Pith. sign in

REVIEW 2 major objections 6 minor 28 references

This paper shows that a graph neural network built from polynomial layers can be attributed exactly by Aumann–Shapley path integration with a fixed, architecture-determined number of evaluation points.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 08:28 UTC pith:UVBGDRBQ

load-bearing objection The exact-integration theory is sound; the headline fidelity result is a confounded system comparison. the 2 major comments →

arxiv 2607.21094 v1 pith:UVBGDRBQ submitted 2026-07-23 cs.LG

A Polynomial Architecture-Attribution Co-Design Framework for Exact Aumann-Shapley Attribution in GNNs

Bizu Feng , Zhimu Yang , Shuming Wang , Shaode Yu , Yuan Cheng , Xiaojun Qian , Zixin Hu This is my paper
classification cs.LG
keywords graph neural networksAumann-Shapley attributionIntegrated Gradientsexact quadratureGauss-Legendre quadraturepolynomial networksmodel-attribution co-designexplainability
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper argues that the numerical approximation error in path-based attribution of graph neural network predictions can be eliminated by designing the network so that its input-to-logit map is a bounded-degree polynomial. It introduces PolyGIN, a graph-isomorphism-style architecture whose layers use only linear operations and a quadratic polynomial activation, and proves that an L-block PolyGIN has output degree at most 2^L. The derivative along the straight-line baseline-to-input path therefore has degree at most 2^L − 1, so 2^(L−1) Gauss–Legendre points evaluate the Aumann–Shapley integral exactly. Feature attributions aggregate into signed node attributions that sum to the logit difference, giving completeness up to floating-point precision. Experiments on five graph benchmarks show PolyGIN matches a standard graph isomorphism network closely in accuracy while the attribution procedure reaches machine-precision completeness with eight evaluations for a four-block network.

Core claim

The central claim is Theorem 1: for a PolyGIN with L polynomial transformation blocks, the scalar logit is a multivariate polynomial in the input features of degree at most 2^L; hence each partial derivative along the straight-line path has degree at most 2^L − 1. Standard Gauss–Legendre quadrature with m = 2^(L−1) points integrates such a univariate polynomial exactly. The result converts integrated-gradients-style attribution from a resolution-dependent numerical approximation into a fixed finite computation with a deterministic evaluation budget, while additive node-level aggregation (summing feature attributions per node) preserves the completeness identity that the feature attributions

What carries the argument

PolyGIN, a message-passing network built so that all steps are linear except a quadratic polynomial activation z + θ⊙z²; graph-isomorphism-style sum aggregation, linear projections, and a learnable diagonal scaling are linear, so compositions keep the input-to-logit map polynomial. The load-bearing identity is the Gauss–Legendre exactness property: an m-point rule integrates polynomials of degree up to 2m − 1 exactly. Combined with the degree bound for PolyGIN, this fixes m = 2^(L−1) and eliminates quadrature truncation error.

Load-bearing premise

Every operation that touches the attributed score must be polynomial or linear in the input features, and the attribution must be taken on the pre-softmax logit along a fixed straight-line path from a chosen baseline; introduce a ReLU, batch norm, softmax, or attention softmax anywhere and the degree bound—and with it the exactness guarantee—no longer holds.

What would settle it

Train a four-block PolyGIN, compute APEX attributions with the prescribed eight Gauss–Legendre points, and compare against a high-resolution numerical integration (say 512-point Gauss–Legendre or Simpson) on the same model. If the two attribution vectors differ by more than floating-point roundoff, or if the completeness error exceeds machine precision for any input, Theorem 1 fails for that model. A simpler check: insert one ReLU activation into the last block and measure the completeness error of an eight-point rule—it should jump well above FP32 precision.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • For any fixed-depth PolyGIN, exact Aumann–Shapley attribution costs exactly 2^(L−1) forward-backward evaluations, independent of any user-chosen integration resolution.
  • Completeness holds to floating-point precision: feature attributions sum to the logit difference, and signed node-level attributions decompose a prediction into supporting and opposing evidence.
  • No convergence sweep or error tolerance is needed; truncation error from path discretization is eliminated under the polynomial assumptions.
  • The polynomial constraint does not cost much accuracy: on the five benchmarks tested, PolyGIN stays within about one percentage point of a standard graph isomorphism network.
  • Fidelity experiments (necessary and sufficient subsets) favor the complete framework over several post-hoc explainers on the evaluated datasets.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The co-design principle could transfer to other continuous attribution frameworks: any path integral whose integrand is polynomial in the path parameter inherits an exact finite quadrature budget, so other baselines or paths may be handled with adjusted degree bounds.
  • The exponential degree growth (2^L) caps practical depth; the paper uses four blocks, and deeper polynomial networks may need additional scaling or regularization to remain stable.
  • The exactness guarantee is tied to pre-softmax logits; users attributing probabilities or using batch norm, ReLU, or attention softmax must accept the same quadrature error the framework removes. A testable extension is to design polynomial analogues of attention or normalization that keep the degree bound.
  • The nearly identical fidelity of numerical integrated gradients and APEX on the same PolyGIN backbone suggests the fidelity gain over other explainers comes from the architecture-plus-attribution combination rather than from exactness alone; the paper argues this in an appendix.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 6 minor

Summary. The paper proposes APEX, a co-design framework pairing a polynomial GNN architecture (PolyGIN) with exact Aumann–Shapley attribution. PolyGIN replaces non-polynomial components with linear operations and the polynomial activation σ_poly(z)=z+θ⊙z², so the pre-softmax logit is a multivariate polynomial in the input features. Theorem 1 gives a degree bound of 2^L for the logit, hence a degree bound of 2^L−1 for the path integrand; Gauss–Legendre quadrature with m=2^{L−1} points then evaluates the attribution integral exactly. Feature-level attributions are summed into signed node-level scores with completeness preserved. Experiments on five graph benchmarks compare predictive accuracy, fidelity, completeness error, and runtime. The theoretical derivation is self-contained and the induction in Theorem 1 is correct; the main weakness is the empirical fidelity comparison, which does not clearly control the attributed backbone.

Significance. If the claims hold, the paper contributes a clean and useful result: a class of GNNs for which Integrated-Gradients-style path integrals are evaluated exactly with a fixed, architecture-determined number of forward–backward evaluations. The proof is elementary but rigorous, and the empirical completeness-error experiment (Figure 5) directly confirms the predicted 8-point exactness for a 4-block PolyGIN. The paper also usefully separates algebraic exactness from 1-WL expressivity in Appendix B. However, the practical relevance advertised in the abstract depends on the fidelity experiments, and those experiments are currently confounded by inadequate reporting of the backbone used for each explainer. The scope is also narrower than the title suggests: exactness holds only for polynomial architectures and pre-softmax logits, a limitation the authors do acknowledge in the conclusion.

major comments (2)
  1. [§4.3, Figures 3–4, Appendix D, Figures 6–7] The main fidelity comparison does not state which trained backbone each explainer is evaluated on. Table 1 reports only GIN vs. PolyGIN accuracy, and §4.3 merely says 'all methods' are evaluated on correctly classified test graphs. Appendix D reveals that when all explainers are evaluated on the same trained PolyGIN models, numerical IG achieves fidelity 'nearly identical' to APEX. This strongly implies that the headline Figures 3–4 compare APEX on PolyGIN against baselines on GIN. If so, the claimed fidelity advantage may be due to the architecture, not to exact Gauss–Legendre quadrature. Please present the same-backbone comparison as the primary fidelity result, or explicitly reframe Figures 3–4 as a system-level comparison of (PolyGIN+APEX) versus (GIN+baseline) and add a table stating the backbone used for every method.
  2. [§3.1, §4.3, Eqs. (10), (13), (14)] APEX attributes the pre-softmax logit f_c, but the fidelity metrics are computed on softmax probabilities p_c after masking/retaining nodes. The paper never justifies that a positive attribution to the logit f_c implies a corresponding change in p_c, since p_c depends on all logits through the softmax denominator. The monotonic link may hold approximately in practice but is not guaranteed. This mismatch is load-bearing for the empirical fidelity claim: an exact logit decomposition can rank nodes differently from a probability-based fidelity protocol. Please either report fidelity based on the attributed logit as well, or state and justify the assumption that logit and probability rankings agree under the perturbation protocol.
minor comments (6)
  1. [Title and Abstract] The title and abstract claim 'exact Aumann–Shapley attribution in GNNs,' but the exactness is conditional on the polynomial architecture and on attributing pre-softmax logits. The conclusion states this explicitly. Please qualify the title/abstract, e.g., 'in polynomial GNNs' or 'under polynomial architectures.'
  2. [§4.4, Eq. (9)] The notation 'm=2^{L−1}' and the degree bound '2^L−1' are visually confusable in plain text, and several sentences render '2 L−1' without superscripts. Use consistent typesetting, e.g., m=2^{L-1} and deg = 2^L-1.
  3. [Figures 3–4] No variance/error bars are shown for the fidelity results, even though Table 1 reports ten seeds. State whether the fidelity curves are averages over seeds and report standard deviations or at least a seed count.
  4. [Table 3] The Mutagenicity label mapping in Table 3 (Label 1 = Non-mutagenic, Label 0 = Mutagenic) is the opposite of the common convention for this dataset. Please verify the encoding and explicitly define the labels used in the visualization; otherwise the qualitative discussion of 'mutagenic' vs. 'non-mutagenic' examples may mislead readers.
  5. [Appendix D] The same-backbone control is important and currently appears only in an appendix. Move a condensed version into the main text or at least add a sentence in §4.3 stating explicitly that a same-backbone control is provided and what it shows.
  6. [§5] There is a duplicated phrase in the final sentence of the completeness-error discussion: '...consistent with floating-point roundoff rather than quadrature truncation error, rather than from approximation error...' Remove the redundancy.

Circularity Check

0 steps flagged

No significant circularity: the APEX exactness result is derived from architecture definitions and classical Gauss-Legendre quadrature, not from its own outputs.

full rationale

The central derivation is self-contained and non-circular. Theorem 1 is a direct induction on Definitions 1-3: aggregation, linear projection, and PolyScaleNorm are linear; PolyActivation z + theta⊙z^2 at most doubles degree, so after L blocks the logit has degree at most 2^L and any partial derivative has degree at most 2^L-1. The quadrature budget m = 2^{L-1} follows from the classical Gauss-Legendre exactness condition (m points integrate degree at most 2m-1), and Eq. (10) is just the mapped GL rule applied to Eq. (2). Node-level completeness (Eq. 12) is summation of feature attributions. Nothing is fitted and then called a prediction: the 8-point result in Fig. 5 is computed from the stated L=4 and checked against FP32 completeness error, not used to set a constant. The paper cites no prior work by its own authors, so no self-citation chain is load-bearing; the Hesse et al. and Gauss-Legendre citations are external standard results. Appendix D provides a same-backbone control and notes that numerical IG on PolyGIN is 'nearly identical' to APEX, which is a consistency check of the same Aumann-Shapley quantity, not circularity. The Conclusion explicitly limits exactness to polynomial logits and excludes ReLU/softmax/batchnorm operations, showing the claim is conditional rather than definitionally forced to fit observed data. The possible fidelity-comparison confound between different backbones is an experimental-design concern, not a circularity in the derivation.

Axiom & Free-Parameter Ledger

0 free parameters · 5 axioms · 3 invented entities

The central derivation is clean and self-contained: no free parameter is fitted to make the degree bound hold, and the only inputs are the polynomial architecture, standard Gauss–Legendre exactness, and the chain rule. The new constructs are explicit architecture components, not postulated entities invoked to save the result.

axioms (5)
  • standard math An m-point Gauss–Legendre rule exactly integrates every polynomial of degree at most 2m−1.
    Used in §3.4 to set m=2^{L−1}; a standard result cited in the paper.
  • domain assumption The attributed model output is the pre-softmax logit, with no non-polynomial output layer in the path being integrated.
    Stated in §3.1; the exactness guarantee explicitly excludes softmax probabilities.
  • domain assumption All PolyGIN operations (GIN sum aggregation, linear maps, PolyScaleNorm, PolyActivation) preserve polynomial dependence on input features, for each fixed graph topology.
    Stated in §3.2 and Definition 3; required for the degree bound in Theorem 1.
  • domain assumption Attribution is taken along the straight-line path from a baseline input to the target input.
    Eq. (1)–(2); a different path would define a different integral, although the algebraic argument would be similar.
  • domain assumption For the WL expressivity analysis: countable input feature universe, bounded node degrees, layer-wise injectivity (Assumption 1), and injective graph-level readout.
    Appendix B; this is a conditional expressivity result and is not needed for the exactness guarantee.
invented entities (3)
  • PolyGIN independent evidence
    purpose: A GIN-style polynomial GNN architecture whose scalar output is a bounded-degree polynomial in the input features, enabling exact quadrature.
    The architecture is fully specified, and its degree bound (Theorem 1) makes a falsifiable prediction confirmed in Fig. 5: 8 evaluation points drive completeness error to FP32 precision for L=4.
  • PolyActivation σ_poly(z)=z+θ⊙z² no independent evidence
    purpose: Polynomial nonlinearity that preserves the degree bound while introducing feature interactions.
    Internal design component defined by Eq. (3); its behavior is fixed by definition and has no separate external validation.
  • PolyScaleNorm no independent evidence
    purpose: Learnable linear scaling intended to mitigate exponential magnitude growth without breaking polynomial form.
    Design component; the paper states it does not by itself guarantee numerical stability, so independent evidence is absent.

pith-pipeline@v1.3.0-alltime-deepseek · 19240 in / 13767 out tokens · 117154 ms · 2026-08-01T08:28:23.391395+00:00 · methodology

0 comments
read the original abstract

We study feature-level and node-level explanations for graph neural networks (GNNs) through the lens of Aumann-Shapley attribution. Path-integral methods such as Integrated Gradients provide an axiomatic formulation of attribution, but their practical use in deep GNNs typically relies on finite-sample numerical approximations to the path integral, requiring a trade-off between quadrature error and computational cost. This paper proposes APEX, a model-attribution co-design framework that makes the attribution integral exactly computable under a polynomial GNN architecture. The key component is PolyGIN, a GIN-style graph network whose message-passing, normalization, and transformation operations preserve a bounded multivariate polynomial form for scalar model scores, such as pre-softmax logits. We show that, for a PolyGIN with $L$ polynomial transformation blocks, the derivative along the attribution path has degree at most $2^L-1$. Therefore, Gauss--Legendre quadrature can evaluate the Aumann--Shapley path integral exactly, up to floating-point precision, with $2^{L-1}$ deterministic evaluation points. The resulting attributions can be computed at the feature level and then aggregated into node-level scores while preserving completeness. Experiments on synthetic and real-world graph benchmarks show that PolyGIN maintains competitive predictive performance, while the complete APEX framework achieves higher attribution fidelity than the compared baselines and substantially reduces the number of evaluations required for path integration.

Figures

Figures reproduced from arXiv: 2607.21094 by Bizu Feng, Shaode Yu, Shuming Wang, Xiaojun Qian, Yuan Cheng, Zhimu Yang, Zixin Hu.

Figure 1
Figure 1. Figure 1: Motivation of APEX. Standard path-integral attribution methods approximate the integral with numerical [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the APEX computation. PolyGIN constrains the model output to a bounded polynomial space. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Necessary explanation comparison measured by Fidelity [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Sufficient explanation comparison measured by Fidelity [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Completeness error with respect to the number of forward evaluations (NFE). Numerical integration base [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Fidelity+ comparison when all explainers are evaluated on the same PolyGIN models. 0.5 0.6 0.7 0.8 0.9 Sparsity 0.00 0.10 0.20 Fidelity￾BBBP 0.5 0.6 0.7 0.8 0.9 Sparsity 0.00 0.10 0.20 0.30 Fidelity￾BACE 0.5 0.6 0.7 0.8 0.9 Sparsity 0.00 0.02 0.04 Fidelity￾Graph-SST2 0.5 0.6 0.7 0.8 0.9 Sparsity 0.00 0.10 0.20 0.30 Fidelity￾Mutagenicity 0.5 0.6 0.7 0.8 0.9 Sparsity 0.20 0.40 0.60 0.80 Fidelity￾BA-shapes AP… view at source ↗
Figure 7
Figure 7. Figure 7: Fidelity− comparison when all explainers are evaluated on the same PolyGIN models. Figures 6 and 7 show that numerical Integrated Gradients achieves fidelity results nearly identical to those of APEX when both are applied to PolyGIN. This agreement is theoretically expected because APEX and numerical IG tar￾get the same Aumann–Shapley attribution, although they evaluate the corresponding path integral diff… view at source ↗
Figure 8
Figure 8. Figure 8: Average runtime for generating one graph explanation. [PITH_FULL_IMAGE:figures/full_fig_p019_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Qualitative visualization of APEX signed node attributions on BA-Shapes. The target node is marked by a [PITH_FULL_IMAGE:figures/full_fig_p020_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Token-level signed attribution visualizations of APEX on Graph-SST2. All examples shown are correctly [PITH_FULL_IMAGE:figures/full_fig_p021_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Signed atom-level attributions on molecular property prediction datasets. Rows correspond to BBBP, [PITH_FULL_IMAGE:figures/full_fig_p022_11.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

28 extracted references · 1 linked inside Pith

  1. [1]

    Graph neural networks in recommender systems: a survey.ACM Computing Surveys, 55(5):1–37, 2022

    Shiwen Wu, Fei Sun, Wentao Zhang, Xu Xie, and Bin Cui. Graph neural networks in recommender systems: a survey.ACM Computing Surveys, 55(5):1–37, 2022

    2022

  2. [2]

    Graph neural networks for social recommendation

    Wenqi Fan, Yao Ma, Qing Li, Yuan He, Eric Zhao, Jiliang Tang, and Dawei Yin. Graph neural networks for social recommendation. InThe world wide web conference, pages 417–426, 2019

    2019

  3. [3]

    Graph convolutional networks for text classification

    Liang Yao, Chengsheng Mao, and Yuan Luo. Graph convolutional networks for text classification. InProceedings of the AAAI conference on artificial intelligence, volume 33, pages 7370–7377, 2019

    2019

  4. [4]

    Dejun Jiang, Zhenxing Wu, Chang-Yu Hsieh, Guangyong Chen, Ben Liao, Zhe Wang, Chao Shen, Dongsheng Cao, Jian Wu, and Tingjun Hou. Could graph neural networks learn better molecular representation for drug dis- covery? a comparison study of descriptor-based and graph-based models.Journal of cheminformatics, 13(1):12, 2021. 10 APEX

    2021

  5. [5]

    Neural message passing for quantum chemistry

    Justin Gilmer, Samuel S Schoenholz, Patrick F Riley, Oriol Vinyals, and George E Dahl. Neural message passing for quantum chemistry. InInternational conference on machine learning, pages 1263–1272. Pmlr, 2017

    2017

  6. [6]

    Axiomatic attribution for deep networks

    Mukund Sundararajan, Ankur Taly, and Qiqi Yan. Axiomatic attribution for deep networks. InInternational conference on machine learning, pages 3319–3328. PMLR, 2017

    2017

  7. [7]

    Princeton University Press, 2015

    Robert J Aumann and Lloyd S Shapley.Values of non-atomic games. Princeton University Press, 2015

    2015

  8. [8]

    The many shapley values for model explanation

    Mukund Sundararajan and Amir Najmi. The many shapley values for model explanation. InInternational conference on machine learning, pages 9269–9278. PMLR, 2020

    2020

  9. [9]

    Explainability in graph neural networks: A taxonomic survey.IEEE transactions on pattern analysis and machine intelligence, 45(5):5782–5799, 2022

    Hao Yuan, Haiyang Yu, Shurui Gui, and Shuiwang Ji. Explainability in graph neural networks: A taxonomic survey.IEEE transactions on pattern analysis and machine intelligence, 45(5):5782–5799, 2022

    2022

  10. [10]

    Gnnexplainer: Generating explanations for graph neural networks.Advances in neural information processing systems, 32, 2019

    Zhitao Ying, Dylan Bourgeois, Jiaxuan You, Marinka Zitnik, and Jure Leskovec. Gnnexplainer: Generating explanations for graph neural networks.Advances in neural information processing systems, 32, 2019

    2019

  11. [11]

    Parameter- ized explainer for graph neural network.Advances in neural information processing systems, 33:19620–19631, 2020

    Dongsheng Luo, Wei Cheng, Dongkuan Xu, Wenchao Yu, Bo Zong, Haifeng Chen, and Xiang Zhang. Parameter- ized explainer for graph neural network.Advances in neural information processing systems, 33:19620–19631, 2020

    2020

  12. [12]

    On explainability of graph neural networks via subgraph explorations

    Hao Yuan, Haiyang Yu, Jie Wang, Kang Li, and Shuiwang Ji. On explainability of graph neural networks via subgraph explorations. InInternational conference on machine learning, pages 12241–12252. PMLR, 2021

    2021

  13. [13]

    Flowx: Towards explainable graph neural networks via message flows.IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(7):4567–4578, 2023

    Shurui Gui, Hao Yuan, Jie Wang, Qicheng Lao, Kang Li, and Shuiwang Ji. Flowx: Towards explainable graph neural networks via message flows.IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(7):4567–4578, 2023

    2023

  14. [14]

    Explainability methods for graph convolutional neural networks

    Phillip E Pope, Soheil Kolouri, Mohammad Rostami, Charles E Martin, and Heiko Hoffmann. Explainability methods for graph convolutional neural networks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10772–10781, 2019

    2019

  15. [15]

    A value for n-person games

    Lloyd S Shapley et al. A value for n-person games. 1953

    1953

  16. [16]

    Fast axiomatic attribution for neural networks.Advances in Neural Information Processing Systems, 34:19513–19524, 2021

    Robin Hesse, Simone Schaub-Meyer, and Stefan Roth. Fast axiomatic attribution for neural networks.Advances in Neural Information Processing Systems, 34:19513–19524, 2021

    2021

  17. [17]

    On the expressive power of deep polynomial neural networks

    Joe Kileel, Matthew Trager, and Joan Bruna. On the expressive power of deep polynomial neural networks. Advances in neural information processing systems, 32, 2019

    2019

  18. [18]

    P-nets: Deep polynomial neural networks

    Grigorios G Chrysos, Stylianos Moschoglou, Giorgos Bouritsas, Yannis Panagakis, Jiankang Deng, and Stefanos Zafeiriou. P-nets: Deep polynomial neural networks. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 7325–7335, 2020

    2020

  19. [19]

    Convolutional neural networks on graphs with fast localized spectral filtering.Advances in neural information processing systems, 29, 2016

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

    2016

  20. [20]

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

    Keyulu Xu, Weihua Hu, Jure Leskovec, and Stefanie Jegelka. How powerful are graph neural networks? In International Conference on Learning Representations, 2019

    2019

  21. [21]

    Calculation of gauss quadrature rules.Mathematics of computation, 23(106):221–230, 1969

    Gene H Golub and John H Welsch. Calculation of gauss quadrature rules.Mathematics of computation, 23(106):221–230, 1969

    1969

  22. [22]

    Robust design with arbitrary distributions using gauss-type quadrature formula.Structural and Multidisciplinary Optimization, 39(3):227–243, 2009

    Sang Hoon Lee, Wei Chen, and Byung Man Kwak. Robust design with arbitrary distributions using gauss-type quadrature formula.Structural and Multidisciplinary Optimization, 39(3):227–243, 2009

    2009

  23. [23]

    Graphframex: Towards systematic evaluation of explainability methods for graph neural networks

    Kenza Amara, Rex Ying, Zitao Zhang, Zhihao Han, Yinan Shan, Ulrik Brandes, Sebastian Schemm, and Ce Zhang. Graphframex: Towards systematic evaluation of explainability methods for graph neural networks. arXiv preprint arXiv:2206.09677, 2022

    Pith/arXiv arXiv 2022

  24. [24]

    Towards robust fidelity for evaluating explainability of graph neural networks

    Xu Zheng, Farhad Shirani, Tianchun Wang, Wei Cheng, Zhuomin Chen, Haifeng Chen, Hua Wei, and Dongsheng Luo. Towards robust fidelity for evaluating explainability of graph neural networks. InInternational Conference on Learning Representations, volume 2024, pages 12250–12275, 2024

    2024

  25. [25]

    Cambridge university press, 2003

    Endre Süli and David F Mayers.An introduction to numerical analysis. Cambridge university press, 2003

    2003

  26. [26]

    Moleculenet: a benchmark for molecular machine learning.Chemical science, 9(2):513–530, 2018

    Zhenqin Wu, Bharath Ramsundar, Evan N Feinberg, Joseph Gomes, Caleb Geniesse, Aneesh S Pappu, Karl Leswing, and Vijay Pande. Moleculenet: a benchmark for molecular machine learning.Chemical science, 9(2):513–530, 2018

    2018

  27. [27]

    Derivation and validation of toxicophores for mutagenicity prediction.Journal of medicinal chemistry, 48(1):312–320, 2005

    Jeroen Kazius, Ross McGuire, and Roberta Bursi. Derivation and validation of toxicophores for mutagenicity prediction.Journal of medicinal chemistry, 48(1):312–320, 2005. 11 APEX

    2005

  28. [28]

    can”, “moving

    Christopher Morris, Martin Ritzert, Matthias Fey, William L Hamilton, Jan Eric Lenssen, Gaurav Rattan, and Martin Grohe. Weisfeiler and leman go neural: Higher-order graph neural networks. InProceedings of the AAAI conference on artificial intelligence, volume 33, pages 4602–4609, 2019. 12 APEX A Dataset Statistics We use BA-Shapes from GNNExplainer [10];...

    2019