Pith. sign in

REVIEW 4 major objections 8 minor 81 references

Reconciling Attribute and Structural Anomalies for Improved Graph Anomaly Detection

T0 review · 4 major / 8 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read TripleAD separates graph anomaly detection into attribute, structure, and mixed channels and reports higher scores than 14 baselines on all five datasets.

desk verdict TripleAD is a sensible divide-and-conquer framework for unsupervised graph anomaly detection, but its novel attribute-mixed curvature is formally under-specified (Eq. 22 can yield negative probabilities) and the reported gains are small with no variance. read the letter →

arxiv 2506.23469 v1 pith:4NVLFNX5 submitted 2025-06-30 cs.LG cs.SI

Chunjing Xiao , Jiahui Lu , Xovee Xu , Fan Zhou , Tianshu Xie , Wei Lu , Lifeng Xu This is my paper · ORCID
classification cs.LGcs.SI
keywords graphanomalydetectionTripleADattributeanomaliesstructuralmixedattribute-mixedcurvaturemutualdistillationneuralnetworks
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

This paper argues that the usual practice of training one graph neural network to catch every kind of anomaly is self-defeating: attribute anomalies and structural anomalies pull the optimization in opposite directions, a 'tug-of-war' that degrades both. To get around it, TripleAD splits the task into three channels, one for attribute, one for structural, and one for mixed anomalies, each with its own reconstruction objective, and then lets the channels exchange knowledge through a mutual-distillation loss. The framework also introduces an attribute-mixed curvature that folds node attributes into a structural curvature score as the reconstruction target for the mixed channel. If the paper's account holds, graph anomaly detection in fraud, healthcare, and social-media settings would have a concrete recipe for treating the three anomaly types separately instead of forcing one model to do everything.

What carries the argument

The load-bearing mechanism is the attribute-mixed curvature $\kappa(i,j)$ of Eqs. (21)-(23), a per-edge score that combines Wasserstein graph curvature with normalized attribute similarity and is used in place of the adjacency matrix for a GCN-based reconstruction. It carries the mixed-anomaly channel: normal-normal pairs are expected to score higher than normal-abnormal pairs, and the reconstruction error of this curvature becomes the third anomaly signal. The other two mechanisms are the multi-scale attribute estimator, which forms an attention-weighted combination of $L$ propagated feature views to avoid over-smoothing, and the link-enhanced structure estimator, which builds a $k$-NN graph so that edge-masked nodes still receive information; a triplet distillation loss ties all three channels together.

What would settle it

Compute the distribution of attribute-mixed curvature values for normal-normal versus normal-abnormal node pairs on Amazon, YelpChi, ACM, and Flickr; if the curves overlap or invert on any of them, the mixed channel's reconstruction error carries no signal and the combined anomaly score of Eq. (34) should lose accuracy.

Watch

Extended reading notes

Core claim

TripleAD's central claim is that attribute anomalies, structural anomalies, and mixed anomalies are best detected by three separate estimation modules whose losses are optimized sequentially rather than jointly, with inter-channel knowledge transfer handled by a triplet-loss distillation. The attribute channel masks a node's attributes and reconstructs them from attention-weighted multi-scale propagated neighbor features; the structure channel masks edges and reconstructs the adjacency matrix with the help of a kNN-enhanced graph that keeps masked nodes reachable; the mixed channel replaces the adjacency matrix with an attribute-mixed curvature and reconstructs that curvature through a GCN. The final anomaly score is a weighted sum of the three reconstruction errors. The paper reports that this architecture outperforms 14 baselines on all five datasets in every metric, with relative improvements of 0.05% to 2.64% over the strongest baseline, and that ablations removing any channel or the distillation step degrade performance.

Load-bearing premise

The mixed channel's signal rests on the assumption that its attribute-mixed curvature score is always lower for abnormal node pairs than for normal ones, yet the paper shows this separation only on CiteSeer and never verifies it on the other four datasets.

Editorial extensions

If this is right

  • Graph anomaly detectors no longer need to compromise between attribute fidelity and structural fidelity; each node can be scored on three separate axes and the scores combined with weights.
  • The multi-scale attribute module supplies a way to reach long-range neighborhood information for reconstruction without stacking deep GNN layers, avoiding over-smoothing.
  • The link-enhanced structure module shows that edge masking need not isolate a node, since a kNN graph built on propagated features keeps information flowing to the target.
  • The attribute-mixed curvature turns both attributes and topology into a single per-edge reconstruction target, so mixed anomalies fall under the same reconstruction-error logic as the other two types.
  • According to the ablations, removing any channel or the distillation step lowers performance, so the reported gain comes from the combination of all three channels plus mutual distillation.

Reading between the lines

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

  • The curvature gap that motivates the mixed channel is shown only on CiteSeer; measuring the same gap on Amazon, YelpChi, ACM, and Flickr would directly test whether the signal transfers.
  • The fixed training order of attribute, then structure, then mixture, is not varied in the paper, so whether the claimed interference relief depends on this curriculum is an open question.
  • The divide-and-conquer-plus-distillation pattern could plausibly carry over to other graph tasks with conflicting objectives, such as joint link prediction and node classification, though TripleAD itself is evaluated only for anomaly detection.
  • The curvature-weighted adjacency matrix and the kNN-enhanced graph are computed as preprocessing and could be reused as input signals for other detectors, an extension the paper does not explore.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 8 minor

Summary. The paper proposes TripleAD, an unsupervised graph anomaly detection framework that decomposes the task into three channels for attribute, structural, and mixed anomalies. The attribute channel uses multi-scale feature propagation to reconstruct masked attributes; the structure channel builds a kNN link-enhanced graph and reconstructs masked edges; the mixed channel introduces an attribute-mixed curvature, computed from a modified Wasserstein distance between local distributions, and reconstructs this curvature using a GCN. The three channels are connected by a mutual distillation strategy based on triplet losses. Experiments on five datasets (Amazon, YelpChi, CiteSeer, ACM, Flickr) report that TripleAD outperforms 15 baselines on AUC-ROC, AUC-PR, and F1 metrics, with relative improvements between 0.05% and 2.64%. The paper also presents ablation and sensitivity analyses to support the contribution of each module.

Significance. If the claims hold, TripleAD would be a useful contribution by showing that explicitly separating anomaly types can reduce interference, and the mutual distillation mechanism provides a way to combine specialized representations. The paper includes extensive comparisons against 15 baselines, ablation studies, and sensitivity analyses, which are strengths. However, the core novelty—attribute-mixed curvature—has a definitional gap, and the empirical evidence for the central 'tug-of-war' claim is indirect because no single-model variant with comparable capacity is tested. The reported improvements are modest and lack statistical validation.

major comments (4)
  1. [III-D, Eq. (22)] The distribution \hat{m}_i(j) in Eq. (22) is not guaranteed to be a valid probability distribution. The entry for x ∈ N_i − N_j equals (1 − δ − S'_ij/|N_i − N_j|)/k_i, which is negative whenever S_ij > (1−δ)^2 |N_i − N_j|. For δ = 0.5 and a pair with only one non-common neighbor, any attribute similarity above 0.25 yields a negative mass. Since S_ij is never defined in the paper, δ is not listed in the hyperparameter search space (Table II), and no clipping or renormalization is mentioned, the Wasserstein distance in Eq. (21) is undefined for many node pairs. Consequently, the attribute-mixed curvature κ(i,j), and hence the entire mixed channel (Eqs. (23)–(26) and (33)–(34)), is not well-defined for these pairs. The authors should either constrain S_ij, choose δ appropriately, or clip/renormalize the distribution, and re-evaluate the experiments.
  2. [IV-D, Unified Versus Separate] The paper uses DOMINANT and CoLA to illustrate the 'tug-of-war' but does not compare TripleAD with a single unified model of comparable capacity trained to reconstruct both attributes and structure simultaneously. The ablation study in §IV-C removes individual components but always retains three channels. Therefore, the observed gains in Table III are not directly attributable to the separation of anomaly types; they could stem from increased model capacity, the link-enhanced graph, or the distillation objective. A single-channel variant with the same total parameter budget should be tested to support the central claim.
  3. [IV-B, Table III] No standard deviations, number of independent runs, or significance tests are reported for any of the results in Table III. Several relative improvements are very small (e.g., 0.05% on Flickr APR), so without statistical evidence the claim of 'superior detection performance across all evaluated datasets' is not supported. The authors should run multiple seeds and report mean ± std or a significance test, especially for the smallest improvements.
  4. [II-B and III-D, Fig. 2] The discriminative power of the attribute-mixed curvature is demonstrated on a single dataset (CiteSeer) via Fig. 2. The paper does not show analogous curvature distributions for Amazon, YelpChi, ACM, or Flickr, yet the mixed-channel score is used on all five datasets. If the separation between normal-normal and abnormal-normal pairs does not hold on a given graph, AS_mix in Eq. (34) adds noise rather than signal. The authors should provide per-dataset evidence or an analysis of when the curvature assumption holds.
minor comments (8)
  1. [III-D, Eq. (22)] The attribute similarity S_ij is used in Eq. (22) but never defined; please specify it (e.g., cosine similarity after normalization).
  2. [III-B and III-C] The masking ratios and strategies for attributes and edges are not specified; please state them (e.g., random 20% masking) to make the reconstruction setup reproducible.
  3. [Table II] The search space in Table II omits several hyperparameters that are tuned in the experiments, including k, γ, η1, η2, m, δ, and the GCN layer count in the mixture module; please list these or state that they are set by grid search elsewhere.
  4. [III-D, Eq. (24)] The symbol L is used both for the number of propagation scales and for the matrix in Eq. (24); please use different symbols to avoid confusion.
  5. [III-D, Eq. (24)] Calling L = \tilde{D}^{-1/2} C \tilde{D}^{-1/2} a 'Laplacian' is misleading because C contains nonnegative entries; it is a normalized adjacency matrix with self-loops.
  6. [References] Reference [56] is missing the article title, and reference [57] appears incomplete; please supply full bibliographic information.
  7. [IV-B] The statement 'In addition, we calculate the runtime of our model...' is not backed by any reported runtime results; please include the results or remove the claim.
  8. [Index Terms] There is a typo: 'graph neutral network' should be 'graph neural network.'

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: TripleAD's predictions are reconstruction errors on external benchmarks, not fitted inputs or imported self-cited constraints.

full rationale

TripleAD's central claim—that three separately optimized channels reduce attribute/structure interference and improve detection—is tested against 14 baselines on five public datasets (Table III), with ablations (Fig. 4) and a unified-versus-separate comparison (Fig. 5). The reconstruction targets X, A, and C are computed directly from the input graph; no anomaly label enters the curvature definition or the training objectives, and hyperparameters are grid-searched on a validation split (Table II). The final anomaly score in Eq. (34) is a weighted sum of three reconstruction errors, not a fitted parameter renamed as a prediction. The paper contains a few self-citations (e.g., [10], [35], [60]) but only for contextual purposes—graph-anomaly applications, a standard FC decoder design, and graph-level anomaly detection—so none is load-bearing for the central derivation. The curvature motivation in Section II-B uses CiteSeer to illustrate separation while CiteSeer also appears in the test set; this is a possible design-choice/test-set overlap and deserves methodological scrutiny, but it is not circular because the curvature formula is fixed and unlabeled. Likewise, the negative-mass issue in Eq. (22) is a mathematical correctness risk, not a circularity: the Wasserstein distance may be undefined for some pairs, but repairing or constraining the distributions would not make the model's predictions equivalent to its inputs by construction. No self-definitional, fitted-input-as-prediction, imported-uniqueness, or ansatz-smuggling step was found.

Assumptions & free parameters 13 free parameters · 5 assumptions · 1 invented entities

The method is a machine learning framework whose performance rests on several domain assumptions about curvature distributions, the reconstruction-error principle, and the separability of anomaly types, plus a large set of tuned hyperparameters. The only invented entity is the attribute-mixed curvature, which lacks independent evidence beyond a single visualization.

free parameters (13)
  • Number of propagation scales L = Grid-searched over {1..8}, optimal not reported
    Controls the number of scales in the multi-scale attribute propagation (Eq. 3); tuned on the validation split.
  • Propagation iterations T = Grid-searched over {1..8}, optimal not reported
    Number of diffusion steps in the structure module (Eqs. 12-14); tuned on validation.
  • Restart probabilities alpha and beta = Grid-searched over {0.001, 0.05, 0.1, 0.2}
    Balance local and global information in attribute and structure propagation; tuned on validation.
  • Hidden dimension h = Grid-searched over {32, 64, 128, 256}
    Dimension of node representations in all modules; tuned on validation.
  • Learning rate = Grid-searched over {0.1, 0.01, 0.001, 0.0001}
    Tuned on validation.
  • Anomaly score weights lambda1, lambda2, lambda3 = Grid-searched over {0.1, 0.2, ..., 0.9}
    Weights in Eq. (34) that combine the three reconstruction errors; tuned on validation.
  • kNN neighbor number k = Tuned, optimal not reported
    Number of neighbors in the link-enhanced graph (Eq. 11); sensitivity shown in Fig. 7.
  • Consistency weight gamma = Tuned, optimal not reported
    Weight of the consistency loss in Eq. (20); sensitivity shown in Fig. 7.
  • Distillation weights eta1 and eta2 = Grid-searched over {0.1..0.9}
    Balance reconstruction and distillation losses in Eqs. (28)-(30); tuned on validation.
  • Triplet margin m = Grid-searched over {0.1, 0.3, 0.5, 1, 1.5}
    Margin in the triplet loss (Eq. 27); sensitivity shown in Fig. 9.
  • Curvature distribution coefficient delta = Not reported
    Coefficient in the probability distribution used to compute attribute-mixed curvature (Eq. 22); chosen by hand, not included in the search table.
  • GCN layer count in mixture module = Not reported
    The paper says 'multiple GCN layers' in Eq. (24) but does not specify the number; a hidden tuning choice.
  • Masking ratio for attributes and edges = Not reported
    The degree of attribute and edge masking in training is never quantified; another unstated tuning choice.
assumptions (5)
  • domain assumption Graph curvature defined by Eq. (1) with the distribution in Eq. (2) is a meaningful measure of interaction and overlap strength.
    Invoked in Section II-B to justify curvature-based anomaly detection; based on prior curvature GNN literature [28]-[31].
  • domain assumption Normal-normal node pairs have higher curvature than normal-abnormal pairs on unseen graphs.
    Assumed throughout the mixed channel; only direct evidence is the CiteSeer visualization in Fig. 2.
  • domain assumption Anomalies are hard to reconstruct by a model trained on the majority normal pattern.
    Standard reconstruction-based anomaly detection assumption used in Eqs. (10), (18), (26), and in the scoring of Section III-F.
  • ad hoc to paper Attribute, structural, and mixed anomalies are separable categories that can be detected independently and then combined.
    The core design premise of the triple-channel framework; motivated by the tug-of-war argument in Section I but not proven.
  • domain assumption The kNN link-enhanced graph built from propagated features preserves the information needed to reconstruct isolated nodes.
    Underpins the structure channel (Eqs. 11-16); assumes semantic similarity approximates structural relevance.
invented entities (1)
  • Attribute-mixed curvature kappa(i,j)
    purpose: A node-pair indicator combining attribute similarity and structural overlap to detect mixed anomalies through reconstruction error.
    Defined in Eqs. (21)-(23). The only independent-looking evidence is the CiteSeer distribution in Fig. 2, which comes from the same paper; no external dataset, theoretical bound, or falsifiable prediction outside the paper is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Reconciling Attribute and Structural Anomalies for Improved Graph Anomaly Detection." pith.science (2026). https://pith.science/paper/4NVLFNX5

@misc{pith2026250623469,
  author       = {Pith},
  title        = {Pith review of: Reconciling Attribute and Structural Anomalies for Improved Graph Anomaly Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4NVLFNX5}},
  note         = {Machine review of arXiv:2506.23469}
}
read the original abstract

Graph anomaly detection is critical in domains such as healthcare and economics, where identifying deviations can prevent substantial losses. Existing unsupervised approaches strive to learn a single model capable of detecting both attribute and structural anomalies. However, they confront the tug-of-war problem between two distinct types of anomalies, resulting in suboptimal performance. This work presents TripleAD, a mutual distillation-based triple-channel graph anomaly detection framework. It includes three estimation modules to identify the attribute, structural, and mixed anomalies while mitigating the interference between different types of anomalies. In the first channel, we design a multiscale attribute estimation module to capture extensive node interactions and ameliorate the over-smoothing issue. To better identify structural anomalies, we introduce a link-enhanced structure estimation module in the second channel that facilitates information flow to topologically isolated nodes. The third channel is powered by an attribute-mixed curvature, a new indicator that encapsulates both attribute and structural information for discriminating mixed anomalies. Moreover, a mutual distillation strategy is introduced to encourage communication and collaboration between the three channels. Extensive experiments demonstrate the effectiveness of the proposed TripleAD model against strong baselines.

Figures

Figures reproduced from arXiv: 2506.23469 by the authors.

Figure 1
Figure 1. The tug-of-war problem between different anomalies. (a) A single [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Curvature Distributions on CiteSeer. Normal-normal pairs (left) [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Overview of the TripleAD framework. The three channels, the multi [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Performance with different components. AmazonYelpChi CiteSeer ACM Flickr DOMINANT 0.2 0.4 0.6 0.8 1.0 t The unified scheme testtet The separate scheme t AmazonYelpChi CiteSeer ACM Flickr CoLA AmazonYelpChi CiteSeer ACM Flickr TripleAD [PITH_FULL_IMAGE:figures/full_fig…
Figure 5
Figure 5. Figure 5: F1-scores using the unified scheme and separate scheme. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 7
Figure 7. Figure 7: Sensitivity analysis for k and γ (AUC). Balancing Parameters η1 and η2. We here discuss the parameters η1 and η2 in the loss function (Eqs. (28), (29) and (30)). As shown in [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: The effects of balancing parameters η1 and η2 on the five datasets in terms of AUC. Amazon YelpChi CiteSeer ACM Flickr 20 40 60 80 100 m = 0.1 m = 0.3 m = 0.5 m = 1 m = 1.5 [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 9
Figure 9. Figure 9: Hyper-parameter m w.r.t. AUC. models (LLMs) [22], [51]. They decompose the language model into several functions, where each “expert” learns on a different input space, effectively enhancing the performance of downstream tasks. For vision learning, the image is parsed …
Figure 10
Figure 10. Figure 10: Relation of the iterative propagation in GNNs and the multi-scale [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

81 extracted references · 78 canonical work pages

  1. [1]

    Generalized video anomaly event detection: Systematic taxonomy and comparison of deep models,

    Y . Liu, D. Yang, Y . Wang, J. Liu, J. Liu, A. Boukerche, P. Sun, and L. Song, “Generalized video anomaly event detection: Systematic taxonomy and comparison of deep models,” ACM Computing Surveys , vol. 56, no. 7, pp. 1–38, 2024

    2024

  2. [2]

    Self-supervised learning for time series analysis: Taxonomy, progress, and prospects,

    K. Zhang, Q. Wen, C. Zhang, R. Cai, M. Jin, Y . Liu, J. Y . Zhang, Y . Liang, G. Pang, D. Song et al. , “Self-supervised learning for time series analysis: Taxonomy, progress, and prospects,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 46, no. 10, pp. 6775– 6794, 2024

    work page 2024

  3. [3]

    Fighting against organized fraudsters using risk diffusion-based parallel graph neural network

    J. Ma, F. Li, R. Zhang, Z. Xu, D. Cheng, Y . Ouyang, R. Zhao, J. Zheng, Y . Zheng, and C. Jiang, “Fighting against organized fraudsters using risk diffusion-based parallel graph neural network.” in International Joint Conference on Artificial Intelligence , 2023, pp. 6138–6146

    work page 2023

  4. [4]

    A comprehensive survey on graph neural networks,

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, and S. Y . Philip, “A comprehensive survey on graph neural networks,” IEEE Transactions on Neural Networks and Learning Systems , vol. 32, no. 1, pp. 4–24, 2020

    work page 2020

  5. [5]

    Comga: Community-aware attributed graph anomaly detection,

    X. Luo, J. Wu, A. Beheshti, J. Yang, X. Zhang, Y . Wang, and S. Xue, “Comga: Community-aware attributed graph anomaly detection,” in ACM International Conference on Web Search and Data Mining , 2022, pp. 657–665

    work page 2022

  6. [6]

    Interleaved sequence rnns for fraud detection,

    B. Branco, P. Abreu, A. S. Gomes, M. S. Almeida, J. T. Ascens ˜ao, and P. Bizarro, “Interleaved sequence rnns for fraud detection,” in Proceedings of the ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2020, pp. 3101–3109

    work page 2020

  7. [7]

    Fang: Leverag- ing social context for fake news detection using graph representation,

    V .-H. Nguyen, K. Sugiyama, P. Nakov, and M.-Y . Kan, “Fang: Leverag- ing social context for fake news detection using graph representation,” in ACM International Conference on Information & Knowledge Man- agement, 2020, pp. 1165–1174

    work page 2020

  8. [8]

    Netwalk: A flexible deep embedding approach for anomaly detection in dynamic networks,

    W. Yu, W. Cheng, C. C. Aggarwal, K. Zhang, H. Chen, and W. Wang, “Netwalk: A flexible deep embedding approach for anomaly detection in dynamic networks,” in Proceedings of the ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2018, pp. 2672–2681

    work page 2018

Show all 81 references
  1. [9]

    Bond: Benchmarking unsupervised outlier node detection on static attributed graphs,

    K. Liu, Y . Dou, Y . Zhao, X. Ding, X. Hu, R. Zhang, K. Ding, C. Chen, H. Peng, K. Shu et al., “Bond: Benchmarking unsupervised outlier node detection on static attributed graphs,” in Advances in Neural Information Processing Systems, vol. 35, 2022, pp. 27 021–27 035

    2022

  2. [10]

    Counter- factual data augmentation with denoising diffusion for graph anomaly detection,

    C. Xiao, S. Pang, X. Xu, X. Li, G. Trajcevski, and F. Zhou, “Counter- factual data augmentation with denoising diffusion for graph anomaly detection,” IEEE Transactions on Computational Social Systems, vol. 11, no. 6, pp. 7555–7567, 2024

    2024

  3. [11]

    Radar: Residual analysis for anomaly detection in attributed networks,

    J. Li, H. Dani, X. Hu, and H. Liu, “Radar: Residual analysis for anomaly detection in attributed networks,” in International Joint Conference on Artificial Intelligence, 2017, pp. 2152–2158

    2017

  4. [12]

    Mixedad: A scalable algorithm for detecting mixed anomalies in attributed graphs,

    M. Zhu and H. Zhu, “Mixedad: A scalable algorithm for detecting mixed anomalies in attributed graphs,” in AAAI Conference on Artificial Intelligence, 2020, pp. 1274–1281

    2020

  5. [13]

    Anomaly de- tection on attributed networks via contrastive self-supervised learning,

    Y . Liu, Z. Li, S. Pan, C. Gong, C. Zhou, and G. Karypis, “Anomaly de- tection on attributed networks via contrastive self-supervised learning,” IEEE Transactions on Neural Networks and Learning Systems , vol. 33, no. 6, pp. 2378–2392, 2021

    2021

  6. [14]

    A comprehensive survey on graph anomaly detection with deep learning,

    X. Ma, J. Wu, S. Xue, J. Yang, C. Zhou, Q. Z. Sheng, H. Xiong, and L. Akoglu, “A comprehensive survey on graph anomaly detection with deep learning,” IEEE Transactions on Knowledge and Data Engineering, vol. 35, no. 12, pp. 12 012–12 038, 2021

    2021

  7. [15]

    Deep anomaly detection on attributed networks,

    K. Ding, J. Li, R. Bhanushali, and H. Liu, “Deep anomaly detection on attributed networks,” in SIAM International Conference on Data Mining, 2019, pp. 594–602

    2019

  8. [16]

    Hybrid-order anomaly detection on attributed networks,

    L. Huang, Y . Zhu, Y . Gao, T. Liu, C. Chang, C. Liu, Y . Tang, and C.-D. Wang, “Hybrid-order anomaly detection on attributed networks,” IEEE Transactions on Knowledge and Data Engineering , vol. 35, no. 12, pp. 12 249–12 263, 2021

    2021

  9. [17]

    A deep multi- view framework for anomaly detection on attributed networks,

    Z. Peng, M. Luo, J. Li, L. Xue, and Q. Zheng, “A deep multi- view framework for anomaly detection on attributed networks,” IEEE Transactions on Knowledge and Data Engineering , vol. 34, no. 6, pp. 2539–2552, 2020

    2020

  10. [18]

    Generative and contrastive self-supervised learning for graph anomaly detection,

    Y . Zheng, M. Jin, Y . Liu, L. Chi, K. T. Phan, and Y .-P. P. Chen, “Generative and contrastive self-supervised learning for graph anomaly detection,” IEEE Transactions on Knowledge and Data Engineering , vol. 35, no. 12, pp. 12 220 – 12 233, 2021

    2021

  11. [19]

    Reconstruction enhanced multi-view contrastive learning for anomaly detection on attributed networks,

    J. Zhang, S. Wang, and S. Chen, “Reconstruction enhanced multi-view contrastive learning for anomaly detection on attributed networks,” in International Joint Conference on Artificial Intelligence, 2022, pp. 2376– 2382

    2022

  12. [20]

    Graph anomaly detection via multi-scale contrastive learning networks with augmented view,

    J. Duan, S. Wang, P. Zhang, E. Zhu, J. Hu, H. Jin, Y . Liu, and Z. Dong, “Graph anomaly detection via multi-scale contrastive learning networks with augmented view,” in AAAI Conference on Artificial Intelligence , 2023, pp. 7459–7467

    2023

  13. [21]

    Embracing change: Continual learning in deep neural networks,

    R. Hadsell, D. Rao, A. A. Rusu, and R. Pascanu, “Embracing change: Continual learning in deep neural networks,” Trends in cognitive sci- ences, vol. 24, no. 12, pp. 1028–1040, 2020

    2020

  14. [22]

    Octavius: Mitigating task inter- ference in mllms via moe,

    Z. Chen, Z. Wang, Z. Wang, H. Liu, Z. Yin, S. Liu, L. Sheng, W. Ouyang, Y . Qiao, and J. Shao, “Octavius: Mitigating task inter- ference in mllms via moe,” in International Conference on Learning Representations, 2024

    2024

  15. [23]

    Graph neural networks beyond compromise between attribute and topology,

    L. Yang, W. Zhou, W. Peng, B. Niu, J. Gu, C. Wang, X. Cao, and D. He, “Graph neural networks beyond compromise between attribute and topology,” in Proceedings of the ACM Web Conference , 2022, pp. 1127–1135

    2022

  16. [24]

    Am-gcn: Adaptive multi-channel graph convolutional networks,

    X. Wang, M. Zhu, D. Bo, P. Cui, C. Shi, and J. Pei, “Am-gcn: Adaptive multi-channel graph convolutional networks,” inProceedings of the ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2020, pp. 1243–1253. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13

    2020

  17. [25]

    Divide and contrast: Self- supervised learning from uncurated data,

    Y . Tian, O. J. Henaff, and A. van den Oord, “Divide and contrast: Self- supervised learning from uncurated data,” in IEEE/CVF International Conference on Computer Vision , 2021, pp. 10 063–10 074

    2021

  18. [26]

    Component divide-and-conquer for real-world image super-resolution,

    P. Wei, Z. Xie, H. Lu, Z. Zhan, Q. Ye, W. Zuo, and L. Lin, “Component divide-and-conquer for real-world image super-resolution,” in European Conference on Computer Vision , 2020, pp. 101–117

    2020

  19. [27]

    A divide-and-conquer approach to the summarization of long documents,

    A. Gidiotis and G. Tsoumakas, “A divide-and-conquer approach to the summarization of long documents,” IEEE/ACM Transactions on Audio, Speech, and Language Processing , vol. 28, pp. 3029–3040, 2020

    2020

  20. [28]

    Ricci curvature of markov chains on metric spaces,

    Y . Ollivier, “Ricci curvature of markov chains on metric spaces,”Journal of Functional Analysis , vol. 256, no. 3, pp. 810–864, 2009

    2009

  21. [29]

    Curvature graph network,

    Z. Ye, K. S. Liu, T. Ma, J. Gao, and C. Chen, “Curvature graph network,” in International Conference on Learning Representations , 2020

    2020

  22. [30]

    Learning stochastic equivalence based on discrete ricci curvature,

    X. Guo, Q. Tian, W. Zhang, W. Wang, and P. Jiao, “Learning stochastic equivalence based on discrete ricci curvature,” in International Joint Conference on Artificial Intelligence , 2021, pp. 1456–1462

    2021

  23. [31]

    Curvature graph neural network,

    H. Li, J. Cao, J. Zhu, Y . Liu, Q. Zhu, and G. Wu, “Curvature graph neural network,” Information Sciences, vol. 592, pp. 50–66, 2022

    2022

  24. [32]

    Measuring and relieving the over-smoothing problem for graph neural networks from the topological view,

    D. Chen, Y . Lin, W. Li, P. Li, J. Zhou, and X. Sun, “Measuring and relieving the over-smoothing problem for graph neural networks from the topological view,” in AAAI Conference on Artificial Intelligence , vol. 34, no. 04, 2020, pp. 3438–3445

    2020

  25. [33]

    Eliciting structural and semantic global knowledge in unsupervised graph contrastive learning,

    K. Ding, Y . Wang, Y . Yang, and H. Liu, “Eliciting structural and semantic global knowledge in unsupervised graph contrastive learning,” in AAAI Conference on Artificial Intelligence , 2023, pp. 7378–7386

    2023

  26. [34]

    Self-supervised representation learning via latent graph prediction,

    Y . Xie, Z. Xu, and S. Ji, “Self-supervised representation learning via latent graph prediction,” in International Conferenceon Machine Learning, 2022, pp. 24 460–24 477

    2022

  27. [35]

    Counterfactual graph learning for anomaly detection on attributed networks,

    C. Xiao, X. Xu, Y . Lei, K. Zhang, S. Liu, and F. Zhou, “Counterfactual graph learning for anomaly detection on attributed networks,” IEEE Transactions on Knowledge and Data Engineering , vol. 35, no. 10, pp. 10 540–10 553, 2023

    2023

  28. [36]

    Learning strong graph neural networks with weak information,

    Y . Liu, K. Ding, J. Wang, V . Lee, H. Liu, and S. Pan, “Learning strong graph neural networks with weak information,” in Proceedings of the ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2023, pp. 1559–1571

    2023

  29. [37]

    Semi-supervised classification with graph convolutional networks,

    T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” in International Conference on Learning Rep- resentations, 2017

    2017

  30. [38]

    Knowledge distillation: A survey,

    J. Gou, B. Yu, S. J. Maybank, and D. Tao, “Knowledge distillation: A survey,” International Journal of Computer Vision , vol. 129, no. 6, pp. 1789–1819, 2021

    2021

  31. [39]

    Knowledge distillation on graphs: A survey,

    Y . Tian, S. Pei, X. Zhang, C. Zhang, and N. V . Chawla, “Knowledge distillation on graphs: A survey,” ACM Computing Surveys , vol. 57, no. 8, pp. 1–16, 2025

    2025

  32. [40]

    Triplet loss for knowledge distillation,

    H. Oki, M. Abe, J. Miyao, and T. Kurita, “Triplet loss for knowledge distillation,” in International Joint Conference on Neural Networks , 2020, pp. 1–7

    2020

  33. [41]

    Self-restrained triplet loss for accurate masked face recognition,

    F. Boutros, N. Damer, F. Kirchbuchner, and A. Kuijper, “Self-restrained triplet loss for accurate masked face recognition,” Pattern Recognition, vol. 124, p. 108473, 2022

    2022

  34. [42]

    Non-negative residual matrix factorization with application to graph anomaly detection,

    H. Tong and C.-Y . Lin, “Non-negative residual matrix factorization with application to graph anomaly detection,” in SIAM International Conference on Data Mining , 2011, pp. 143–153

    2011

  35. [43]

    Rethinking graph neural networks for anomaly detection,

    J. Tang, J. Li, Z. Gao, and J. Li, “Rethinking graph neural networks for anomaly detection,” in International Conferenceon Machine Learning , 2022, pp. 21 076–21 089

    2022

  36. [44]

    Inductive anomaly detection on attributed networks,

    K. Ding, J. Li, N. Agarwal, and H. Liu, “Inductive anomaly detection on attributed networks,” in International Joint Conference on Artificial Intelligence, 2021, pp. 1288–1294

    2021

  37. [45]

    Scalable anomaly ranking of attributed neighborhoods,

    B. Perozzi and L. Akoglu, “Scalable anomaly ranking of attributed neighborhoods,” in SIAM International Conference on Data Mining , 2016, pp. 207–215

    2016

  38. [46]

    Anomalous: A joint modeling approach for anomaly detection on attributed networks,

    Z. Peng, M. Luo, J. Li, H. Liu, Q. Zheng et al. , “Anomalous: A joint modeling approach for anomaly detection on attributed networks,” in International Joint Conference on Artificial Intelligence, 2018, pp. 3513– 3519

    2018

  39. [47]

    Anomalydae: Dual autoencoder for anomaly detection on attributed networks,

    H. Fan, F. Zhang, and Z. Li, “Anomalydae: Dual autoencoder for anomaly detection on attributed networks,” in IEEE International Con- ference on Acoustics, Speech and Signal Processing , 2020, pp. 5685– 5689

    2020

  40. [48]

    Ada-gad: Anomaly- denoised autoencoders for graph anomaly detection,

    J. He, Q. Xu, Y . Jiang, Z. Wang, and Q. Huang, “Ada-gad: Anomaly- denoised autoencoders for graph anomaly detection,” in AAAI Confer- ence on Artificial Intelligence , vol. 38, no. 8, 2024, pp. 8481–8489

    2024

  41. [49]

    Gad- nr: Graph anomaly detection via neighborhood reconstruction,

    A. Roy, J. Shu, J. Li, C. Yang, O. Elshocht, J. Smeets, and P. Li, “Gad- nr: Graph anomaly detection via neighborhood reconstruction,” in ACM International Conference on Web Search and Data Mining , 2024, pp. 576–585

    2024

  42. [50]

    Federated graph anomaly detection via contrastive self-supervised learning,

    X. Kong, W. Zhang, H. Wang, M. Hou, X. Chen, X. Yan, and S. K. Das, “Federated graph anomaly detection via contrastive self-supervised learning,” IEEE Transactions on Neural Networks and Learning Systems, pp. 1–14, 2024

    2024

  43. [51]

    Lifelong language pretraining with distribution-specialized experts,

    W. Chen, Y . Zhou, N. Du, Y . Huang, J. Laudon, Z. Chen, and C. Cui, “Lifelong language pretraining with distribution-specialized experts,” in International Conferenceon Machine Learning , 2023, pp. 5383–5395

    2023

  44. [52]

    A divide-and-conquer method for scalable robust multitask learning,

    Y . Pan, R. Xia, J. Yin, and N. Liu, “A divide-and-conquer method for scalable robust multitask learning,” IEEE Transactions on Neural Networks and Learning Systems , vol. 26, no. 12, pp. 3163–3175, 2015

    2015

  45. [53]

    Graph learning: A survey,

    F. Xia, K. Sun, S. Yu, A. Aziz, L. Wan, S. Pan, and H. Liu, “Graph learning: A survey,” IEEE Transactions on Artificial Intelligence, vol. 2, no. 2, pp. 109–127, 2021

    2021

  46. [54]

    Augmentation-free self-supervised learning on graphs,

    N. Lee, J. Lee, and C. Park, “Augmentation-free self-supervised learning on graphs,” in AAAI Conference on Artificial Intelligence, vol. 36, no. 7, 2022, pp. 7372–7380

    2022

  47. [55]

    Accelerated local anomaly detection via resolving attributed networks,

    N. Liu, X. Huang, and X. Hu, “Accelerated local anomaly detection via resolving attributed networks,” in International Joint Conference on Artificial Intelligence, 2017, pp. 2337–2343

    2017

  48. [56]

    H. Wu, Y . Wu, N. Li, M. Yang, J. Zhang, M. K. Ng, and J. Long, Machine Learning, vol. 113, no. 9, pp. 6247–6272, 2024

    2024

  49. [57]

    Multiplex graph representation learning via dual correlation reduction,

    Y . Mo, Y . Chen, Y . Lei, L. Peng, X. Shi, C. Yuan, and X. Zhu, “Multiplex graph representation learning via dual correlation reduction,” IEEE Transactions on Knowledge and Data Engineering, vol. 35, no. 12, pp. 12 814–12 827, 2023

    2023

  50. [58]

    Self- attentive attributed network embedding through adversarial learning,

    W. Yu, W. Cheng, C. Aggarwal, B. Zong, H. Chen, and W. Wang, “Self- attentive attributed network embedding through adversarial learning,” in IEEE International Conference on Data Mining , 2019, pp. 758–767

    2019

  51. [59]

    Addressing het- erophily in graph anomaly detection: A perspective of graph spectrum,

    Y . Gao, X. Wang, X. He, Z. Liu, H. Feng, and Y . Zhang, “Addressing het- erophily in graph anomaly detection: A perspective of graph spectrum,” in Proceedings of the ACM Web Conference , 2023, pp. 1528–1538

    2023

  52. [60]

    Motif- consistent counterfactuals with adversarial refinement for graph-level anomaly detection,

    C. Xiao, S. Pang, W. Tai, Y . Huang, G. Trajcevski, and F. Zhou, “Motif- consistent counterfactuals with adversarial refinement for graph-level anomaly detection,” in Proceedings of the ACM SIGKDD Conference on Knowledge Discovery and Data Mining , 2024, p. 3518–3526

    2024

  53. [61]

    Alleviat- ing structural distribution shift in graph anomaly detection,

    Y . Gao, X. Wang, X. He, Z. Liu, H. Feng, and Y . Zhang, “Alleviat- ing structural distribution shift in graph anomaly detection,” in ACM International Conference on Web Search and Data Mining , 2023, pp. 357–365

    2023

  54. [62]

    Learning graph deep autoencoder for anomaly detection in multi-attributed networks,

    M. Shao, Y . Lin, Q. Peng, J. Zhao, Z. Pei, and Y . Sun, “Learning graph deep autoencoder for anomaly detection in multi-attributed networks,” Knowledge-Based Systems, vol. 260, p. 110084, 2023

    2023

  55. [63]

    Anomman: Detect anomalies on multi-view attributed networks,

    L.-H. Chen, H. Li, W. Zhang, J. Huang, X. Ma, J. Cui, N. Li, and J. Yoo, “Anomman: Detect anomalies on multi-view attributed networks,” Information Sciences, vol. 628, pp. 1–21, 2023

    2023

  56. [64]

    Arise: Graph anomaly detection on attributed networks via substructure awareness,

    J. Duan, B. Xiao, S. Wang, H. Zhou, and X. Liu, “Arise: Graph anomaly detection on attributed networks via substructure awareness,” IEEE Transactions on Neural Networks and Learning Systems , vol. 35, no. 12, pp. 18 172–18 185, 2023

    2023

  57. [65]

    Samcl: Subgraph-aligned multiview contrastive learning for graph anomaly detection,

    J. Hu, B. Xiao, H. Jin, J. Duan, S. Wang, Z. Lv, S. Wang, X. Liu, and E. Zhu, “Samcl: Subgraph-aligned multiview contrastive learning for graph anomaly detection,” IEEE Transactions on Neural Networks and Learning Systems , vol. 36, no. 1, pp. 1664–1676, 2023

    2023

  58. [66]

    Anemone: Graph anomaly detection with multi-scale contrastive learning,

    M. Jin, Y . Liu, Y . Zheng, L. Chi, Y .-F. Li, and S. Pan, “Anemone: Graph anomaly detection with multi-scale contrastive learning,” in ACM International Conference on Information & Knowledge Management , 2021, pp. 3122–3126

    2021

  59. [67]

    Data augmentation for deep graph learning: A survey,

    K. Ding, Z. Xu, H. Tong, and H. Liu, “Data augmentation for deep graph learning: A survey,” ACM SIGKDD Explorations Newsletter , vol. 24, no. 2, pp. 61–77, 2022

    2022

  60. [68]

    Learning on attribute-missing graphs,

    X. Chen, S. Chen, J. Yao, H. Zheng, Y . Zhang, and I. W. Tsang, “Learning on attribute-missing graphs,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 44, no. 2, pp. 740–757, 2020

    2020

  61. [69]

    Heterogeneous graph neural net- work via attribute completion,

    D. Jin, C. Huo, C. Liang, and L. Yang, “Heterogeneous graph neural net- work via attribute completion,” in Proceedings of The Web Conference , 2021, pp. 391–400

    2021

  62. [70]

    Initializing then refining: A simple graph attribute imputation network,

    W. Tu, S. Zhou, X. Liu, Y . Liu, Z. Cai, E. Zhu, Z. Changwang, and J. Cheng, “Initializing then refining: A simple graph attribute imputation network,” in International Joint Conference on Artificial Intelligence , 2022, pp. 3494–3500

    2022

  63. [71]

    Missing data imputation with adversarially-trained graph convolutional networks,

    I. Spinelli, S. Scardapane, and A. Uncini, “Missing data imputation with adversarially-trained graph convolutional networks,” Neural Networks , vol. 129, pp. 249–260, 2020. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 14

    2020

  64. [72]

    Graph convolutional networks for graphs containing missing features,

    H. Taguchi, X. Liu, and T. Murata, “Graph convolutional networks for graphs containing missing features,” Future Generation Computer Systems, vol. 117, pp. 155–168, 2021

    2021

  65. [73]

    Handling missing data via max-entropy regularized graph autoencoder,

    Z. Gao, Y . Niu, J. Cheng, J. Tang, L. Li, T. Xu, P. Zhao, F. Tsung, and J. Li, “Handling missing data via max-entropy regularized graph autoencoder,” in AAAI Conference on Artificial Intelligence , 2023, pp. 7651–7659

    2023

  66. [74]

    S2gae: Self-supervised graph autoencoders are generalizable learners with graph masking,

    Q. Tan, N. Liu, X. Huang, S.-H. Choi, L. Li, R. Chen, and X. Hu, “S2gae: Self-supervised graph autoencoders are generalizable learners with graph masking,” in ACM International Conference on Web Search and Data Mining , 2023, pp. 787–795

    2023

  67. [75]

    Inductive and unsupervised representation learning on graph structured objects,

    L. Wang, B. Zong, Q. Ma, W. Cheng, J. Ni, W. Yu, Y . Liu, D. Song, H. Chen, and Y . Fu, “Inductive and unsupervised representation learning on graph structured objects,” in International Conference on Learning Representations, 2020

    2020

  68. [76]

    Neo-gnns: Neigh- borhood overlap-aware graph neural networks for link prediction,

    S. Yun, S. Kim, J. Lee, J. Kang, and H. J. Kim, “Neo-gnns: Neigh- borhood overlap-aware graph neural networks for link prediction,” in Advances in Neural Information Processing Systems , 2021, pp. 13 683– 13 694

    2021

  69. [77]

    Neural bellman- ford networks: A general graph neural network framework for link prediction,

    Z. Zhu, Z. Zhang, L.-P. Xhonneux, and J. Tang, “Neural bellman- ford networks: A general graph neural network framework for link prediction,” in Advances in Neural Information Processing Systems , 2021, pp. 29 476–29 490

    2021

  70. [78]

    Linkless link prediction via relational distillation,

    Z. Guo, W. Shiao, S. Zhang, Y . Liu, N. V . Chawla, N. Shah, and T. Zhao, “Linkless link prediction via relational distillation,” in International Conferenceon Machine Learning , 2023, pp. 12 012–12 033

    2023

  71. [79]

    Learning from counterfactual links for link prediction,

    T. Zhao, G. Liu, D. Wang, W. Yu, and M. Jiang, “Learning from counterfactual links for link prediction,” in International Conferenceon Machine Learning, 2022, pp. 26 911–26 926

    2022

  72. [80]

    Self-supervised temporal graph learning with temporal and structural intensity alignment,

    M. Liu, K. Liang, Y . Zhao, W. Tu, S. Zhou, X. Gan, X. Liu, and K. He, “Self-supervised temporal graph learning with temporal and structural intensity alignment,” IEEE Transactions on Neural Networks and Learning Systems , vol. 36, no. 4, pp. 6355–6367, 2024

    2024

  73. [81]

    Learning robust representations with graph denoising policy network,

    L. Wang, W. Yu, W. Wang, W. Cheng, W. Zhang, H. Zha, X. He, and H. Chen, “Learning robust representations with graph denoising policy network,” in IEEE International Conference on Data Mining , 2019, pp. 1378–1383

    2019

Pith tools

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