Pith. sign in

REVIEW 3 major objections 7 minor 22 references

MetaGMT: Improving Actionable Interpretability of Graph Multilinear Networks via Meta-Learning Filtration

T0 review · 3 major / 7 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A meta-learning filtration loop makes Graph Multilinear Network explanations more faithful and more stable, with gains in explanation ROC of up to 8 points.

desk verdict The proposed meta-filtration loop is a no-op as written—the inner-loop loss has no dependence on the parameters it updates—so the empirical gains are unexplained and the central claim does not hold. read the letter →

arxiv 2505.19445 v1 pith:3VXTGERR submitted 2025-05-26 cs.LG

Rishabh Bhattacharya , Hari Shankar , Vaishnavi Shivkumar , Ponnurangam Kumaraguru This is my paper · ORCID
classification cs.LG
keywords graphneuralnetworksinterpretabilitymeta-learningbi-leveloptimizationspuriouscorrelationsexplanationfidelityMultilinearsubgraphexplanations
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 claims that the explanations produced by Graph Multilinear Networks (GMT), an inherently interpretable GNN architecture, can be made more faithful by adding a meta-learning filtration step. The proposed MetaGMT trains the explainer with a bi-level objective: an inner loop adapts a copy of the classifier on a local subgraph using the sparsity-promoting explanation loss, and an outer loop asks whether the adapted classifier still predicts the original graph correctly. Explanations that rely on spurious correlations fail this adaptation test and are therefore penalized. Across BA-2Motifs, MUTAG, and SP-Motif, the authors report higher explanation ROC and Precision@K than vanilla GMT, with lower variance across seeds and at most a small trade-off in classification accuracy on the spurious benchmark.

What carries the argument

The load-bearing mechanism is the bi-level meta-adaptation loop built on top of the Subgraph Multilinear Extension (SubMT) objective: the extractor produces edge attention scores, a functional copy of the classifier is adapted for N_inner steps on the subgraph using only the explanation (sparsity) loss L_expl, and the outer loss measures whether the adapted classifier still predicts the full graph's label. This filtration signal is carried by the gradients of the outer loss back through the inner updates via automatic differentiation, jointly updating the extractor and classifier. The key identity is that a good explanation should be transferable: if attention marks the true motif, adapting the classifier on the subgraph should not hurt full-graph prediction; if attention marks spurious structure, adaptation degrades the meta-prediction and is penalized.

What would settle it

Instrument the inner loop to print the gradient of L_expl with respect to the classifier parameters; if that gradient is identically zero (because L_expl depends only on the extractor and input features, not on the classifier), then the inner loop cannot change the classifier and the bi-level mechanism is inert, meaning any observed gains must come from the outer classification loss alone.

Watch

Extended reading notes

Core claim

MetaGMT augments the standard GMT training objective with a meta-learning filter that evaluates the usefulness of an explanation rather than only its sparsity. Concretely, the extractor's attention scores on a 1-hop subgraph around a randomly sampled node are used to run several inner-loop steps of classifier adaptation, minimizing only the information/sparsity loss. The adapted classifier is then applied to the full graph, and the classification loss on the true label provides the outer-loop signal for both the classifier and the extractor. The paper's central claim is that this bi-level procedure steers attention away from spuriously correlated edges, because such edges will not support successful adaptation from subgraph to full graph. The empirical claim is that MetaGMT-LIN and MetaGMT-SAM improve explanation AUC-ROC and Precision@5 on BA-2Motifs and SP-Motif (e.g., 80.16% vs 72.14% X-ROC on SP-Motif b=0.7, and 57.82% vs 48.66% X-Prec@5), while keeping classification accuracy competitive except at the highest bias level b=0.9, where MetaGMT-LIN accuracy drops to 54.53%.

Load-bearing premise

Minimizing the label-free sparsity loss on a 1-hop subgraph around one randomly sampled node, and then measuring full-graph classification after a few inner steps, is a reliable proxy for whether the attention marks causally relevant edges.

Editorial extensions

If this is right

  • MetaGMT-LIN and MetaGMT-SAM yield higher explanation AUC-ROC and Precision@5 than their GMT counterparts on BA-2Motifs and SP-Motif, with the largest gains in high-bias settings (about 8 points of X-ROC on SP-Motif b=0.7).
  • The variance across random seeds shrinks substantially (e.g., X-Prec@5 on BA-2Motifs from ±10.60 to ±3.79), so explanations become more reproducible across training runs.
  • Classification accuracy stays competitive on BA-2Motifs and MUTAG, and even improves to 99.50% on BA-2Motifs; the trade-off is concentrated in the most biased SP-Motif configuration.
  • The meta-objective provides a label-free, self-supervised signal for explanation quality, which could allow filtering spurious edges without requiring ground-truth motif labels beyond evaluation.

Reading between the lines

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

  • If the reported gains hold, the same bi-level filtration could be grafted onto other attention-based or mask-based GNN explainers, not just GMT, turning any sparse explainer into one that is explicitly selected for causal transferability.
  • The drop in classification accuracy at SP-Motif b=0.9 (66.90% to 54.53%) suggests the meta-filter may over-penalize edges that are predictive in the biased training distribution, which could be a testable failure mode for deployment in high-shift settings.
  • Because the inner loop samples one node per graph, the filtration signal is noisy; increasing the number of sampled nodes or using multiple subgraphs per outer step could either sharpen or wash out the reported gains, and this is a directly testable extension.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 7 minor

Summary. This paper proposes MetaGMT, an augmentation of the inherently interpretable Graph Multilinear Network (GMT) with a bi-level meta-learning objective. The claimed intuition (Section 4.1) is that an explanation is faithful when adapting the classifier on the k-hop subgraph attention improves full-graph prediction, so spurious edges should be filtered out because they fail to support such adaptation. The authors define an inner loop that adapts functional copies of the classifier with the subgraph information loss and an outer loop that computes the classification loss on the full graph, and they report X-ROC, X-Prec@5, and classification accuracy on BA-2Motifs, MUTAG, and SP-Motif (b = 0.5, 0.7, 0.9). The empirical tables show consistent X-ROC and X-Prec gains for MetaGMT-LIN on SP-Motif (e.g., +8.0 X-ROC at b = 0.7) with reduced variance across 10 seeds, at the cost of lower classification accuracy that the Appendix reports but the Abstract does not qualify.

Significance. The idea of filtering a model's internal attention by asking whether it supports classifier adaptation is reasonable and practically motivated, and the paper makes concrete falsifiable predictions: explanation-fidelity gains in high-bias regimes and reduced seed-to-seed variance. The experimental design has genuine strengths: hyperparameters are kept consistent between the GMT baselines and the MetaGMT variants, results are averaged over 10 fixed seeds, and the code is provided in an anonymous repository. The Appendix's honest flagging of the accuracy trade-off is also to the paper's credit. However, the contribution's value hinges entirely on the meta-learning mechanism being effective, and as written (Eqs. 4-5 and Algorithm 1) that mechanism is mathematically inert; the empirical improvements are therefore unexplained as written, and several headline claims ('significantly improves,' 'maintains competitive classification accuracy') outrun the reported evidence. If the mechanism were correctly specified, the experiments re-verified, and the significance claims supported by tests, this would be a solid, if incremental, contribution to inherently interpretable GNNs.

major comments (3)
  1. [§4.2, Eqs. (4)–(5); Algorithm 1] The stress-test concern is confirmed: as written, the inner-loop adaptation has no effect on the training objective, which removes the paper's central mechanism. L_expl = L_info(A_sub) (Eq. 4) depends only on the extractor's attention scores A_sub, computed by M_ext(θ_ext) from the subgraph's initial node representations; the extractor parameters θ_ext are held fixed during the inner loop (Section 4.2, step 1), and Algorithm 1 computes A_sub before the loop begins. Nothing in L_expl therefore depends on the adapted classifier parameters φ_clf, so ∇_{φ_clf} L_expl = 0 for every inner step, φ*_clf = θ_clf, and the outer loss reduces to the ordinary first-order classification loss L_pred(M_clf(θ_clf)(G), y). No filtration signal from the inner loop ever reaches the extractor or the classifier. A further consequence is that the sparsity loss L_info is never optimized at all, so MetaGMT as written is equivalent to GMT training without its sparsity regularizer, and any differences from the baseline would be attributable to the missing regularizer rather than to meta-learning. This is compounded by an internal inconsistency: Section 4.1's rationale and the Figure 1 caption state that the inner loop trains the model to predict the correct class, which describes a prediction loss that Eqs. (4)–(5) never define. The authors must either make explicit how A_sub depends on φ_clf through the initial node representations, or replace Eq. (4) with a classifier-dependent inner loss, and must reconcile Algorithm 1 with the figure's description.
  2. [Appendix A.1, Table 5; Abstract; §6] The claim that MetaGMT 'maintains competitive classification accuracy' is contradicted by the paper's own accuracy table. At SP-Motif b = 0.9, MetaGMT-LIN reaches 54.53% versus 66.90% for GMT-LIN, a drop of 12.4 points; at b = 0.5 and b = 0.7 the drops are 3.5 and 4.4 points, and MUTAG also declines (91.56% to 91.12%). This is inconsistent with Section 6's statement that 'improvements in interpretability come with either preserved or slightly improved predictive performance,' and the trade-off is acknowledged only in the Appendix, whose own caption notes 'potential trade-offs between classification performance and explanation quality.' The Abstract and Section 6 must be revised to qualify the accuracy claim to the settings where it holds, and the classification accuracy table should appear in the main text.
  3. [§5.2, Tables 1–4] The claimed 'significant improvements' in explanation quality are not supported by any statistical test. Several headline gaps are within one standard deviation of the baselines (e.g., BA-2Motifs X-ROC: 98.37 ± 1.08 versus 97.97 ± 1.46; X-Prec@5: 91.96 ± 3.79 versus 88.58 ± 10.60), and MetaGMT is worse than GMT on MUTAG for both X-ROC (92.60 versus 93.13) and X-Prec@5 (28.55 versus 28.66) as well as for GMT-SAM on MUTAG (93.58 versus 94.79). The variance-reduction claims also need explicit testing. Since both methods are averaged over the same 10 seeds (0–9), paired tests (paired t-test or Wilcoxon signed-rank on per-seed metrics, and an appropriate test on the variances) are straightforward and should be reported before the word 'significant' is used in the Abstract or in Section 5.2.
minor comments (7)
  1. [Table 4] The GMT-SAM MUTAG entry '2 8.65±1.43' contains a stray space and should read '28.65±1.43'; the header 'SPURIOUS-M OTIF' in Table 2 and 'SP-Motif(B= 0.5)' in Tables 3–4 should also be formatted consistently.
  2. [§5.1; References] The SP-Motif dataset is attributed to 'Wang et al., Learning invariant graph representations via virtual environment inference,' but the canonical source of the SP-Motif benchmark is Wu et al., 'Handling Distribution Shifts on Graphs: An Invariance Perspective' (ICLR 2022); the listed citation is incomplete and points to a different title. The reference for Veličković et al. also lacks a year and venue.
  3. [Appendix A.2, Table 7] MUTAG is configured with 'Use Edge Attention: No,' yet Tables 1–4 report edge-based explanation metrics for MUTAG; please clarify how attention scores are produced for explanations when the classifier does not consume edge attention.
  4. [§A.2, Tables 6–7] The meta-parameters (α = 0.01, N_inner = 3, k = 1) are fixed across all datasets with no sensitivity analysis or ablation; even a small ablation over N_inner and α would substantiate the claim that the depth of the inner loop drives the reported gains.
  5. [Abstract; Table 1] The Abstract's 'up to 8% of Explanation ROC on SP-Motif 0.5' does not match Table 1: the 8.0-point gain occurs at b = 0.7 (80.16 versus 72.14), while at b = 0.5 the gain is 7.4 points; please align the Abstract with the table.
  6. [§6] The meta-objective is called 'self-supervised,' but the outer loss uses ground-truth labels; only the inner sparsity loss is label-free, so the terminology should be clarified.
  7. [§5.2] The Abstract's phrase 'compared to baseline methods' overstates the comparison, since only the two GMT variants being augmented (GMT-LIN and GMT-SAM) serve as baselines; adding at least one post-hoc explainer (e.g., GNNExplainer) or a non-meta inherently interpretable baseline (e.g., GSAT) would strengthen the claim.

Circularity Check

1 steps flagged · score 7.0 of 10

Inner-loop adaptation is a mathematical no-op: Lexpl depends only on extractor outputs, so the claimed meta-filtration reduces to standard GMT by construction.

  1. self definitional [Section 4.2, Eq. 4-5; Algorithm 1 lines 9-10]
    "An explanation loss Lexpl is defined on the subgraph, specifically using the information loss (Eq. 3) applied to the subgraph attention scores: Lexpl(Asub)=Linfo(Asub) (4) ... Update ϕ(n+1)clf ← Optinner(ϕ(n)clf, ∇ϕ(n)clf Lexpl)."

    Eq. 4 defines Lexpl solely in terms of Asub, which is produced by the fixed extractor Mext(θext) from the initial node embeddings. It has no dependence on the classifier parameters ϕ(n)clf being adapted. Therefore ∇_{ϕ(n)clf} Lexpl(Asub)=0 at every inner step, so Eq. 5 and Algorithm 1 line 10 leave ϕ(n)clf unchanged and ϕ*clf = θclf exactly. The outer 'meta-prediction' ŷmeta = Mclf(ϕ*clf)(G) then coincides with the ordinary GMT prediction Mclf(θclf)(G), and the gradients gclf, gext are just the gradients of the standard classification loss. The proposed meta-filtration signal—penalizing explanations that fail to support classifier adaptation—never enters the objective.

full rationale

MetaGMT does not use ground-truth explanation labels during training, so its reported X-ROC and X-Prec@5 values are not fitted to the benchmark targets; in that respect the fidelity evaluation is self-contained. There is no load-bearing self-citation. However, the paper's central bi-level filtration mechanism is formally inert as written: Eq. 4 defines Lexpl solely on Asub, which depends only on the extractor, so the inner-loop gradient with respect to the classifier parameters vanishes, the adapted classifier equals the original classifier, and the outer loss is the standard GMT classification loss. The claimed meta-filtration therefore reduces by construction to ordinary GMT training. Any observed differences from GMT in the experimental tables cannot be attributed to the stated algorithm. This is a definitional reduction of the contribution's mechanism, warranting a high circularity score, even though the empirical comparisons themselves are not circular.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The central claim rests on an unproven proxy that inner-loop adaptation on a sparsity-only loss measures explanation faithfulness, plus a sampling assumption that a single 1-hop neighborhood is informative for graph-level tasks. Hyperparameters are hand-set or validation-tuned, and the method relies on higher-order gradient software without analysis.

free parameters (5)
  • Inner learning rate alpha = 0.01
    Chosen by hand; scales the inner-loop classifier adaptation that drives the filter.
  • Inner steps N_inner = 3
    Chosen by hand; controls how strongly the classifier adapts to the subgraph explanation before the outer loss.
  • Subgraph hops k = 1
    Chosen by hand; determines whether the sampled local context includes the predictive motif.
  • Sparsity schedule target r = 1.0 to 0.5 or 0.7 with decay -0.1 every 10 epochs
    Hand-set per dataset; directly shapes attention sparsity and thus the explanation mask.
  • Loss weights and outer learning rate = lambda_pred=1, lambda_info=1; lr 1e-3 or 3e-3
    Following GMT defaults or tuned slightly on a validation set.
assumptions (3)
  • ad hoc to paper An explanation is faithful if adapting the classifier with the sparsity loss on the subgraph improves full-graph prediction.
    Core heuristic in Section 4.1; no proof or ablation, and it is the mechanism that is supposed to penalize spurious edges.
  • domain assumption A uniform random node's 1-hop subgraph contains enough decision-relevant edges for the inner adaptation to be informative.
    Section 4.2 samples one node and uses k=1; in graph-level tasks the motif may not be in that neighborhood.
  • domain assumption The higher library's gradients through N_inner inner steps are finite and provide a useful signal for both classifier and extractor.
    Algorithm 1 and Eq. 7 rely on this; no gradient statistics or stability analysis are given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MetaGMT: Improving Actionable Interpretability of Graph Multilinear Networks via Meta-Learning Filtration." pith.science (2026). https://pith.science/paper/3VXTGERR

@misc{pith2026250519445,
  author       = {Pith},
  title        = {Pith review of: MetaGMT: Improving Actionable Interpretability of Graph Multilinear Networks via Meta-Learning Filtration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3VXTGERR}},
  note         = {Machine review of arXiv:2505.19445}
}
read the original abstract

The growing adoption of Graph Neural Networks (GNNs) in high-stakes domains like healthcare and finance demands reliable explanations of their decision-making processes. While inherently interpretable GNN architectures like Graph Multi-linear Networks (GMT) have emerged, they remain vulnerable to generating explanations based on spurious correlations, potentially undermining trust in critical applications. We present MetaGMT, a meta-learning framework that enhances explanation fidelity through a novel bi-level optimization approach. We demonstrate that MetaGMT significantly improves both explanation quality (AUC-ROC, Precision@K) and robustness to spurious patterns, across BA-2Motifs, MUTAG, and SP-Motif benchmarks. Our approach maintains competitive classification accuracy while producing more faithful explanations (with an increase up to 8% of Explanation ROC on SP-Motif 0.5) compared to baseline methods. These advancements in interpretability could enable safer deployment of GNNs in sensitive domains by (1) facilitating model debugging through more reliable explanations, (2) supporting targeted retraining when biases are identified, and (3) enabling meaningful human oversight. By addressing the critical challenge of explanation reliability, our work contributes to building more trustworthy and actionable GNN systems for real-world applications.

Figures

Figures reproduced from arXiv: 2505.19445 by the authors.

Figure 1
Figure 1. The complete pipeline of MetaGMT. We select a node in the graph and extract its k-hop neighbourhood. Then, we use the explainer to find the attention scores. Using the explanation loss defined in 3 and 4, We iteratively train our model (Inner Loop) to predict the correct class. This prediction is compared to ground truth (Outer Loop), and the loss calculated is used to update our explainer and our model. implicitly … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 18 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    Evaluating explainability for graph neural networks

    Agarwal, C., Lakkaraju, H., and Zitnik, M. Evaluating explainability for graph neural networks. In Advances in Neural Information Processing Systems (NeurIPS), volume 35, pp.\ 35528--35541, 2022

    work page 2022

  3. [3]

    How Interpretable Are Interpretable Graph Neural Networks?

    Chen, Y., Bian, Y., Han, B., and Cheng, J. How interpretable are interpretable graph neural networks? In International Conference on Machine Learning (ICML), 2024. URL https://arxiv.org/abs/2406.07955

    work page Pith review arXiv 2024

  4. [4]

    and Shen, Y

    Deng, Z. and Shen, Y. Self-interpretable graph learning with sufficient and necessary explanations. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pp.\ 16721--16729, 2023

    work page 2023

  5. [5]

    Explainable graph neural networks: A survey

    Dwivedi, R., Yuan, H., Yu, S., Tan, C., and Luo, J. Explainable graph neural networks: A survey. ACM Transactions on Knowledge Discovery from Data, 17 0 (8): 0 1--50, 2023

    work page 2023

  6. [6]

    Graph neural networks for social recommendation

    Fan, W., Ma, Y., Li, Q., He, Y., Zhao, E., Tang, J., and Yin, D. Graph neural networks for social recommendation. In The World Wide Web Conference, pp.\ 417--426, 2019

    work page 2019

  7. [7]

    and Lenssen, J

    Fey, M. and Lenssen, J. E. Fast graph representation learning with pytorch geometric. In ICLR Workshop on Representation Learning on Graphs and Manifolds, 2019

    work page 2019

  8. [8]

    Utilising graph machine learning within drug discovery and development

    Gaudelet, T., Mallet, V., Mineva, T., et al. Utilising graph machine learning within drug discovery and development. Briefings in Bioinformatics, 22 0 (6): 0 bbab159, 2021

    work page 2021

Show all 22 references
  1. [9]

    Kipf, T. N. and Welling, M. Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907, 2016

    2016 arXiv

  2. [10]

    Parameterized explainer for graph neural network

    Luo, D., Cheng, W., Xu, D., Yu, W., Zong, B., Chen, H., and Zhang, X. Parameterized explainer for graph neural network. In Advances in Neural Information Processing Systems (NeurIPS), volume 33, pp.\ 19708--19719, 2020

    2020

  3. [11]

    Miao, S., Pan, S., Zhao, X., Zheng, Y., Wang, X., and Yu, P. S. Interpretable and generalizable graph learning via stochastic attention mechanism. In International Conference on Machine Learning (ICML), pp.\ 15678--15691. PMLR, 2022

    2022

  4. [12]

    E., Kolouri, S., Rostami, M., Martin, C

    Pope, P. E., Kolouri, S., Rostami, M., Martin, C. E., and Hoffmann, H. Explainability methods for graph convolutional neural networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp.\ 10772--10781, 2019

    2019

  5. [13]

    A meta-learning approach for training explainable graph neural networks

    Spinelli, I., Scardapane, S., and Uncini, A. A meta-learning approach for training explainable graph neural networks. IEEE Transactions on Neural Networks and Learning Systems, 2021. URL https://arxiv.org/abs/2109.09426

    2021 arXiv

  6. [14]

    and Pratt, L

    Thrun, S. and Pratt, L. Learning to learn. Kluwer Academic Publishers, 1998

    1998

  7. [15]

    Graph attention networks

    Veli c kovi \'c , P., Cucurull, G., Casanova, A., Romero, A., Lio, P., and Bengio, Y. Graph attention networks. In International Conference on Learning Representations (ICLR)

  8. [16]

    and Thai, M

    Vu, M. and Thai, M. T. Pgm-explainer: Probabilistic graphical model explanations for graph neural networks. In Advances in Neural Information Processing Systems (NeurIPS), volume 33, pp.\ 12039--12050, 2020

    2020

  9. [17]

    Learning invariant graph representations via virtual environment inference

    Wang, Y., Chen, Y., Huang, J., Liu, Z., Zhang, J., Cheng, X., Li, Y., and Zhong, Z. Learning invariant graph representations via virtual environment inference

  10. [18]

    How powerful are graph neural networks? In International Conference on Learning Representations (ICLR), 2019

    Xu, K., Hu, W., Leskovec, J., and Jegelka, S. How powerful are graph neural networks? In International Conference on Learning Representations (ICLR), 2019

    2019

  11. [19]

    Graph neural network for fraud detection: A review and prospect

    Yang, J., Liu, Z., Zhao, W., Zhang, Z., and Cui, B. Graph neural network for fraud detection: A review and prospect. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management, pp.\ 4978--4982, 2022

    2022

  12. [20]

    GNNExplainer : Generating explanations for graph neural networks

    Ying, R., Bourgeois, D., You, J., Zitnik, M., and Leskovec, J. GNNExplainer : Generating explanations for graph neural networks. In Advances in Neural Information Processing Systems (NeurIPS), volume 32, 2019

    2019

  13. [21]

    On explainability of graph neural networks via subgraph explorations

    Yuan, H., Yu, H., Gui, S., and Ji, S. On explainability of graph neural networks via subgraph explorations. In International Conference on Machine Learning (ICML), pp.\ 12241--12252. PMLR, 2021

    2021

  14. [22]

    Hierarchical graph pooling with structure learning

    Zhang, Z., Bu, J., Ester, M., Zhang, J., Yao, C., Yu, Z., and Wang, C. Hierarchical graph pooling with structure learning. arXiv preprint arXiv:1911.05954, 2019

    1911 arXiv

Pith tools

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