Pith. sign in

REVIEW 5 major objections 11 minor 43 references

NAEx: A Plug-and-Play Framework for Explaining Network Alignment

T0 review · 5 major / 11 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A model-agnostic, post hoc explainer can now generate faithful sparse explanations for any neural embedding-based network alignment model, including on previously unseen node pairs.

desk verdict A genuinely first inductive NA explainer with a coherent design and decent FID results, but the headline gain is overstated and the FTH metric is partly circular. read the letter →

arxiv 2508.04731 v1 pith:CSV37LJX submitted 2025-08-05 cs.LG cs.IRcs.SI

Shruti Saxena , Arijit Khan , Joydeep Chandra This is my paper
classification cs.LGcs.IRcs.SI
keywords networkalignmentexplainabilitygraphneuralnetworksmodel-agnosticexplanationposthocinductivesubgraphmasksmutualinformation
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 claims that network alignment predictions from any neural embedding-based model can be explained by finding a small set of edges and node features in both networks that preserve the alignment if kept. It introduces NAEx, a framework that learns these explanatory subgraphs jointly across the source and target network, so cross-network dependencies are not lost. Because the explanation network is shared across all edges and features, NAEx can explain new node pairs without retraining, which earlier explanation methods cannot do. The authors argue this makes NAEx the first general-purpose, model-agnostic, post hoc explainer for alignment models, and they demonstrate it on four alignment models over three datasets with new alignment-specific evaluation metrics.

What carries the argument

The load-bearing mechanism is the joint, parameterized edge-and-feature masking over both networks: a shared MLP scores each edge from the concatenation of its endpoint embeddings and the anchor node's embedding, turning graphs into edge distributions from which subgraphs are sampled with the binary-concrete trick, and a shared feature mask selects attributes with Gaussian reparameterization. These masks are optimized by an objective combining alignment consistency (cross-entropy with a sigmoid-of-cosine surrogate for alignment probability), a subgraph-level contrastive loss, and sparsity regularizers. This shared parameterization is what gives NAEx its inductive capability: once trained, ex

What would settle it

Construct a synthetic pair of networks where the true alignment of an anchor pair depends on a planted set of edges and features, train a GNN-based aligner, run NAEx, and check whether the selected subgraph and features match the planted set; if explanations with high NAEx faithfulness do not preserve the aligner's top-1 predictions or miss the planted structure, the framework's faithfulness claim is refuted.

Watch

Extended reading notes

Core claim

NAEx treats explanation as minimizing the conditional entropy of the alignment prediction given masked subgraphs, approximated by a cross-entropy loss between the original alignment score and the score computed from the masked graphs. The alignment score itself is modeled as the sigmoid of the cosine similarity between node embeddings (Eqs. 15-16), a surrogate that lets the framework train with gradient descent. Edge masks are generated from a shared MLP via the binary-concrete reparameterization trick; feature masks are learned jointly with a Gaussian reparameterization. A subgraph contrastive loss pulls the pooled embeddings of the two explanatory subgraphs together for aligned pairs, and

Load-bearing premise

The framework assumes the sigmoid of the cosine similarity between embeddings faithfully represents how the NA model decides alignments; if the real model uses an argmax rule or bidirectional consistency, explanations tuned to this surrogate may not reflect the model's actual decision.

Editorial extensions

If this is right

  • NAEx can be attached to any existing GNN-based alignment model without retraining or altering it, turning black-box embeddings into inspectable subgraph-and-feature explanations.
  • Explanations are generated jointly for both networks, so users can directly compare the structural and semantic reasons an alignment holds.
  • Because the explainer is inductive, once trained it can explain thousands of new anchor pairs in time proportional to the graph edge count, enabling large-scale auditing.
  • The alignment-specific metrics (fidelity, faithfulness, sparsity) provide a ground-truth-free way to benchmark any alignment explainer.
  • The separate edge and feature masks let practitioners see whether a model leans on structure or attributes, supporting model selection and debugging.

Reading between the lines

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

  • If the sigmoid-of-cosine surrogate diverges from an argmax-based decision rule, the reported fidelity numbers may overstate how faithfully NAEx captures the actual model; a direct test against top-1 alignment preservation would settle this.
  • The subgraph contrastive loss imposes similarity between the two explanations; this could make explanations look consistent even when the base model's reasoning is asymmetric, so the loss may need to be relaxed or validated on asymmetric aligners.
  • The framework's inductive generalization suggests a natural extension to dynamic networks: retraining the shared MLP on temporal snapshots could produce explanations that track why alignments emerge or break over time.
  • NAEx could also serve as a diagnostic for overfitting: if a trained aligner's explanations repeatedly highlight peripheral edges or noisy features, that is evidence the model is exploiting spurious correlations.
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 / 11 minor

Summary. The paper proposes NAEx, a post hoc, model-agnostic framework for explaining neural embedding-based network alignment (NA) models. NAEx jointly learns edge masks and feature masks over a source and target graph pair, parameterizing explanation generation with a shared MLP and binary-concrete sampling. The training objective combines an alignment-consistency cross-entropy term (using a sigmoid of cosine similarity between embeddings as a surrogate for alignment probability), a contrastive subgraph-similarity term, and sparsity regularizers. The authors evaluate NAEx with four NA models (JORA, SANA, HTC, ASSISTANT) on three datasets, reporting fidelity (FID), faithfulness (FTH), and sparsity. They claim the method is inductive, efficient, and outperforms baselines including dNAME and GNNExplainer.

Significance. Explaining network alignment is an under-studied but important problem, and NAEx addresses it with a clean, general architecture: shared edge/feature masks over both graphs, a probabilistic subgraph sampling scheme, and a cross-network objective. If the faithfulness claims can be substantiated, NAEx would be a useful plug-and-play tool for diagnosing and comparing NA models, and its inductive inference property is a genuine practical advantage over perturbation-based baselines. The paper introduces tailored evaluation metrics and conducts experiments over multiple datasets and models, which is a strength. However, the current evidence does not fully establish faithfulness to the actual NA model because both the training loss and the primary faithfulness metric rely on the same sigmoid surrogate, and the headline quantitative claims are not clearly supported by the tables.

major comments (5)
  1. [§4.4.1 and §5.4 (Eqs. 15-17, Eq. 23)] The faithfulness metric FTH is essentially the negative exponential of the KL divergence between the original and explained sigmoid score distributions (Eq. 23). This is the same quantity that the alignment-consistency loss Lalign (Eq. 17) drives to be similar. Hence high FTH scores are, to a large degree, a measure of how well the training objective was optimized, not of how faithfully the explanation preserves the actual NA model's decision. The paper must report an evaluation that is not circular, e.g., fidelity computed against the true argmax predictions (Eq. 1), or a metric that does not use the surrogate. As written, the FTH-based claims do not establish faithfulness to the real alignment model.
  2. [§4.4.1 (Eqs. 15-16)] The paper replaces the NA model's prediction rule, which is argmax over pairwise similarities (Eq. 1), with a sigmoid of the cosine similarity between anchor embeddings. This is an unverified assumption. The surrogate may disagree with the argmax decision, especially after subgraph masking or when bidirectional consistency is enforced. The authors provide no theoretical or empirical justification that maximizing the sigmoid likelihood preserves the top-1 ranking of the original model. This is load-bearing because both the training loss and the FTH metric depend on this surrogate. Please add experiments comparing surrogate scores with actual model outputs, or replace the surrogate with a more faithful softmax/ranking-preserving formulation.
  3. [§6.1 and Table 4] The abstract and contributions claim '~32% improvement in explanation fidelity' over dNAME, but this number is not derivable from Table 4. The largest FID gain versus dNAME at any sparsity level shown is about 10.8% (AllMovie-IMDB, S=0.7); the largest FTH gain is about 35.2% (ACM-DBLP, S=0.7). The text also reports an 'average improvement of 2.87−65.74% in FID' without specifying how the range is computed or over which cells. Please state explicitly which quantity is being averaged, whether relative or absolute, and report the per-cell gains. The 'up to 95% speedup' claim in the abstract is likewise not substantiated by Figure 4, which shows training/inference time but no percentage calculation.
  4. [§6.1 and Table 4] The comparison in Table 4 is presented at sparsity levels S=0.5, 0.6, 0.7, but the paper does not explain how NAEx is constrained to produce explanations at exactly these sparsity levels. The loss in Eq. 13 contains L1 regularizers with weights lambda1 and lambda2, but no mechanism or tuning procedure is described for achieving a target sparsity. Baselines such as dNAME and GNNExplainer, unless similarly constrained, may be at an unfair disadvantage. Please specify the protocol used to match sparsity across methods, or otherwise justify that the comparison is apples-to-apples.
  5. [§5.4 (Eq. 23)] The FTH definition is technically problematic. For candidates not in the explained subgraph, the paper sets shat_uv = 0 and re-normalizes qhat_u. This can produce zero probabilities, making the KL divergence undefined (log 0). The paper does not specify any smoothing or fallback. Also, computing qu and qhat_u over all |Vt| candidates is prohibitively expensive for large graphs; the paper does not state how this is implemented in practice. Please provide a well-defined, computable formulation and clarify the exact procedure used in the experiments.
minor comments (11)
  1. [§3.1] Typo: 'where where' appears in the sentence defining h_s_u and h_t_v.
  2. [§5.2] Typo: 'mdoels' should be 'models'; also 'Foresquare' is spelled inconsistently with 'Foresq' elsewhere.
  3. [§4.4.3] Typo: 'relevent' should be 'relevant'.
  4. [§6.2] Typo: 'siginificant' should be 'significant'.
  5. [Table 2] The dataset statistics are poorly formatted; for instance 'Foresq - Twitter (F-T)5,313 - 5,120 76,972 - 1,64,919 0 1,609' is hard to parse. Please use a cleaner table layout with clear columns.
  6. [§6.2 and Figure 3] The ablation study reports FID/FTH values only as bars without numerical labels, and the legend is minimal. The text states 'NAEx-3 performs similar' and 'NAEx-4 shows a significant improvement' but the figure does not allow the reader to verify the magnitude of these effects. Please include numerical values or a table.
  7. [§5.3] The list of NA models that NAEx can explain includes many methods without citations in that list; also, the claim that it is 'first general-purpose' framework is strong and would benefit from a more precise comparison with existing approaches in the text, not only in the table.
  8. [§4.4.1] Equation 17 sums over negative samples but does not specify how many negatives are used or how they are sampled. This is important for reproducibility.
  9. [§4.3] Equation 11 uses cross-entropy H(Y, Yhat) but the notation is not explicitly defined; it would help to state that this is the empirical cross-entropy approximated by Monte Carlo samples.
  10. [§5.4 (Eq. 24)] The sparsity metric uses 2-hop neighborhoods as the denominator, but the rationale for choosing 2-hop rather than the full graph or some other local neighborhood is not justified.
  11. [General] The paper repeatedly refers to the Supplementary for Algorithm 1 and other details, but the supplementary material is not included in the submission. Please provide it or incorporate the essential details in the main text.

Circularity Check

1 steps flagged · score 6.0 of 10

Faithfulness metric FTH is defined using the same sigmoid-cosine surrogate (Eqs. 15-16) that the alignment-consistency loss minimizes, making the reported FTH an in-sample measure of the training objective rather than independent evidence of faithfulness to the actual argmax-based NA model.

  1. fitted input called prediction [Sec. 4.4.1 (Eqs. 15-17) and Sec. 5.4 (Eq. 23)]
    "P (Y = 1 | Gs, Gt) ≈ σ(sim(hs_u, ht_v)) = suv (15) ... P ( ˆY = 1 | ˆGs, ˆGt) ≈ σ(sim(ˆhs_u, ˆht_v)) = ˆsuv (16) ... Lalign(u,v) = suv log ˆsuv + Σ_{(u′,v′)̸∈A} su′v′ log ˆsu′v′ . (17) ... FTH = 1/|I| Σ_{(u,v)∈I} exp^{−KL(qu| ˆqu)} (23) where KL(qu| ˆqu) is the KL divergence between qu = [suv1 , suv2 , ...suv|Vt | ] and ˆqu = [ˆsuv1 , ˆsuv2 , ...ˆsuv|Vt | ]."

    The surrogate probabilities s_uv and \hat{s}_uv from Eqs. 15-16 are simultaneously (a) the targets of the alignment-consistency training loss Lalign (Eq. 17, a cross-entropy between s and \hat{s}) and (b) the arguments of the faithfulness metric FTH (Eq. 23, exp(-KL(q||\ hat{q}))). Because KL(q||\hat{q}) = -H(q) + H(q,\hat{q}), and H(q) is fixed for a given anchor/model, maximizing FTH is equivalent to minimizing the same cross-entropy that Lalign minimizes. Thus high FTH on the anchors used for training is an in-sample reflection of the training objective, not independent confirmation that the explanation preserves the original prediction. Moreover, the original model is defined by argmax over candidates (Eq. 1) with possible mutual consistency, while Eqs. 15-16 replace its output by a si

full rationale

NAEx's core mask-learning mechanism is not circular: it derives from a mutual-information objective, uses jointly learned edge/feature masks, and FID (Eq. 22) can be computed from the actual model's predictions on reduced graphs, providing independent signal. No load-bearing self-citation was found; citations to GNNExplainer/PGExplainer are standard prior work, and the authors' own survey/SAlign citations are not used to justify the main derivation. The significant circularity is in the evaluation of faithfulness: FTH (Eq. 23) uses exactly the sigmoid-approximated score distributions s and \ hat{s} (Eqs. 15-16) that the alignment-consistency loss Lalign (Eq. 17) is trained to match. Since FTH is a monotone function of the cross-entropy minimized in Eq. 17, reporting FTH as evidence that explanations are faithful to the original NA model is partly circular: it measures fit to the surrogate, while the paper does not establish that the surrogate preserves the argmax decision rule (Eq. 1) or bidirectional consistency. This affects the central faithfulness claim, but the framework retains independent components (FID, sparsity, inductive generalization, baseline comparisons), so the overall circularity is partial rather than total. Score 6 reflects one central evaluation metric reducing to the training objective by construction.

Assumptions & free parameters 6 free parameters · 4 assumptions · 2 invented entities

The central claim depends mostly on the standard machinery of GNN explainers (concrete relaxation, Monte Carlo cross-entropy, sparsity regularization) and on a few ad hoc modeling choices: the sigmoid proxy for alignment probability, the contrastive subgraph-similarity loss, and undisclosed loss weights and temperatures.

free parameters (6)
  • lambda1, lambda2
    Weights for subgraph similarity and sparsity terms in the loss (Eq. 13); values not disclosed.
  • beta
    Temperature for binary concrete sampling (Eq. 8), chosen by hand; value not disclosed.
  • tau
    Temperature in the contrastive subgraph similarity loss (Eq. 18); value not disclosed.
  • K
    Number of Monte Carlo samples for the expectation in Eq. 10 and Eq. 11; value not disclosed.
  • KF
    Maximum number of retained features in the feature mask, constraining the sum of F (Section 4.2.3); value not disclosed.
  • B = 5 to 100
    Number of anchors selected by Submodular Pick for the transductive evaluation (Section 6.1); varied in experiments.
assumptions (4)
  • domain assumption Edges are conditionally independent in the probabilistic graph model (Eqs. 5 and 6).
    Inherited from PGExplainer [18] to make subgraph sampling tractable; if edge interactions are strongly coupled, the factorized mask may miss joint dependencies.
  • domain assumption The mutual information objective and its cross-entropy surrogate correctly capture explanation fidelity (Eqs. 2, 4, and 11).
    Borrowed from GNNExplainer [35]; its validity for the network alignment setting is assumed.
  • ad hoc to paper The sigmoid of cosine similarity between anchor embeddings is a valid proxy for the NA model's alignment probability (Eqs. 15 and 16).
    The original NA model in Eq. 1 uses argmax over similarities; the sigmoid conversion is an assumption that may not reflect the actual decision rule.
  • ad hoc to paper The contrastive subgraph similarity loss enforces meaningful structural comparison between explanations (Eq. 18).
    The use of tanh-pooled graph embeddings and a contrastive loss is introduced ad hoc to make explanations comparable; its benefit is only shown empirically.
invented entities (2)
  • Shared edge-importance network Psi (MLP across both graphs)
    purpose: Predicts edge relevance scores for inductive explanation generation (Eq. 7).
    Validated only on internal benchmarks; no external falsifiable prediction is made.
  • Learnable feature mask F with Gaussian corruption W
    purpose: Global feature selection for explanations (Eq. 9).
    An internal mechanism of the framework; no independent evidence outside the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of NAEx: A Plug-and-Play Framework for Explaining Network Alignment." pith.science (2026). https://pith.science/paper/CSV37LJX

@misc{pith2026250804731,
  author       = {Pith},
  title        = {Pith review of: NAEx: A Plug-and-Play Framework for Explaining Network Alignment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CSV37LJX}},
  note         = {Machine review of arXiv:2508.04731}
}
read the original abstract

Network alignment (NA) identifies corresponding nodes across multiple networks, with applications in domains like social networks, co-authorship, and biology. Despite advances in alignment models, their interpretability remains limited, making it difficult to understand alignment decisions and posing challenges in building trust, particularly in high-stakes domains. To address this, we introduce NAEx, a plug-and-play, model-agnostic framework that explains alignment models by identifying key subgraphs and features influencing predictions. NAEx addresses the key challenge of preserving the joint cross-network dependencies on alignment decisions by: (1) jointly parameterizing graph structures and feature spaces through learnable edge and feature masks, and (2) introducing an optimization objective that ensures explanations are both faithful to the original predictions and enable meaningful comparisons of structural and feature-based similarities between networks. NAEx is an inductive framework that efficiently generates NA explanations for previously unseen data. We introduce evaluation metrics tailored to alignment explainability and demonstrate NAEx's effectiveness and efficiency on benchmark datasets by integrating it with four representative NA models.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

43 extracted references · 42 canonical work pages

  1. [1]

    Agarwal, O

    C. Agarwal, O. Queen, H. Lakkaraju, and M. Zitnik. Evaluating ex- plainability for graph neural networks.Scientific Data, 10(1):144, 2023

    work page 2023

  2. [2]

    Carmagnola and F

    F. Carmagnola and F. Cena. User identification for cross-system per- sonalisation. Information Sciences, 179(1):16–32, 2009

    work page 2009

  3. [3]

    J. Chen, K. Amara, J. Yu, and R. Ying. Generative explanations for graph neural network: Methods and evaluations. arXiv preprint arXiv:2311.05764, 2023

    work page Pith review arXiv 2023

  4. [4]

    Conte, P

    D. Conte, P. Foggia, C. Sansone, and M. Vento. Graph matching appli- cations in pattern recognition and image processing. In International Conference on Image Processing, ICIP, pages 21–24. IEEE, 2003

    work page 2003

  5. [5]

    V . P. Dwivedi, C. K. Joshi, A. T. Luu, T. Laurent, Y . Bengio, and X. Bresson. Benchmarking graph neural networks. Journal of Machine Learning Research, 24(43):1–48, 2023

    work page 2023

  6. [6]

    Heimann, H

    M. Heimann, H. Shen, T. Safavi, and D. Koutra. Regal: Representation learning-based graph alignment. In Proceedings of the 27th ACM inter- national conference on information and knowledge management, pages 117–126, 2018

    work page 2018

  7. [7]

    Huang, M

    Q. Huang, M. Yamada, Y . Tian, D. Singh, and Y . Chang. Graphlime: Local interpretable model explanations for graph neural networks.IEEE Transactions on Knowledge and Data Engineering , 35(7):6968–6972, 2022

    work page 2022

  8. [8]

    T. T. Huynh, C. T. Duong, T. T. Nguyen, V . T. Van, A. Sattar, H. Yin, and Q. V . H. Nguyen. Network alignment with holistic embeddings. IEEE Transactions on Knowledge and Data Engineering , 35(2):1881–1894, 2021

    work page 2021

Show all 43 references
  1. [9]

    Kakkad, J

    J. Kakkad, J. Jannu, K. Sharma, C. Aggarwal, and S. Medya. A survey on explainability of graph neural networks. arXiv preprint arXiv:2306.01958, 2023

    2023 arXiv

  2. [10]

    Khan and E

    A. Khan and E. B. Mobaraki. Interpretability methods for graph neural networks. In 2023 IEEE 10th International Conference on Data Science and Advanced Analytics (DSAA), pages 1–4. IEEE, 2023

    2023

  3. [11]

    Koutra, H

    D. Koutra, H. Tong, and D. Lubensky. Big-align: Fast bipartite graph alignment. In 2013 IEEE 13th international conference on data mining, pages 389–398. IEEE, 2013

    2013

  4. [12]

    Kuchaiev and N

    O. Kuchaiev and N. Pržulj. Integrative network alignment reveals large regions of global network similarity in yeast and human. Bioinformat- ics, 27(10):1390–1396, 2011

    2011

  5. [13]

    G. Li, M. Müller, B. Ghanem, and V . Koltun. Training graph neural networks with 1000 layers. In International conference on machine learning, pages 6437–6449. PMLR, 2021

    2021

  6. [14]

    Y . Li, H. Cai, and H. Liu. Attention-based multi-layer network repre- sentation learning framework for network alignment. Information Pro- cessing & Management, 62(3):104009, 2025

    2025

  7. [15]

    Liang, Y

    Z. Liang, Y . Rong, C. Li, Y . Zhang, Y . Huang, T. Xu, X. Ding, and J. Huang. Unsupervised large-scale social network alignment via cross network embedding. In Proceedings of the 30th ACM International Conference on Information & Knowledge Management , pages 1008– 1017, 2021

    2021

  8. [16]

    L. Liu, W. Zuo, and T. Peng. Detecting outlier pairs in complex network based on link structure and semantic relationship.Expert Syst. Appl., 69: 40–49, 2017

    2017

  9. [17]

    L. Liu, B. Qu, B. Chen, A. Hanjalic, and H. Wang. Modelling of infor- mation diffusion on social networks with applications to wechat. Phys- ica A: Statistical Mechanics and its Applications, 496:318–329, 2018

    2018

  10. [18]

    D. Luo, W. Cheng, D. Xu, W. Yu, B. Zong, H. Chen, and X. Zhang. Parameterized explainer for graph neural network. In Proceedings of the 34th International Conference on Neural Information Processing Systems, NIPS ’20, Red Hook, NY , USA, 2020. Curran Associates Inc. ISBN 9781713829546

    2020

  11. [19]

    J.-D. Park, C. Tran, W.-Y . Shin, and X. Cao. Gradalign+: Empower- ing gradual network alignment using attribute augmentation. In Pro- ceedings of the 31st ACM international conference on information & knowledge management, pages 4374–4378, 2022

    2022

  12. [20]

    J. Peng, F. Xiong, S. Pan, L. Wang, and X. Xiong. Robust network alignment with the combination of structure and attribute embeddings. In 2023 IEEE International Conference on Data Mining (ICDM), pages 498–507. IEEE, 2023

    2023

  13. [21]

    why should i trust you?

    M. T. Ribeiro, S. Singh, and C. Guestrin. " why should i trust you?" explaining the predictions of any classifier. In Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining, pages 1135–1144, 2016

    2016

  14. [22]

    Saxena and J

    S. Saxena and J. Chandra. Salign: A graph neural attention framework for aligning structurally heterogeneous networks. Journal of Artificial Intelligence Research, 77:949–969, 2023

    2023

  15. [23]

    Saxena and J

    S. Saxena and J. Chandra. A survey on network alignment: approaches, applications and future directions. In Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence , pages 8216– 8224, 2024

    2024

  16. [24]

    Seo, J.-H

    D.-H. Seo, J.-H. Lim, W.-Y . Shin, and S.-W. Kim. Leveraging trustwor- thy node attributes for effective network alignment. In Proceedings of the 33rd ACM International Conference on Information and Knowledge Management, pages 2004–2013, 2024

    2004

  17. [25]

    Q. Sun, X. Lin, Y . Zhang, W. Zhang, and C. Chen. Towards higher-order topological consistency for unsupervised network alignment. In 2023 IEEE 39th International Conference on Data Engineering (ICDE) , pages 177–190. IEEE, 2023

    2023

  18. [26]

    R. Tang, Z. Yong, S. Jiang, X. Chen, Y . Liu, Y .-C. Zhang, G.-Q. Sun, and W. Wang. Network alignment. Physics Reports, 1107:1–45, 2025

    2025

  19. [27]

    W. Tang, F. Su, H. Sun, Q. Qi, J. Wang, S. Tao, and H. Yang. Weakly su- pervised entity alignment with positional inspiration. In Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining, WSDM ’23, page 814–822, New York, NY , USA, 2023. Asso- ci...

    2023

  20. [28]

    X. Tian, Z. Sun, and W. Hu. Generating explanations to understand and repair embedding-based entity alignment. In 2024 IEEE 40th Inter- national Conference on Data Engineering (ICDE) , pages 2205–2217. IEEE, 2024

    2024

  21. [29]

    B. D. Trisedya, F. D. Salim, J. Chan, D. Spina, F. Scholer, and M. Sanderson. i-align: an interpretable knowledge graph alignment model. Data Mining and Knowledge Discovery , 37(6):2494–2516, 2023

    2023

  22. [30]

    H. T. Trung, N. T. Toan, T. Van Vinh, H. T. Dat, D. C. Thang, N. Q. V . Hung, and A. Sattar. A comparative study on network alignment tech- niques. Expert Systems with Applications, 140:112883, 2020

    2020

  23. [31]

    H. T. Trung, T. Van Vinh, N. T. Tam, H. Yin, M. Weidlich, and N. Q. V . Hung. Adaptive network alignment with unsupervised and multi-order convolutional networks. In 2020 IEEE 36th International Conference on Data Engineering (ICDE), pages 85–96. IEEE, 2020

    2020

  24. [32]

    Velickovic, G

    P. Velickovic, G. Cucurull, A. Casanova, A. Romero, P. Lio, Y . Bengio, et al. Graph attention networks. stat, 1050(20):10–48550, 2017

    2017

  25. [33]

    Z. Yan, L. Liu, X. Li, W. K. Cheung, Y . Zhang, Q. Liu, and G. Wang. To- wards improving embedding based models of social network alignment via pseudo anchors. IEEE Trans. Knowl. Data Eng., 35(4):4307–4320, 2023

    2023

  26. [34]

    J. Yeo, H. Park, S. Lee, E. W. Lee, and S.-w. Hwang. Xina: explainable instance alignment using dominance relationship. IEEE Transactions on Knowledge and Data Engineering, 32(2):388–401, 2018

    2018

  27. [35]

    Z. Ying, D. Bourgeois, J. You, M. Zitnik, and J. Leskovec. Gnnex- plainer: Generating explanations for graph neural networks. In Ad- vances in Neural Information Processing Systems (NeurIPS) , pages 9240–9251, 2019

    2019

  28. [36]

    H. Yuan, J. Tang, X. Hu, and S. Ji. Xgnn: Towards model-level ex- planations of graph neural networks. In Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data min- ing, pages 430–438, 2020

    2020

  29. [37]

    H. Yuan, H. Yu, S. Gui, and S. Ji. Explainability in graph neural net- works: A taxonomic survey. IEEE Trans. Pattern Anal. Mach. Intell. , 45(5):5782–5799, 2023

    2023

  30. [38]

    Zhang and H

    S. Zhang and H. Tong. Final: Fast attributed network alignment. In Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining, pages 1345–1354, 2016

    2016

  31. [39]

    Zhang, H

    S. Zhang, H. Tong, L. Jin, Y . Xia, and Y . Guo. Balancing consistency and disparity in network alignment. In Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining , pages 2212–2222, 2021

    2021

  32. [40]

    Zhang, S

    Z. Zhang, S. Gao, S. Su, L. Sun, and R. Chen. Mining: Multi-granularity network alignment based on contrastive learning.IEEE Transactions on Knowledge and Data Engineering, 2023

    2023

  33. [41]

    Zheng, L

    C. Zheng, L. Pan, and P. Wu. Jora: Weakly supervised user identity linkage via jointly learning to represent and align. IEEE transactions on neural networks and learning systems, 35(3):3900–3911, 2022

    2022

  34. [42]

    F. Zhou, Z. Wen, G. Trajcevski, K. Zhang, T. Zhong, and F. Liu. Disen- tangled network alignment with matching explainability. In IEEE IN- FOCOM 2019-IEEE conference on computer communications , pages 1360–1368. IEEE, 2019

    2019

  35. [43]

    Y . Zhou, J. Ren, R. Jin, Z. Zhang, J. Zheng, Z. Jiang, D. Yan, and D. Dou. Unsupervised adversarial network alignment with reinforce- ment learning. ACM Trans. Knowl. Discov. Data, 16(3), 2021

    2021

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.