Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Hierarchical Uncertainty-Aware Graph Neural Network

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read The paper claims that uncertainty computed at node, community, and global scales, fed back into message passing, keeps graph neural networks accurate on heterophilic graphs and under adversarial attacks, with error bounded by $(1-q)^k$.

desk verdict A plausible architectural contribution with real but small empirical gains, whose theoretical guarantees do not hold up on inspection and whose abstract overclaims a self-supervised component that is absent from the method. read the letter →

arxiv 2504.19820 v2 pith:TSAVVWLO submitted 2025-04-28 cs.LG cs.IR

Yoonhyuk Choi , Jiho Choi , Taewook Ko , Chong-Kwon Kim This is my paper · ORCID
classification cs.LGcs.IR
keywords graphneuralnetworksheterophilyhierarchicalstructurelearninguncertainty-awaremessage-passingPAC-Bayesgeneralizationboundssemi-supervisednodeclassificationadversarialrobustnessuncertaintycalibration
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

HU-GNN is an architecture for semi-supervised node classification that treats uncertainty as a signal computed at three structural scales: individual nodes, learned communities, and a global graph node. The paper claims this unifies multi-scale representation learning, principled uncertainty estimation, and self-supervised embedding diversity in one end-to-end framework. The central claim is that uncertainty scores recomputed at every layer and fed back into attention-weighted message passing allow the model to down-weight misleading one-hop neighbors in heterophilic graphs and draw instead on community or global evidence, improving accuracy and robustness. The authors support this with a PAC-Bayes generalization bound, a contraction result for the uncertainty updates, and a robustness bound in which misclassification decays like $(1-q)^k$ when learned communities capture two-hop same-label structure. Experiments on six benchmark graphs show accuracy gains on both homophilic and heterophilic datasets and smaller losses under edge dropping, structural attacks, and feature noise.

What carries the argument

The load-bearing mechanism is the uncertainty estimator $f_u(\cdot)$, which maps a mean squared deviation of neighboring or member embeddings to a score in $[0,1]$, plus uncertainty-modulated attention, $m_{ij} \propto \exp(a^\top[\tilde h_i \| \tilde h_j])\exp(-u_j)$, and a final representation $h_i^{\mathrm{final}} = \lambda_i h_i + \lambda_{C_m} h_{C_m} + \lambda_G h_G$ with $\lambda_v \propto \exp(a_v)\exp(-u_v)$. This lets a node with high local uncertainty lean on community or global evidence. The theory then rests on three results: a PAC-Bayes bound using an effective degree $\tilde{\Delta}_i$, a contraction constant $c<1$ for the joint feature-uncertainty update, and Theorem 3's bound $(1-q)^k$ on misclassification under heterophily.

What would settle it

Take a heterophilic graph with one-hop label agreement $p<0.5$ and two-hop label agreement $q$ at most 0.5; if HU-GNN fails to beat a standard GCN there, Theorem 3's regime does not apply. Alternatively, inspect the trained cluster assignments and compare them with true two-hop same-label neighborhoods: high overlap with $q>0.5$ but no $(1-q)^k$-type error decay would falsify the bound, as would strong accuracy with low overlap.

Watch

Extended reading notes

Core claim

The paper's central discovery is that uncertainty can be organized hierarchically and propagated as a latent variable, and that doing so keeps GNNs accurate in heterophilic settings. Concretely, HU-GNN computes local uncertainty from the variance of neighbor features, community uncertainty from intra-cluster feature variance, and global uncertainty from spread across communities; each layer's uncertainty values rescale the attention weights feeding the next aggregation. The decisive theoretical claim is Theorem 3: in a binary classification setting with one-hop homophily $p<0.5$ but community-level (two-hop) homophily $q>0.5$, the probability that HU-GNN misclassifies a node is on the order of $(1-q)^k$, where $k$ counts corroborating two-hop neighbors or community members, whereas a standard GNN errs with probability on the order of $(1-p)$ or worse. If this holds, a GNN can remain accurate on heterophilic graphs not by changing its propagation operator but by learning when to distrust immediate neighbors and ascend the hierarchy.

Load-bearing premise

The heterophily guarantee assumes that the learned community grouping effectively captures two-hop neighborhoods whose members share the target node's label with probability $q>0.5$, and the paper states this assumption without constructing or verifying such communities.

Editorial extensions

If this is right

  • On heterophilic graphs where direct neighbors mostly disagree in label, HU-GNN should keep accuracy high by shifting weight from uncertain one-hop neighbors to community and global evidence, which standard GCN and GAT cannot do.
  • Uncertainty gating yields a tighter PAC-Bayes generalization bound: high-degree nodes contribute only their effective degree $\tilde{\Delta}_i$, so noisy neighbors inflate the bound far less than in a standard GNN.
  • Because the joint feature-uncertainty update is contractive, repeated or deeper uncertainty propagation converges to a fixed point instead of oscillating or diverging.
  • Under edge dropout, structural poisoning, and feature perturbation, HU-GNN's per-layer uncertainty scores should absorb much of the corruption, producing smaller accuracy drops than flat uncertainty baselines.
  • The design subsumes GAT and local-global GNNs as special cases, so as community or global components are ablated, accuracy should degrade gradually rather than collapse.

Reading between the lines

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

  • The paper leaves implicit that Theorem 3's condition turns heterophily robustness into a clustering-quality problem: if learned communities capture two-hop same-label structure, any uncertainty-gated GNN should inherit the bound, which is testable by swapping in different clustering methods while keeping the rest of HU-GNN fixed.
  • A natural extension the paper motivates but does not evaluate is to use the community and global uncertainty scores as out-of-distribution or anomaly detectors for nodes, since the scores already summarize how anomalous a node looks relative to its cluster and the whole graph.
  • The bound's dependence on $q$ rather than $p$ suggests that on heterophilic benchmarks, HU-GNN's accuracy should correlate with how well its clusters align with true two-hop same-label neighborhoods; measuring that alignment directly would separate the effect of clustering from the effect of uncertainty weighting.
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 / 5 minor

Summary. The paper proposes HU-GNN, a graph neural network that combines local uncertainty-aware message passing, differentiable community pooling, and a global context node. The authors claim three theoretical contributions: a PAC-Bayes generalization bound with an uncertainty-dependent effective degree, a convergence theorem for the joint feature-uncertainty update, and a robustness bound showing that heterophily misclassification error decays as (1-q)^k. The empirical section reports accuracy on six benchmarks, ablations of the hierarchical components, and robustness under three perturbation types. The architecture and pseudo-code are clearly described, but the theoretical results are presented as proof sketches, and the central heterophily theorem rests on an assumption that essentially guarantees the claimed conclusion. As a result, the advertised theoretical contributions are not established in the manuscript.

Significance. If the theoretical claims were properly supported, the idea of propagating uncertainty across hierarchical scales would be a useful contribution to robust GNN design, and the explicit treatment of heterophily via community-level evidence is a reasonable direction. The paper also provides a useful implementation-level description with pseudo-code and a complexity analysis, and it promises a public code release. However, the theoretical significance claimed in the abstract is not currently supported: the theorems are either too imprecise to verify or assume the desired conclusion. The empirical gains over strong baselines are small (0.3–1.2 percentage points) and are reported without standard deviations or repeated-run statistics, so the experimental evidence is also not fully convincing. The manuscript therefore has a promising core idea but is not yet a sound paper in its present form.

major comments (4)
  1. [§5.1, Eq. (23)] Theorem 1 is not stated precisely enough to be verified: the effective degree Δ~_i, which is central to the claimed improvement, is never formally defined, the term H[u_L] is undefined, and the symbol δ denotes both the confidence level and the complexity term. The paragraph after Eq. (23) concedes that the precise form of the bound is technical and gives only an intuitive explanation. No derivation connects the actual HU-GNN updates in Eqs. (4)–(18) to this bound, so the advertised PAC-Bayes generalization guarantee is not established.
  2. [§5.2, Theorem 2] The convergence theorem is conditional on an assumption that is not proved. The proof states that f_u is chosen to be a contraction or non-expanding map, but the algorithm in Eq. (8) applies a nonlinear f_u to variances, and no contraction constant c or norm is derived for the full feature-uncertainty map F. The informal claim that averaging operators are typically contractive does not cover variance-based nonlinearities or the composed update. Consequently, the statement that HU-GNN's uncertainty updates converge is not supported for the actual architecture.
  3. [§5.3, Theorem 3] The heterophily robustness bound is circular. The theorem assumes that the community grouping effectively captures two-hop neighbors with community-level homophily q>0.5, but this is precisely the property that must be proved from the architecture. The community assignment in Eq. (9) is a softmax over learned weight vectors applied to local embeddings, and nothing in the construction forces clusters to contain predominantly same-label two-hop nodes. Eq. (12) mean-pools all assigned member embeddings, including misleading one-hop neighbors, and Eq. (16) is a weighted average of local, community, and global embeddings, not a majority vote over k independent witnesses. The claimed exponential (1-q)^k misclassification bound therefore does not follow from the model's updates; it reduces to a restatement of the assumption that informative two-hop structure has been captured.
  4. [§6.5, Eq. (24)] The calibration evaluation is not reproducible as described. The ECE-based update rule in Eq. (24) does not state whether ECE is computed on the training, validation, or test split, and the 20–35% ECE reduction reported in §6.2 is not accompanied by repeated-run statistics. If the test set was used to tune β2, the calibration claim is invalid; as written, the reader cannot determine which split is used, so the degree to which the reported calibration improvement is genuine remains unclear.
minor comments (5)
  1. [§4.2, Eq. (12) and Algorithm 1, line 12] Eq. (12) writes h_Cm = Σ_j W_C h_j without normalization, while Algorithm 1 divides by |C1|; the text calls this mean pooling, so the formula should be corrected to be consistent with the algorithm.
  2. [§3.1 and §4.2] The symbol m denotes the number of edges in §3.1 and the number of communities in Eq. (9) onward; this dual use will confuse readers and should be resolved with distinct notation.
  3. [§5.3, Theorem 3] The theorem describes the model as a two-layer HU-GNN consisting of one community layer and one global layer, but §4.1 sets the total number of local layers to two and §4.2–4.3 add community and global stages; the layer count in the theorem should match the architecture.
  4. [Table 3 and Figure 2] No standard deviations or significance tests are reported for the classification accuracies, and the ablation study does not state how many random splits or seeds were used; the reported gains, some of which are under one percentage point, should be accompanied by variance estimates.
  5. [References] Several references list only 'Others' instead of full author lists (e.g., [15], [19], [61], [68]), which prevents readers from verifying the cited work; these should be completed.

Circularity Check

3 steps flagged · score 6.0 of 10

Theorem 3's heterophily bound assumes the two-hop homophilic communities it purports to guarantee; Theorem 2 assumes contraction; ECE gains are tuned against ECE.

  1. self definitional [Section 5.3, Theorem 3]
    "Suppose the community grouping effectively captures two-hop neighbors. In that case, the model identifies the uncertainty associated with immediate (1-hop) neighbors, assigning them lower weights, and thus relies more on the informative two-hop neighbors. Under mild conditions, the probability that HU-GNN misclassifies a node can be bounded by a term on the order of (1−q)^k, where k relates to the number of two-hop neighbors or community members offering corroborating evidence."

    q is defined in the setup as the probability that a two-hop neighbor shares the same label, so (1−q)^k is just the probability that k independent two-hop witnesses are wrong. The theorem's only substantive condition, 'community grouping effectively captures two-hop neighbors,' is exactly what would make q the community-level homophily and supply k corroborating witnesses. The architecture supplies no such mechanism: Eq. 9 assigns communities by a softmax over learned weight vectors on local embeddings, Eq. 12 mean-pools all members, and Eq. 16-18 forms a weighted average rather than a majority vote over independent witnesses. Thus the robustness bound is a restatement of the assumption, not a consequence of the model.

  2. self definitional [Section 5.2, Theorem 2 and preceding assumptions]
    "the uncertainty update function f_u(·) is chosen to be contraction mappings or at least non-expanding in a suitable norm. ... There exists a non-negative constant c < 1 such that for any two uncertainty states U and U′ (e.g. at two iterations), their difference is contractive under F: ∥F(U)−F(U′)∥∞ ≤ c∥U−U′∥∞."

    The theorem's conclusion, that F is contractive with constant c < 1, is precisely the assumption (b) that f_u is chosen to be a contraction mapping. The proof sketch appeals to averaging, but the actual update in Eq. 8 is f_u applied to the mean squared deviation of neighbor features, not a weighted average of previous uncertainties, so no argument shows this composite map inherits contraction. The convergence guarantee is assumed rather than derived.

1 more flagged steps
  1. fitted input called prediction [Section 6.5 (Eq. 24) and Section 6.2]
    "The selection criterion is the minimal Expected Calibration Error (ECE) ... β2←β2×(1.2, if ECE > 0.05, 0.8, if ECE < 0.02) ... HU-GNN lowers Expected Calibration Error (ECE, §6.5) by 20–35% relative to strong baselines."

    The calibration coefficient β2 is selected and adaptively updated by directly minimizing or thresholding ECE, and the reported 20–35% ECE reduction is then presented as evidence of calibration. Because the metric being reported is the same quantity used to tune the hyperparameters, the improvement is forced by construction unless a separately held-out test ECE is explicitly reported, which the paper does not do.

full rationale

The empirical benchmark results are independent and self-contained: the model is evaluated against standard baselines on Cora, Citeseer, Pubmed, Actor, Chameleon, and Squirrel, and those comparisons are not circular. The self-citations in the paper ([8], [9], [29], [30]) are related-work references and are not load-bearing for the central derivation. However, the two main formal guarantees are circular. Theorem 3's heterophily robustness bound reduces to its own assumption that learned communities capture two-hop same-label structure with q > 0.5; the bound O((1−q)^k) is a restatement of that assumption, and the architecture's softmax community assignment and mean-pooling do not implement independent corroborating witnesses. Theorem 2 explicitly assumes f_u is a contraction and then concludes that the uncertainty update is contractive. Additionally, the calibration improvement is obtained by tuning β2 against ECE, so the reported ECE reduction is not an independent validation. The PAC-Bayes bound in Theorem 1 is stated only informally with O(...) terms and a proof sketch, but this is better characterized as a rigor gap than a circular step. Overall, because the central robustness and convergence claims reduce to assumptions or fitted objectives, a score of 6 is appropriate: the paper is not entirely circular, but its advertised theoretical contributions are partially circular by construction.

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

The central claim rests on free hyperparameters (beta_1, beta_2, M) that are tuned on validation data, and on ad hoc assumptions about the contraction of f_u and the quality of learned communities. The latter assumptions are essentially the conclusions the theorems try to establish, creating a circular burden. No new physical or external entities are introduced beyond a virtual global node.

free parameters (3)
  • beta_1 (sharpness weight) = 0.3
    Tuned on validation ECE and accuracy; controls the uncertainty sharpening term in the loss (Eq. 19).
  • beta_2 (calibration weight) = 0.1
    Tuned on validation ECE with an adaptive schedule (Eq. 24); controls the calibration penalty.
  • Number of communities M = not reported
    A critical hyperparameter for the hierarchical structure, but the paper does not specify how it is chosen or its value for any dataset.
assumptions (4)
  • ad hoc to paper The uncertainty update function f_u is a contraction or non-expansive map (Theorem 2).
    Needed for the convergence proof; the paper states it as an assumption rather than deriving it from the actual update equations.
  • ad hoc to paper Learned communities capture two-hop neighborhoods with homophily q > 0.5 (Theorem 3).
    This is the central premise of the heterophily robustness bound, yet no mechanism guarantees the learned clusters have this property.
  • domain assumption Two-hop neighbors are more homophilic than one-hop neighbors in heterophilic graphs.
    Borrowed from EvenNet and H2GCN; used to motivate the community-level argument in Section 5.3.
  • standard math Activations and weight matrices are Lipschitz continuous.
    A standard regularity assumption in GNN analysis, invoked for the convergence argument.
invented entities (1)
  • Global context node
    purpose: Summarizes the entire graph and tracks global distributional shift.
    A virtual node added to the architecture; it has no falsifiable handle outside the model's internal representations and performance.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hierarchical Uncertainty-Aware Graph Neural Network." pith.science (2026). https://pith.science/paper/TSAVVWLO

@misc{pith2026250419820,
  author       = {Pith},
  title        = {Pith review of: Hierarchical Uncertainty-Aware Graph Neural Network},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TSAVVWLO}},
  note         = {Machine review of arXiv:2504.19820}
}
read the original abstract

Recent research on graph neural networks (GNNs) has explored mechanisms for capturing local uncertainty and exploiting graph hierarchies to mitigate data sparsity and leverage structural properties. However, the synergistic integration of these two approaches remains underexplored. This work introduces a novel architecture, the Hierarchical Uncertainty-Aware Graph Neural Network (HU-GNN), which unifies multi-scale representation learning, principled uncertainty estimation, and self-supervised embedding diversity within a single end-to-end framework. Specifically, HU-GNN adaptively forms node clusters and estimates uncertainty at multiple structural scales from individual nodes to higher levels. These uncertainty estimates guide a robust message-passing mechanism and attention weighting, effectively mitigating noise and adversarial perturbations while preserving predictive accuracy on semi-supervised classification tasks. We also offer key theoretical contributions, including a probabilistic formulation, rigorous uncertainty-calibration guarantees, and formal robustness bounds. Extensive experiments on standard benchmarks demonstrate that our model achieves state-of-the-art robustness and interpretability.

Figures

Figures reproduced from arXiv: 2504.19820 by the authors.

Figure 1
Figure 1. The overall framework of HU-GNN, illustrating message [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 3
Figure 3. (RQ4) Hyperparameter analysis on the Cora and Chameleon dataset, varying 𝛽1 (x-axis) and 𝛽2 (y-axis) in Equa￾tion 19. Here, the z-axis represents validation accuracy were also observed on heterophilic datasets: Actor, Chameleon, and Squirrel, where community and global cues cut Metattack damage by 37% relative to UnGSL. Edge-level guards such as GNNGuard depend on fixed similarity rules, and crafted edges can still … view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. LiveGraph: Active-Structure Neural Re-ranking for Exercise Recommendation

    cs.IR 2026-02 unverdicted novelty 5.0 of 10

    LiveGraph re-ranks exercise recommendations with a dynamic concept kernel and uncertainty-aware meta-RL fusion, claiming improved accuracy and diversity, with only accuracy tables shown.

Reference graph

Works this paper leans on

83 extracted references · 49 canonical work pages · cited by 1 Pith paper

  1. [1]

    Sami Abu-El-Haija, Bryan Perozzi, Amol Kapoor, Nazanin Alipourfard, Kristina Lerman, Hrayr Harutyunyan, Greg Ver Steeg, and Aram Galstyan. 2019. Mixhop: Higher-order graph convolutional architectures via sparsified neighborhood mixing. In international conference on machine learning . PMLR, 21–29

    2019

  2. [2]

    Deyu Bo, Xiao Wang, Chuan Shi, and Huawei Shen. 2021. Beyond low-frequency information in graph convolutional networks. arXiv preprint arXiv:2101.00797 (2021)

    arXiv 2021

  3. [3]

    Shaked Brody, Uri Alon, and Eran Yahav. 2021. How attentive are graph attention networks? arXiv preprint arXiv:2105.14491 (2021)

    arXiv 2021

  4. [4]

    Cen Chen, Kenli Li, Wei Wei, Joey Tianyi Zhou, and Zeng Zeng. 2021. Hierarchical graph neural networks for few-shot learning. IEEE Transactions on Circuits and Systems for Video Technology 32, 1 (2021), 240–252

    2021

  5. [5]

    Ming Chen, Zhewei Wei, Zengfeng Huang, Bolin Ding, and Yaliang Li. 2020. Simple and deep graph convolutional networks. In International Conference on Machine Learning. PMLR, 1725–1735

    2020

  6. [6]

    Yuhan Chen, Yihong Luo, Jing Tang, Liang Yang, Siya Qiu, Chuan Wang, and Xiaochun Cao. 2023. LSGNN: Towards General Graph Neural Network in Node Classification by Local Similarity. arXiv preprint arXiv:2305.04225 (2023)

    arXiv 2023

  7. [7]

    Eli Chien, Jianhao Peng, Pan Li, and Olgica Milenkovic. 2020. Adaptive universal generalized pagerank graph neural network. arXiv preprint arXiv:2006.07988 (2020)

    arXiv 2020

  8. [8]

    Yoonhyuk Choi, Jiho Choi, Taewook Ko, Hyungho Byun, and Chong-Kwon Kim. 2022. Finding Heterophilic Neighbors via Confidence-based Subgraph Matching for Semi-supervised Node Classification. InProceedings of the 31st ACM International Conference on Information & Knowledge Management . 283–292

    work page 2022

Show all 83 references
  1. [9]

    Yoonhyuk Choi, Jiho Choi, Taewook Ko, and Chong-Kwon Kim. 2023. Is signed message essential for graph neural networks. arXiv preprint arXiv:2301.08918 (2023)

    2023 arXiv

  2. [10]

    Michaël Defferrard, Xavier Bresson, and Pierre Vandergheynst. 2016. Convolu- tional neural networks on graphs with fast localized spectral filtering. Advances in neural information processing systems 29 (2016)

    2016

  3. [11]

    Tyler Derr, Yao Ma, and Jiliang Tang. 2018. Signed graph convolutional networks. In 2018 IEEE International Conference on Data Mining (ICDM) . IEEE, 929–934

    2018

  4. [12]

    Zheng Fang, Lingjun Xu, Guojie Song, Qingqing Long, and Yingxue Zhang. 2022. Polarized Graph Neural Networks. In Proceedings of the ACM Web Conference

    2022

  5. [13]

    Boyuan Feng, Yuke Wang, and Yufei Ding. 2021. Uag: Uncertainty-aware attention graph neural network for defending adversarial attacks. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 35. 7404–7412

    2021

  6. [14]

    Mauro Forti and Paolo Nistri. 2003. Global convergence of neural networks with discontinuous neuron activations. IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications 50, 11 (2003), 1421–1435

    2003

  7. [15]

    Fuchsgruber and Others

    F. Fuchsgruber and Others. 2024. Graph Energy-Based Models for Multi-Scale Uncertainty. Proceedings of the 38th Conference on Neural Information Processing Systems (NeurIPS) (2024)

    2024

  8. [16]

    Pascal Germain, Alexandre Lacasse, François Laviolette, and Mario Marchand

  9. [17]

    Yuhe Guo and Zhewei Wei. 2022. Clenshaw Graph Neural Networks. arXiv preprint arXiv:2210.16508 (2022)

    2022 arXiv

  10. [18]

    Shen Han, Zhiyao Zhou, Jiawei Chen, Zhezheng Hao, Sheng Zhou, Gang Wang, Yan Feng, Chun Chen, and Can Wang. 2025. Uncertainty-Aware Graph Structure Learning. arXiv preprint arXiv:2502.12618 (2025)

    2025 arXiv

  11. [19]

    He and Others

    B. He and Others. 2024. Scattering Graph Representations: A Unifying Perspective on Graph Contrastive Learning. Proceedings of the 38th Conference on Neural Information Processing Systems (NeurIPS) (2024)

    2024

  12. [20]

    Dongxiao He, Lianze Shan, Jitao Zhao, Hengrui Zhang, Zhen Wang, and Weixiong Zhang. 2024. Exploitation of a Latent Mechanism in Graph Contrastive Learning: Representation Scattering. Advances in Neural Information Processing Systems 37 (2024), 115351–115376

    2024

  13. [21]

    Mikael Henaff, Joan Bruna, and Yann LeCun. 2015. Deep convolutional networks on graph-structured data. arXiv preprint arXiv:1506.05163 (2015)

    2015 arXiv

  14. [22]

    Hans Hao-Hsun Hsu, Yuesong Shen, Christian Tomani, and Daniel Cremers

  15. [23]

    Junjie Huang, Huawei Shen, Liang Hou, and Xueqi Cheng. 2019. Signed graph attention networks. In International Conference on Artificial Neural Networks . Springer, 566–577

    2019

  16. [24]

    Kexin Huang, Ying Jin, Emmanuel Candes, and Jure Leskovec. 2023. Uncertainty quantification over graph with conformalized graph neural networks. Advances in Neural Information Processing Systems 36 (2023), 26699–26721

    2023

  17. [25]

    Huang and Others

    X. Huang and Others. 2024. Cluster-GT: Hierarchical Graph Transformers with Adaptive Clustering. Proceedings of the 38th Conference on Neural Information Processing Systems (NeurIPS) (2024)

    2024

  18. [26]

    Eric Jang, Shixiang Gu, and Ben Poole. 2016. Categorical reparameterization with gumbel-softmax. arXiv preprint arXiv:1611.01144 (2016)

    2016 arXiv

  19. [27]

    Thomas N Kipf and Max Welling. 2016. Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907 (2016)

    2016 arXiv

  20. [28]

    Johannes Klicpera, Aleksandar Bojchevski, and Stephan Günnemann. 2018. Pre- dict then propagate: Graph neural networks meet personalized pagerank. arXiv preprint arXiv:1810.05997 (2018)

    2018 arXiv

  21. [29]

    Taewook Ko, Yoonhyuk Choi, and Chong-Kwon Kim. 2023. Signed directed graph contrastive learning with laplacian augmentation.arXiv preprint arXiv:2301.05163 (2023)

    2023 arXiv

  22. [30]

    Taewook Ko, Yoonhyuk Choi, and Chong-Kwon Kim. 2023. Universal graph con- trastive learning with a novel laplacian perturbation. In Uncertainty in Artificial Intelligence. PMLR, 1098–1108

    2023

  23. [31]

    Volodymyr Kuleshov and Shachi Deshpande. 2022. Calibrated and sharp uncer- tainties in deep learning via density estimation. In International Conference on Machine Learning. PMLR, 11683–11693

    2022

  24. [32]

    Soo Yong Lee, Fanchen Bu, Jaemin Yoo, and Kijung Shin. 2023. Towards deep attention in graph neural networks: Problems and remedies. In International Conference on Machine Learning . PMLR, 18774–18795

    2023

  25. [33]

    Runlin Lei, Zhen Wang, Yaliang Li, Bolin Ding, and Zhewei Wei. 2022. EvenNet: Ignoring Odd-Hop Neighbors Improves Robustness of Graph Neural Networks. arXiv preprint arXiv:2205.13892 (2022)

    2022 arXiv

  26. [34]

    Bingheng Li, Erlin Pan, and Zhao Kang. 2024. Pc-conv: Unifying homophily and heterophily with two-fold filtering. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 38. 13437–13445

    2024

  27. [35]

    Xiang Li, Renyu Zhu, Yao Cheng, Caihua Shan, Siqiang Luo, Dongsheng Li, and Weining Qian. 2022. Finding Global Homophily in Graph Neural Networks When Meeting Heterophily. arXiv preprint arXiv:2205.07308 (2022)

    2022 arXiv

  28. [36]

    Renjie Liao, Raquel Urtasun, and Richard Zemel. 2020. A pac-bayesian approach to generalization bounds for graph neural networks. arXiv preprint arXiv:2012.07690 (2020). Hierarchical Uncertainty-Aware Graph Neural Network CIKM ’25, November 10–14, 2025, Seoul, South Korea

    2020 arXiv

  29. [37]

    Xixun Lin, Wenxiao Zhang, Fengzhao Shi, Chuan Zhou, Lixin Zou, Xiangyu Zhao, Dawei Yin, Shirui Pan, and Yanan Cao. 2024. Graph neural stochastic diffu- sion for estimating uncertainty in node classification. In Forty-first International Conference on Machine Learning

    2024

  30. [38]

    Yang Liu, Xiang Ao, Fuli Feng, and Qing He. 2022. Ud-gnn: Uncertainty-aware debiased training on semi-homophilous graphs. In Proceedings of the 28th ACM SIGKDD conference on knowledge discovery and data mining . 1131–1140

    2022

  31. [39]

    Sitao Luan, Chenqing Hua, Qincheng Lu, Jiaqi Zhu, Mingde Zhao, Shuyuan Zhang, Xiao-Wen Chang, and Doina Precup. 2022. Revisiting heterophily for graph neural networks. arXiv preprint arXiv:2210.07606 (2022)

    2022 arXiv

  32. [40]

    Dongsheng Luo, Wei Cheng, Wenchao Yu, Bo Zong, Jingchao Ni, Haifeng Chen, and Xiang Zhang. 2021. Learning to drop: Robust graph neural network via topological denoising. In Proceedings of the 14th ACM International Conference on Web Search and Data Mining. 779–787

    2021

  33. [41]

    Behnam Neyshabur, Srinadh Bhojanapalli, and Nathan Srebro. 2017. A pac- bayesian approach to spectrally-normalized margin bounds for neural networks. arXiv preprint arXiv:1707.09564 (2017)

    2017 arXiv

  34. [42]

    Hoang Nt and Takanori Maehara. 2019. Revisiting graph neural networks: All we have is low-pass filters. arXiv preprint arXiv:1905.09550 (2019)

    2019 arXiv

  35. [43]

    Hongbin Pei, Bingzhe Wei, Kevin Chen-Chuan Chang, Yu Lei, and Bo Yang

  36. [44]

    Marius-Constantin Popescu, Valentina E Balas, Liliana Perescu-Popescu, and Nikos Mastorakis. 2009. Multilayer perceptron and neural networks. WSEAS Transactions on Circuits and Systems 8, 7 (2009), 579–588

    2009

  37. [45]

    Yu Rong, Wenbing Huang, Tingyang Xu, and Junzhou Huang. 2019. Dropedge: Towards deep graph convolutional networks on node classification.arXiv preprint arXiv:1907.10903 (2019)

    2019 arXiv

  38. [46]

    Benedek Rozemberczki, Ryan Davies, Rik Sarkar, and Charles Sutton. 2019. Gem- sec: Graph embedding with self clustering. In Proceedings of the 2019 IEEE/ACM international conference on advances in social networks analysis and mining. 65–72

    2019

  39. [47]

    Uday Shankar Shanthamallu, Jayaraman J Thiagarajan, and Andreas Spanias

  40. [48]

    Weili Shi, Xueying Yang, Xujiang Zhao, Haifeng Chen, Zhiqiang Tao, and Sheng Li. 2023. Calibrate graph neural networks under out-of-distribution nodes via deep q-learning. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Management . 2270–2279

    2023

  41. [49]

    Saurabh Sihag, Gonzalo Mateos, Corey McMillan, and Alejandro Ribeiro. 2022. coVariance neural networks. Advances in neural information processing systems 35 (2022), 17003–17016

    2022

  42. [50]

    Yifei Sun, Haoran Deng, Yang Yang, Chunping Wang, Jiarong Xu, Renhong Huang, Linfeng Cao, Yang Wang, and Lei Chen. 2022. Beyond Homophily: Structure- aware Path Aggregation Graph Neural Network. InProceedings of the Thirty-First International Joint Conference on Artificial Int...

    2022 doi

  43. [51]

    Jie Tang, Jimeng Sun, Chi Wang, and Zi Yang. 2009. Social influence analysis in large-scale networks. In Proceedings of the 15th ACM SIGKDD international conference on Knowledge discovery and data mining . 807–816

    2009

  44. [52]

    Puja Trivedi, Mark Heimann, Rushil Anirudh, Danai Koutra, and Jayaraman J Thiagarajan. 2024. Accurate and scalable estimation of epistemic uncertainty for graph neural networks. arXiv preprint arXiv:2401.03350 (2024)

    2024 arXiv

  45. [53]

    Petar Velickovic, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, and Yoshua Bengio. 2017. Graph attention networks. stat 1050 (2017), 20

    2017

  46. [54]

    Velickovic, W

    P. Velickovic, W. Fedus, W. L. Hamilton, P. Liò, A. Bojchevski, and A. Sørensen

  47. [55]

    Tao Wang, Di Jin, Rui Wang, Dongxiao He, and Yuxiao Huang. 2022. Power- ful graph convolutional networks with adaptive propagation mechanism for homophily and heterophily. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 36. 4210–4218

    2022

  48. [56]

    Xiyuan Wang and Muhan Zhang. 2022. How powerful are spectral graph neural networks. In International Conference on Machine Learning . PMLR, 23341–23362

    2022

  49. [57]

    John Winn, Christopher M Bishop, and Tommi Jaakkola. 2005. Variational message passing. Journal of Machine Learning Research 6, 4 (2005)

    2005

  50. [58]

    Lirong Wu, Haitao Lin, Yufei Huang, Tianyu Fan, and Stan Z Li. 2023. Extracting low-/high-frequency knowledge from graph neural networks and injecting it into mlps: An effective gnn-to-mlp distillation framework. In Proceedings of the AAAI Conference on Artificial Intelligence...

    2023

  51. [59]

    Ning Wu, Xin Wayne Zhao, Jingyuan Wang, and Dayan Pan. 2020. Learning effective road network representation with hierarchical graph neural networks. In Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining . 6–14

    2020

  52. [60]

    X. Wu, L. Song, and Others. 2019. Adversarial Examples on Graph Data: Deep Insights into Attack and Defense. Proceedings of the 28th International Joint Conference on Artificial Intelligence (IJCAI) (2019)

    2019

  53. [61]

    Xi and Others

    B. Xi and Others. 2024. A Comprehensive Benchmark for Adversarial Robustness of GNNs. Proceedings of the 38th Conference on Neural Information Processing Systems (NeurIPS) (2024)

    2024

  54. [62]

    Keyulu Xu, Weihua Hu, Jure Leskovec, and Stefanie Jegelka. 2018. How powerful are graph neural networks? arXiv preprint arXiv:1810.00826 (2018)

    2018 arXiv

  55. [63]

    Yuchen Yan, Yuzhong Chen, Huiyuan Chen, Minghua Xu, Mahashweta Das, Hao Yang, and Hanghang Tong. 2024. From trainable negative depth to edge heterophily in graphs. Advances in Neural Information Processing Systems 36 (2024)

    2024

  56. [64]

    Yujun Yan, Milad Hashemi, Kevin Swersky, Yaoqing Yang, and Danai Koutra

  57. [65]

    R. Ying, J. You, C. Morris, R. Xiong, W. Hamilton, and J. Leskovec. 2018. Hierar- chical Graph Representation Learning with Differentiable Pooling. Advances in Neural Information Processing Systems (NeurIPS) (2018)

    2018

  58. [66]

    Zhizhi Yu, Bin Feng, Dongxiao He, Zizhen Wang, Yuxiao Huang, and Zhiyong Feng. 2024. LG-GNN: local-global adaptive graph neural network for modeling both homophily and heterophily. In Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence . 2515–2523

    2024

  59. [67]

    Zhang, X

    J. Zhang, X. Zhou, R. Jia, D. Pei, and L. Song. 2020. GNNGuard: Defending Graph Neural Networks against Adversarial Attacks. Advances in Neural Information Processing Systems (NeurIPS) (2020)

    2020

  60. [68]

    Zhang and Others

    Q. Zhang and Others. 2023. RUNG: Robust Unbiased GNN with Adversarial Attacks via IRLS Aggregation. Proceedings of the 37th Conference on Neural Information Processing Systems (NeurIPS) (2023)

    2023

  61. [69]

    Zhao Zhang, Fuzhen Zhuang, Hengshu Zhu, Zhiping Shi, Hui Xiong, and Qing He

  62. [70]

    arXiv preprint arXiv:2102.06462 (2021)

    Two sides of the same coin: Heterophily and oversmoothing in graph convolutional neural networks. arXiv preprint arXiv:2102.06462 (2021)

    2021 arXiv

  63. [71]

    Xujiang Zhao, Feng Chen, Shu Hu, and Jin-Hee Cho. 2020. Uncertainty aware semi-supervised learning on graph data. Advances in Neural Information Process- ing Systems 33 (2020), 12827–12836

    2020

  64. [72]

    Xin Zheng, Miao Zhang, Chunyang Chen, Qin Zhang, Chuan Zhou, and Shirui Pan. 2023. Auto-heg: Automated graph neural network on heterophilic graphs. arXiv preprint arXiv:2302.12357 (2023)

    2023 arXiv

  65. [73]

    Jiong Zhu, Ryan A Rossi, Anup Rao, Tung Mai, Nedim Lipka, Nesreen K Ahmed, and Danai Koutra. 2021. Graph neural networks with heterophily. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 35. 11168–11176

    2021

  66. [74]

    Jiong Zhu, Yujun Yan, Lingxiao Zhao, Mark Heimann, Leman Akoglu, and Danai Koutra. 2020. Beyond homophily in graph neural networks: Current limitations and effective designs. Advances in Neural Information Processing Systems 33 (2020), 7793–7804

    2020

  67. [75]

    Zügner and S

    D. Zügner and S. Günnemann. 2019. Adversarial Attacks on Graph Neural Networks via Meta Learning.International Conference on Learning Representations (ICLR) (2019)

    2019

  68. [76]

    In Proceedings of the AAAI conference on artificial intelligence , Vol

    Relational graph neural network with hierarchical attention for knowledge graph completion. In Proceedings of the AAAI conference on artificial intelligence , Vol. 34. 9612–9619

  69. [77]

    Kai Zhao, Qiyu Kang, Yang Song, Rui She, Sijie Wang, and Wee Peng Tay

  70. [2009]

    InProceedings of the 26th Annual International Conference on Machine Learning

    PAC-Bayesian learning of linear classifiers. InProceedings of the 26th Annual International Conference on Machine Learning . 353–360

  71. [2019]

    International Conference on Learning Representations (ICLR) (2019)

    Deep Graph Infomax. International Conference on Learning Representations (ICLR) (2019)

    2019

  72. [2020]

    arXiv preprint arXiv:2002.05287 (2020)

    Geom-gcn: Geometric graph convolutional networks. arXiv preprint arXiv:2002.05287 (2020)

    2020 arXiv

  73. [2021]

    In Proceedings of the AAAI Conference on Artificial Intelligence , Vol

    Uncertainty-matching graph neural networks to defend against poisoning attacks. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 35. 9524–9532

  74. [2022]

    What makes graph neural networks miscalibrated? Advances in Neural Information Processing Systems 35 (2022), 13775–13786

    2022

  75. [2023]

    arXiv preprint arXiv:2305.16780 (2023)

    Graph neural convection-diffusion with heterophily. arXiv preprint arXiv:2305.16780 (2023)

    2023 arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.