated claim. Please correct the axis labels if they actually denote multiples of √N, or rerun the analysis with α values that include √N.
  4. [Sec. 4.4, Fig. 3] The hyperparameter α is selected by comparing test-set AUC on the same evaluation datasets used to report final results. This constitutes test-set overfitting: the final numbers are obtained after peeking at the test labels, while the baseline hyperparameters are not given the same tuning opportunity. To support the reported comparisons, select α on a separate validation split or via nested cross-validation, or show that performance is insensitive to α across a range that clearly includes √N.
  5. [Sec. 3.2, Eqs. (2),(6)] The abstract and Section 3 claim the refinement produces 'optimal levels of granularity,' but no optimality argument is given for the greedy binary split using Q=2|E|/|V| as the quality measure. The split-retention rule (Eq. 6) only ensures that one child has higher edge density than the parent; this is a local heuristic. The claim 'optimal granularity' should be weakened to 'adaptive granularity' unless a supporting analysis is provided. This is central because all landmarks and distances are built on this refinement hierarchy.
minor comments (5)
  1. [Eq. (12)] The distance weight formula can divide by zero when |Depth(x_i)-Depth(x_{i+1})| = Depth_max (e.g., an edge connecting the root and a deepest leaf). Please clarify the intended behavior or add a small regularization term.
  2. [Notation] The symbol T is used both for the hierarchical tree (Sec. 3.2) and for the heat-kernel normalizing parameter (Eq. 8). This is confusing; please use distinct notation.
  3. [Table 3] The table caption does not define MG, CV, and DV, although they are defined in the text. Adding definitions to the caption would improve readability.
  4. [Sec. 4.1] The text says 'we computed the average value of the evaluation metrics across different datasets,' but no average row is provided in Table 1. Either add it or remove the sentence.
  5. [Table 1] The column heads 'HITS@K' should be replaced with a defined metric name (e.g., 'Hits@K') and the K value should be specified, as raised in Major Comment 2.

Circularity Check

0 steps flagged · score 0.0 of 10

No derivation-to-fit circularity; MGLP is an empirical heuristic, and its SOTA claim is internally inconsistent but not circular.

full rationale

I walked the claimed derivation chain: Granular-Ball refinement (Sec. 3.2) uses Q(G)=2|E|/|V| and the split condition (Eq. 6) as a heuristic stopping rule; Eqs. 10-12 define a distance encoder; no equation forces a link-prediction output to equal any fitted quantity or training target by construction. The components (MG, CV, DV) are ablated empirically (Table 3), and results are benchmarked against external baselines, so self-citations to Xia (2019) and Xie et al. (2023) supply vocabulary and heuristics but not the measured performance. Two non-circular concerns should be recorded: (i) Sec. 4.4 selects sqrt(N) initial central nodes by test-set AUC (Fig. 3) while Table 1 uses the same test folds; this is test-set hyperparameter selection, a validity/reproducibility flaw, not an algebraic reduction. (ii) Sec. 4.2's claim that MGLP 'consistently surpasses all 17 baselines' is contradicted by Table 1 on Facebook (HPLC 99.69 vs MGLP 99.66). HITS@K for DDI/COLLAB is never defined (no K), another reporting flaw. These are correctness issues, not circularity: no central claim reduces by construction to its inputs, so the circularity score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The central claim rests on heuristic design choices: the quality metric, the split rule, the depth-based distance weighting, and the number of initial landmarks. Most of these are introduced ad hoc to make the method work, with no independent theoretical grounding. The HCG is an algorithmic construct, not a discovered entity.

free parameters (3)
  • α (initial central node count) = √N (N = #nodes), validated on test sets
    The number of initial granular-ball central nodes is set to √N, drawing on Xie et al. (2020), and is validated via test-set experiments in Section 4.4.
  • Heat-kernel parameter T (Eq. 8) = Not specified
    Edge weights in the Hierarchical Central Graph use a heat kernel exp(-d(u,v)^2/T) but no value or tuning procedure for T is given.
  • Number of Laplacian eigenvectors used as center vectors = Not specified
    The paper says eigenvectors of the HCG Laplacian are extracted, but never states how many are used or how they are combined into the final center vector.
assumptions (4)
  • domain assumption Granular-ball quality Q=2|E|/|V| measures homophily
    Eq. 2 defines quality purely by average internal degree; no evidence is provided that this corresponds to feature-based homophily or community structure.
  • ad hoc to paper Greedy split rule (Eq. 6) yields optimal granularity
    The split condition only requires one child to have higher quality than the parent; there is no convergence guarantee or proof that this produces 'optimal' levels of granularity.
  • domain assumption Nodes assigned to nearest central node form connected subgraphs
    Eq. 4 assigns by shortest-path distance but does not enforce connectivity; Definition 1 assumes granular-balls are connected subgraphs, which the assignment heuristic may violate.
  • domain assumption Shortest-path distances with depth-based edge weights remain meaningful
    Eq. 11-12 define a weighted distance where edges across tree levels are amplified; this assumes the resulting distances preserve structural proximity in the original graph.
invented entities (1)
  • Hierarchical Central Graph (HCG)
    purpose: Encodes relative positions among granular-ball central nodes via a weighted graph and Laplacian eigenvectors.
    The HCG is a construction internal to the algorithm, not a new physical entity; it has no falsifiable handle outside the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multi-Granularity Position Embedding of Graphs via Granular-Ball for Link Prediction." pith.science (2026). https://pith.science/paper/AHEMVFXK

@misc{pith2026260729115,
  author       = {Pith},
  title        = {Pith review of: Multi-Granularity Position Embedding of Graphs via Granular-Ball for Link Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AHEMVFXK}},
  note         = {Machine review of arXiv:2607.29115}
}
read the original abstract

Link prediction aims to identify potential or future connections within a given graph structure. Position information is essential for link prediction, as it distinguishes homogeneous nodes through their relative relationships, facilitating the accurate capture of structural patterns and implicit connections. Previous studies derive node positional information as distances to single-granularity landmarks, defined as the centers of homophilic regions, while neglecting the multi-granularity nature of homophilic structures and their hierarchical interrelations. We propose the Multi-Granularity Position Embedding of Graphs via Granular-Ball for Link Prediction (MGLP) method to obtain multi-granularity position embedding of graphs. Specifically, MGLP introduces an Adaptive Granular-Ball Graph Refinement mechanism to adaptively refine the graph into homophilic subdomains with optimal levels of granularity. The central nodes within subdomains are treated as landmarks, which form a Hierarchical Central Graph. Moreover, a novel Multi-granularity Hierarchical Distance encoding mechanism is proposed to capture both the homophilic structures within a graph and their hierarchical correlations, improving the discriminative power of nodes. Experimental results demonstrate that the multi-granularity position embedding generated by our method exhibits excellent performance and strong competitiveness compared to baseline algorithms for link prediction. Our codes are available in https://anonymous.4open.science/r/MGLP-D3C5/.

Figures

Figures reproduced from arXiv: 2607.29115 by the authors.

Figure 1
Figure 1. The comparison of Uni-Granularity and Multi-Granularity Position Embedding. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Overview of MGLP. I. Granularizing the original graph, select the central node of each granular-ball. II. Construct the HCG and compute center vectors. III. Using MGHD and the granularization results to compute the distance vectors, with the node’s hierarchical level represented by the color in the HCG. IV. Concatenate center vectors and distance vectors to form node position embeddings, and put these embeddings int… view at source ↗
Figure 3
Figure 3. Parametric Analysis of Initial Granular-Ball Central Nodes [PITH_FULL_IMAGE:figures/full_fig_p015_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

126 extracted references · 10 linked inside Pith

  1. [1]

    arXiv e-prints , pages=

    Semi-Supervised Classification with Graph Convolutional Networks , author=. arXiv e-prints , pages=

  2. [2]

    International Conference on Learning Representations , year=

    Semi-Supervised Classification with Graph Convolutional Networks , author=. International Conference on Learning Representations , year=

  3. [3]

    International Conference on Learning Representations , year=

    How Powerful are Graph Neural Networks? , author=. International Conference on Learning Representations , year=

  4. [4]

    Advances in neural information processing systems , volume=

    Inductive representation learning on large graphs , author=. Advances in neural information processing systems , volume=

  5. [5]

    Mathematics of control, signals and systems , volume=

    Approximation by superpositions of a sigmoidal function , author=. Mathematics of control, signals and systems , volume=. 1989 , publisher=

    1989

  6. [6]

    International conference on machine learning , pages=

    Position-aware graph neural networks , author=. International conference on machine learning , pages=. 2019 , organization=

    2019

  7. [7]

    Advances in Neural Information Processing Systems , volume=

    Distance encoding: Design provably more powerful neural networks for graph representation learning , author=. Advances in Neural Information Processing Systems , volume=

  8. [8]

    Advances in neural information processing systems , volume=

    Link prediction based on graph neural networks , author=. Advances in neural information processing systems , volume=

Show all 126 references
  1. [9]

    The European Physical Journal B , volume=

    Predicting missing links via local information , author=. The European Physical Journal B , volume=. 2009 , publisher=

    2009

  2. [10]

    science , volume=

    Emergence of scaling in random networks , author=. science , volume=. 1999 , publisher=

    1999

  3. [11]

    arXiv preprint arXiv:1611.07308 , year=

    Variational graph auto-encoders , author=. arXiv preprint arXiv:1611.07308 , year=

    arXiv

  4. [12]

    Proceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

    node2vec: Scalable feature learning for networks , author=. Proceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

  5. [13]

    Computer , volume=

    Matrix factorization techniques for recommender systems , author=. Computer , volume=. 2009 , publisher=

    2009

  6. [14]

    Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

    Deepwalk: Online learning of social representations , author=. Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

  7. [15]

    Advances in Neural Information Processing Systems , volume=

    Labeling trick: A theory of using graph neural networks for multi-node representation learning , author=. Advances in Neural Information Processing Systems , volume=

  8. [16]

    Advances in Neural Information Processing Systems , volume=

    Subgraph neural networks , author=. Advances in Neural Information Processing Systems , volume=

  9. [17]

    International Conference on Learning Representations , year=

    GLASS: GNN with Labeling Tricks for Subgraph Representation Learning , author=. International Conference on Learning Representations , year=

  10. [18]

    Advances in neural information processing systems , volume=

    Open graph benchmark: Datasets for machine learning on graphs , author=. Advances in neural information processing systems , volume=

  11. [19]

    Applied and Computational Harmonic Analysis , volume=

    Wavelets on graphs via spectral graph theory , author=. Applied and Computational Harmonic Analysis , volume=. 2011 , publisher=

    2011

  12. [20]

    International conference on complex networks and their applications , pages=

    Fluid communities: A competitive, scalable and diverse community detection algorithm , author=. International conference on complex networks and their applications , pages=. 2017 , organization=

    2017

  13. [21]

    International conference on machine learning , pages=

    Neural message passing for quantum chemistry , author=. International conference on machine learning , pages=. 2017 , organization=

    2017

  14. [22]

    Joshi and Thomas Laurent and Yoshua Bengio and Xavier Bresson , title=

    Vijay Prakash Dwivedi and Chaitanya K. Joshi and Thomas Laurent and Yoshua Bengio and Xavier Bresson , title=. CoRR , volume=. 2020 , cdate=

    2020

  15. [23]

    2021 , url=

    Weihua Hu and Matthias Fey and Hongyu Ren and Maho Nakata and Yuxiao Dong and Jure Leskovec , booktitle=. 2021 , url=

    2021

  16. [24]

    Communications of the ACM , volume=

    WordNet: a lexical database for English , author=. Communications of the ACM , volume=. 1995 , publisher=

    1995

  17. [25]

    Proceedings of the 2008 ACM SIGMOD international conference on Management of data , pages=

    Freebase: a collaboratively created graph database for structuring human knowledge , author=. Proceedings of the 2008 ACM SIGMOD international conference on Management of data , pages=

    2008

  18. [26]

    Scientific american , volume=

    Scale-free networks , author=. Scientific american , volume=. 2003 , publisher=

    2003

  19. [27]

    Israel Journal of Mathematics , volume=

    On Lipschitz embedding of finite metric spaces in Hilbert space , author=. Israel Journal of Mathematics , volume=. 1985 , publisher=

    1985

  20. [28]

    Combinatorica , volume=

    The geometry of graphs and some of its algorithmic applications , author=. Combinatorica , volume=. 1995 , publisher=

    1995

  21. [29]

    Journal of the ACM (JACM) , volume=

    Fibonacci heaps and their uses in improved network optimization algorithms , author=. Journal of the ACM (JACM) , volume=. 1987 , publisher=

    1987

  22. [30]

    Advances in Neural Information Processing Systems , volume=

    Neo-gnns: Neighborhood overlap-aware graph neural networks for link prediction , author=. Advances in Neural Information Processing Systems , volume=

  23. [31]

    2021 , url=

    Global Attention Improves Graph Networks Generalization , author=. 2021 , url=

    2021

  24. [32]

    Graph Attention Networks

    Veli. Graph Attention Networks. International Conference on Learning Representations , year=

  25. [33]

    arXiv preprint arXiv:1710.10903 , year=

    Graph attention networks , author=. arXiv preprint arXiv:1710.10903 , year=

    arXiv

  26. [34]

    Proceedings of the 37th International Conference on Machine Learning , pages =

    Contrastive Multi-View Representation Learning on Graphs , author =. Proceedings of the 37th International Conference on Machine Learning , pages =. 2020 , editor =

    2020

  27. [35]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    Line graph neural networks for link prediction , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

  28. [36]

    International conference on machine learning , pages=

    Representation learning on graphs with jumping knowledge networks , author=. International conference on machine learning , pages=. 2018 , organization=

    2018

  29. [37]

    Kingma and Jimmy Ba , title=

    Diederik P. Kingma and Jimmy Ba , title=. 2015 , cdate=

    2015

  30. [38]

    International Conference on Learning Representations , year=

    On the Equivalence between Positional Node Embeddings and Structural Graph Representations , author=. International Conference on Learning Representations , year=

  31. [39]

    Proceedings of the 18th ACM conference on Information and knowledge management , pages=

    Fast shortest path distance estimation in large networks , author=. Proceedings of the 18th ACM conference on Information and knowledge management , pages=

  32. [40]

    Proceedings

    Predicting Internet network distance with coordinates-based approaches , author=. Proceedings. Twenty-First Annual Joint Conference of the IEEE Computer and Communications Societies , volume=. 2002 , organization=

    2002

  33. [41]

    IEEE/ACM Transactions On Networking , volume=

    IDMaps: A global Internet host distance estimation service , author=. IEEE/ACM Transactions On Networking , volume=. 2001 , publisher=

    2001

  34. [42]

    24th International Conference on Distributed Computing Systems, 2004

    PIC: Practical Internet coordinates for distance estimation , author=. 24th International Conference on Distributed Computing Systems, 2004. Proceedings. , pages=. 2004 , organization=

    2004

  35. [43]

    Proceedings of the 3rd ACM SIGCOMM Conference on Internet Measurement , pages=

    Constructing internet coordinate system based on delay measurement , author=. Proceedings of the 3rd ACM SIGCOMM Conference on Internet Measurement , pages=

  36. [44]

    ACM SIGCOMM Computer Communication Review , volume=

    Vivaldi: A decentralized network coordinate system , author=. ACM SIGCOMM Computer Communication Review , volume=. 2004 , publisher=

    2004

  37. [45]

    Physical Review E , volume=

    Average path length in random networks , author=. Physical Review E , volume=. 2004 , publisher=

    2004

  38. [46]

    Physical review E , volume=

    Random graphs with arbitrary degree distributions and their applications , author=. Physical review E , volume=. 2001 , publisher=

    2001

  39. [47]

    Graph Neural Networks: Foundations, Frontiers, and Applications , pages=

    Graph neural networks , author=. Graph Neural Networks: Foundations, Frontiers, and Applications , pages=. 2022 , publisher=

    2022

  40. [48]

    Physical Review E , volume=

    Class of correlated random networks with hidden variables , author=. Physical Review E , volume=. 2003 , publisher=

    2003

  41. [49]

    Reviews of modern physics , volume=

    Statistical mechanics of complex networks , author=. Reviews of modern physics , volume=. 2002 , publisher=

    2002

  42. [50]

    On the evolution of random graphs , author=. Publ. Math. Inst. Hung. Acad. Sci , volume=

  43. [51]

    Proceedings of the 39th International Conference on Machine Learning , pages =

    Learning from Counterfactual Links for Link Prediction , author =. Proceedings of the 39th International Conference on Machine Learning , pages =. 2022 , editor =

    2022

  44. [52]

    International Conference on Learning Representations , year=

    Neural Link Prediction with Walk Pooling , author=. International Conference on Learning Representations , year=

  45. [53]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    A multi-scale approach for graph link prediction , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  46. [54]

    Advances in Neural Information Processing Systems , volume=

    Graph contrastive learning with augmentations , author=. Advances in Neural Information Processing Systems , volume=

  47. [55]

    45th Annual IEEE Symposium on Foundations of Computer Science , pages=

    Triangulation and embedding using small sets of beacons , author=. 45th Annual IEEE Symposium on Foundations of Computer Science , pages=. 2004 , organization=

    2004

  48. [56]

    arXiv preprint arXiv:1903.02428 , year=

    Fast graph representation learning with PyTorch Geometric , author=. arXiv preprint arXiv:1903.02428 , year=

    1903 arXiv

  49. [57]

    Proceedings of the ACM Web Conference 2022 , pages=

    ClusterSCL: Cluster-Aware Supervised Contrastive Learning on Graphs , author=. Proceedings of the ACM Web Conference 2022 , pages=

    2022

  50. [58]

    Twenty-Second International Joint Conference on Artificial Intelligence , year=

    Theoretical justification of popular link prediction heuristics , author=. Twenty-Second International Joint Conference on Artificial Intelligence , year=

  51. [59]

    Proceedings of the twelfth international conference on Information and knowledge management , pages=

    The link prediction problem for social networks , author=. Proceedings of the twelfth international conference on Information and knowledge management , pages=

  52. [60]

    Psychometrika , volume=

    A new status index derived from sociometric analysis , author=. Psychometrika , volume=. 1953 , publisher=

    1953

  53. [61]

    Physical review letters , volume=

    Scale-free networks are ultrasmall , author=. Physical review letters , volume=. 2003 , publisher=

    2003

  54. [62]

    International Conference on Learning Representations , year=

    Equivariant and Stable Positional Encoding for More Powerful Graph Neural Networks , author=. International Conference on Learning Representations , year=

  55. [63]

    arXiv preprint arXiv:2003.00982 , year=

    Benchmarking graph neural networks , author=. arXiv preprint arXiv:2003.00982 , year=

    2003 arXiv

  56. [64]

    Advances in Neural Information Processing Systems , volume=

    Rethinking graph transformers with spectral attention , author=. Advances in Neural Information Processing Systems , volume=

  57. [65]

    Annals of Mathematics , pages=

    On Poisson's summation formula , author=. Annals of Mathematics , pages=. 1941 , publisher=

    1941

  58. [66]

    AAAI Workshop on Deep Learning on Graphs: Methods and Applications , year=

    A Generalization of Transformer Networks to Graphs , author=. AAAI Workshop on Deep Learning on Graphs: Methods and Applications , year=

  59. [67]

    Neural computation , volume=

    Laplacian eigenmaps for dimensionality reduction and data representation , author=. Neural computation , volume=. 2003 , publisher=

    2003

  60. [68]

    Computer science review , volume=

    Graph clustering , author=. Computer science review , volume=. 2007 , publisher=

    2007

  61. [69]

    Journal of statistical mechanics: theory and experiment , volume=

    Fast unfolding of communities in large networks , author=. Journal of statistical mechanics: theory and experiment , volume=. 2008 , publisher=

    2008

  62. [70]

    The Eleventh International Conference on Learning Representations , year=

    Graph Neural Networks for Link Prediction with Subgraph Sketching , author=. The Eleventh International Conference on Learning Representations , year=

  63. [71]

    IEEE transactions on neural networks , volume=

    The graph neural network model , author=. IEEE transactions on neural networks , volume=. 2008 , publisher=

    2008

  64. [72]

    SIAM Journal on Applied Mathematics , volume=

    Spectra of Laplacian matrices of weighted graphs: structural genericity properties , author=. SIAM Journal on Applied Mathematics , volume=. 2018 , publisher=

    2018

  65. [73]

    Physical review letters , volume=

    Skeleton and fractal scaling in complex networks , author=. Physical review letters , volume=. 2006 , publisher=

    2006

  66. [74]

    Journal of the ACM (JACM) , volume=

    Highway dimension and provably efficient shortest path algorithms , author=. Journal of the ACM (JACM) , volume=. 2016 , publisher=

    2016

  67. [75]

    Proceedings of the ACM Web Conference 2023 , pages=

    xGCN: An Extreme Graph Convolutional Network for Large-scale Social Link Prediction , author=. Proceedings of the ACM Web Conference 2023 , pages=

    2023

  68. [76]

    Proceedings of the Web Conference 2021 , pages=

    Multi-level hyperedge distillation for social linking prediction on sparsely observed networks , author=. Proceedings of the Web Conference 2021 , pages=

    2021

  69. [77]

    Proceedings of the Web Conference 2021 , pages=

    Multiplex bipartite network embedding using dual hypergraph convolutional networks , author=. Proceedings of the Web Conference 2021 , pages=

    2021

  70. [78]

    Proceedings of the Web Conference 2021 , pages=

    Hashing-accelerated graph neural networks for link prediction , author=. Proceedings of the Web Conference 2021 , pages=

    2021

  71. [79]

    nature , volume=

    Collective dynamics of ‘small-world’networks , author=. nature , volume=. 1998 , publisher=

    1998

  72. [80]

    Advances in Neural Information Processing Systems , volume=

    Neural bellman-ford networks: A general graph neural network framework for link prediction , author=. Advances in Neural Information Processing Systems , volume=

  73. [81]

    Proceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

    Using structure indices for efficient approximation of network properties , author=. Proceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

  74. [82]

    , author=

    Computing the shortest path: A search meets graph theory. , author=. SODA , volume=

  75. [83]

    Proceedings of the 3rd ACM SIGCOMM conference on Internet measurement , pages=

    Virtual landmarks for the internet , author=. Proceedings of the 3rd ACM SIGCOMM conference on Internet measurement , pages=

  76. [84]

    Proceedings of the 2013 ACM SIGMOD International Conference on Management of Data , pages=

    Fast exact shortest-path distance queries on large networks by pruned landmark labeling , author=. Proceedings of the 2013 ACM SIGMOD International Conference on Management of Data , pages=

    2013

  77. [85]

    IEEE Transactions on Knowledge and Data Engineering , volume=

    Approximate shortest distance computing: A query-dependent local landmark scheme , author=. IEEE Transactions on Knowledge and Data Engineering , volume=. 2012 , publisher=

    2012

  78. [86]

    Proceedings of the 20th ACM international conference on Information and knowledge management , pages=

    Fast fully dynamic landmark-based estimation of shortest path distances in very large graphs , author=. Proceedings of the 20th ACM international conference on Information and knowledge management , pages=

  79. [87]

    3rd Workshop on Online Social Networks (WOSN 2010) , year=

    Orion: shortest path estimation for large social graphs , author=. 3rd Workshop on Online Social Networks (WOSN 2010) , year=

    2010

  80. [88]

    1998 , institution=

    The pagerank citation ranking: Bring order to the web , author=. 1998 , institution=

    1998

  81. [89]

    Journal of Granular Computing , volume=

    Granular ball computing: A new computational model for data analysis , author=. Journal of Granular Computing , volume=

  82. [90]

    IEEE Transactions on Knowledge and Data Engineering , volume=

    An efficient spectral clustering algorithm based on granular-ball , author=. IEEE Transactions on Knowledge and Data Engineering , volume=. 2023 , publisher=

    2023

  83. [91]

    Physica A: statistical mechanics and its applications , volume=

    Link prediction in complex networks: A survey , author=. Physica A: statistical mechanics and its applications , volume=. 2011 , publisher=

    2011

  84. [92]

    Annual review of sociology , volume=

    Birds of a feather: Homophily in social networks , author=. Annual review of sociology , volume=. 2001 , publisher=

    2001

  85. [93]

    Social networks , volume=

    Friends and neighbors on the web , author=. Social networks , volume=. 2003 , publisher=

    2003

  86. [94]

    Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , pages=

    Knowledge-aware graph neural networks with label smoothness regularization for recommender systems , author=. Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , pages=

  87. [95]

    The world wide web conference , pages=

    Hierarchical temporal convolutional networks for dynamic recommender systems , author=. The world wide web conference , pages=

  88. [96]

    Proceedings of the IEEE , volume=

    A review of relational machine learning for knowledge graphs , author=. Proceedings of the IEEE , volume=. 2015 , publisher=

    2015

  89. [97]

    Journal of machine learning research , volume=

    Mixed membership stochastic blockmodels , author=. Journal of machine learning research , volume=

  90. [98]

    Bioinformatics , volume=

    Boostgapfill: Improving the fidelity of metabolic network reconstructions through integrated constraint and pattern-based methods , author=. Bioinformatics , volume=. 2017 , publisher=

    2017

  91. [99]

    Advances in neural information processing systems , volume=

    Protein interface prediction using graph convolutional networks , author=. Advances in neural information processing systems , volume=

  92. [100]

    Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

    Graph convolutional neural networks for web-scale recommender systems , author=. Proceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

  93. [101]

    Proceedings of the 57th annual meeting of the association for computational linguistics , pages=

    Encoding social information with graph convolutional networks forpolitical perspective detection in news media , author=. Proceedings of the 57th annual meeting of the association for computational linguistics , pages=

  94. [102]

    AI open , volume=

    Graph neural networks: A review of methods and applications , author=. AI open , volume=. 2020 , publisher=

    2020

  95. [103]

    Advances in Neural Information Processing Systems , volume=

    Twibot-22: Towards graph-based twitter bot detection , author=. Advances in Neural Information Processing Systems , volume=

  96. [104]

    Kipf and Max Welling , title =

    Thomas N. Kipf and Max Welling , title =. CoRR , volume =. 2016 , url =. 1609.02907 , timestamp =

    2016 arXiv

  97. [105]

    CoRR , volume =

    Shirui Pan and Ruiqi Hu and Guodong Long and Jing Jiang and Lina Yao and Chengqi Zhang , title =. CoRR , volume =. 2018 , url =. 1802.04407 , timestamp =

    2018 arXiv

  98. [106]

    2022 , eprint=

    Hyperspherical Variational Auto-Encoders , author=. 2022 , eprint=

    2022

  99. [107]

    Proceedings of the 2017 ACM on Conference on Information and Knowledge Management , pages =

    Wang, Chun and Pan, Shirui and Long, Guodong and Zhu, Xingquan and Jiang, Jing , title =. Proceedings of the 2017 ACM on Conference on Information and Knowledge Management , pages =. 2017 , isbn =. doi:10.1145/3132847.3132967 , abstract =

    2017

  100. [108]

    The semantic web: 15th international conference, ESWC 2018, Heraklion, Crete, Greece, June 3--7, 2018, proceedings 15 , pages=

    Modeling relational data with graph convolutional networks , author=. The semantic web: 15th international conference, ESWC 2018, Heraklion, Crete, Greece, June 3--7, 2018, proceedings 15 , pages=. 2018 , organization=

    2018

  101. [109]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    End-to-end structure-aware convolutional networks for knowledge base completion , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  102. [110]

    arXiv preprint arXiv:2110.04375 , year=

    Neural link prediction with walk pooling , author=. arXiv preprint arXiv:2110.04375 , year=

    arXiv

  103. [111]

    Science , volume=

    Topological structure in visual perception , author=. Science , volume=. 1982 , publisher=

    1982

  104. [112]

    Granular Computing , volume=

    DGCC: data-driven granular cognitive computing , author=. Granular Computing , volume=. 2017 , publisher=

    2017

  105. [113]

    IEEE Transactions on Neural Networks and Learning Systems , volume=

    An efficient and adaptive granular-ball generation method in classification problem , author=. IEEE Transactions on Neural Networks and Learning Systems , volume=. 2022 , publisher=

    2022

  106. [114]

    IEEE Transactions on Knowledge and Data Engineering , volume=

    GBNRS: A novel rough set algorithm for fast adaptive attribute reduction in classification , author=. IEEE Transactions on Knowledge and Data Engineering , volume=. 2020 , publisher=

    2020

  107. [115]

    IEEE Transactions on Neural Networks and Learning Systems , volume=

    Granular ball sampling for noisy label classification or imbalanced classification , author=. IEEE Transactions on Neural Networks and Learning Systems , volume=. 2021 , publisher=

    2021

  108. [116]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    A fast adaptive k-means with no bounds , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

  109. [117]

    IEEE Transactions on Neural Networks and Learning Systems , year=

    GBRS: A Unified Granular-Ball Learning Model of Pawlak Rough Set and Neighborhood Rough Set , author=. IEEE Transactions on Neural Networks and Learning Systems , year=

  110. [118]

    arXiv preprint arXiv:2303.02388 , year=

    Graph-based representation for image based on granular-ball , author=. arXiv preprint arXiv:2303.02388 , year=

    arXiv

  111. [119]

    arXiv preprint arXiv:2303.12807 , year=

    Granular-ball optimization algorithm , author=. arXiv preprint arXiv:2303.12807 , year=

    arXiv

  112. [120]

    arXiv preprint arXiv:2210.03120 , year=

    Gbsvm: Granular-ball support vector machine , author=. arXiv preprint arXiv:2210.03120 , year=

    arXiv

  113. [121]

    2006 , publisher=

    Elements of information theory , author=. 2006 , publisher=

    2006

  114. [122]

    Journal of machine learning research , volume=

    An introduction to variable and feature selection , author=. Journal of machine learning research , volume=

  115. [123]

    Information Sciences , volume=

    A new internal index based on density core for clustering validation , author=. Information Sciences , volume=. 2020 , publisher=

    2020

  116. [124]

    Proceedings of the ACM on Web Conference 2024 , pages=

    Hierarchical Position Embedding of Graphs with Landmarks and Clustering for Link Prediction , author=. Proceedings of the ACM on Web Conference 2024 , pages=

    2024

  117. [125]

    Letourneau and J

    M. Letourneau and J. W. Sharp, AMS-StyleGuide-online.pdf, American Mathematical Society, Providence, RI, USA, [Online]. Available: http://www.ams.org/arc/styleguide/index.html

  118. [126]

    Leslie Lamport, : a document preparation system, Addison Wesley, Massachusetts, 2nd edition, 1994

    1994

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.