Pith. sign in

REVIEW 4 major objections 6 minor 1 cited by

Multigraph Message Passing with Bi-Directional Multi-Edge Aggregations

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read By aggregating parallel edges before combining neighbors, MEGA-GNN keeps permutation equivariance, gains universality under a strict total edge order, and improves minority-class F1 by up to 13.31 percentage points on financial…

desk verdict Genuinely new two-stage multigraph architecture with strong AML results, but the universality theorem rests on a proof sketch that omits the active/finished set mechanism. read the letter →

arxiv 2412.00241 v2 pith:5C6Z2OLI submitted 2024-11-29 cs.LG

H. Çağr{i} Bilgi , Lydia Y. Chen , Kubilay Atasu This is my paper · ORCID
classification cs.LG
keywords multigraphneuralnetworksmulti-edgeaggregationmessagepassingpermutationequivarianceuniversalityfinancialtransactiongraphsanti-moneylaunderingedgeclassification
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

Most graph neural networks assume at most one edge between two nodes, and the two multigraph-specific models proposed so far each sacrifice something: one depends on an arbitrary edge-port numbering that breaks permutation equivariance, while the other flattens parallel edges into a single edge and loses per-edge predictions. MEGA-GNN instead performs a two-stage aggregation in every message-passing layer: parallel edges between the same pair of nodes are aggregated first, and messages from distinct neighbors are combined second, with a symmetric pass over reversed edges for directed graphs. The paper proves that this framework is permutation equivariant whenever both aggregators are permutation invariant, and universal when the edges admit a strict total ordering such as timestamps. If the proofs hold, this closes the gap between expressivity and symmetry for multigraph networks, and the reported experiments show up to 13.31 percentage-point improvements in minority-class F1 over prior state-of-the-art on anti-money-laundering transaction classification while matching it on phishing-account detection.

What carries the argument

The load-bearing object is the artificial node placed on every directed pair that supports at least one edge, together with the two-stage layer built around it. Stage one aggregates the multiset of parallel edge features at the artificial node; stage two aggregates the resulting per-neighbor embeddings at the destination node; reverse edges get their own artificial nodes for bi-directional message passing. This construction does two jobs at once: because both aggregation functions are permutation invariant, the layer is equivariant by composition, and because the first stage can compute a minimum over edge labels, a strict total ordering on edges can drive a BFS-style unique node ID assignment, which is the step that upgrades the framework from equivariant to universal.

What would settle it

Run the proof's node-identification algorithm on a connected multigraph in which two parallel edges have identical feature vectors; with no strict total order the 'minimum edge label' step cannot separate proposals from the same neighbor, so duplicate node IDs become possible, and a universal approximation claim on that input would fail.

Watch

Extended reading notes

Core claim

The central claim is that placing an artificial node between each ordered pair of nodes that has at least one edge turns a multigraph's parallel edges into a well-behaved intermediate aggregation level. In each layer the model first computes one embedding per such pair by a permutation-invariant aggregation over the parallel-edge features, then aggregates these per-neighbor embeddings at the receiving node, and updates the original edge features individually, so no edge-level information is discarded. The paper states this two-stage scheme is permutation equivariant even without any edge ordering, and that with a strict total ordering of edges it is also universal: the first-stage aggregation can select the minimum edge label, which lets the network mimic a breadth-first search that assigns unique node identities, satisfying the known sufficient conditions for universal approximation of graph functions. Empirically, the framework instantiated with GIN, PNA, or GenAgg aggregators outperforms the prior port-numbered and collapsed-edge methods on minority-class F1 for illicit transaction detection, and matches the best prior method on phishing account detection.

Load-bearing premise

The universality proof assumes that edges can be put in a strict total order by their features, so no two parallel edges tie, and that the earlier search-based node-identification construction carries over unchanged when the old port numbers are replaced by that ordering.

Editorial extensions

If this is right

  • Multigraph edge classification becomes a first-class task: each original edge keeps its own latent features and can be labelled directly, instead of being collapsed before message passing.
  • Permutation equivariance and universality are not mutually exclusive: a strict total edge order such as a timestamp gives universality without breaking equivariance.
  • The same framework covers node, edge, and graph classification because every layer updates both node and edge representations, with a permutation-invariant readout for graph-level outputs.
  • On the four AML datasets the two-stage models raise minority-class F1 by 9.25 percentage points on average on the high-illicit sets and 13.31 on the low-illicit sets relative to the strongest prior method, and on the ETH phishing dataset they match the best prior model while clearly exceeding the collapsed-edge baseline.

Reading between the lines

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

  • A consequence the paper leaves implicit is that rounding or hashing edge features so parallel edges tie should erode the universality guarantee, and the size of that empirical drop would quantify how much of the AML gain comes from the strict-ordering assumption versus the two-stage structure itself.
  • The artificial-node construction is a general hierarchical aggregation pattern, so the same two-stage idea could transfer to hypergraphs or to graphs with grouped edge relationships, though the paper only draws that analogy in passing.
  • Because the asymptotic cost remains O(|E|d), equal to single-stage messaging, the added expressivity is not bought with extra asymptotic compute; this makes the framework a natural drop-in candidate for large transaction graphs.
  • The paper reports that Ego-IDs help edge classification but hurt node classification on the Ethereum data, suggesting that external identifier features and structural aggregation interact differently across tasks and deserve a dedicated study.
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

4 major / 6 minor

Summary. The paper proposes MEGA-GNN, a message-passing framework for attributed directed multigraphs in which each layer first aggregates parallel edges at artificial nodes (Eq. 4), then aggregates messages from distinct neighbors (Eq. 5), with optional bi-directional message passing (Section 3.4) and per-edge updates (Eq. 6). The main theoretical claims are permutation equivariance (Theorem 3.1) and universality under a strict total edge ordering (Theorem 3.2, via a BFS-based node-ID construction in Lemma 3.1). Empirically, the authors report minority-class F1 gains of up to about 13 percentage points over Multi-GNN and GFP baselines on four AML transaction datasets, comparable or slightly better F1 on the ETH phishing dataset, an ablation study, and a throughput comparison.

Significance. If the theoretical claims can be made rigorous, the paper would be a useful contribution to the relatively sparse literature on GNNs for multigraphs: it proposes a simple and natural two-stage aggregation, identifies a genuine limitation of prior work (loss of permutation equivariance in Multi-GNN's port numbering), and demonstrates substantial gains on financial transaction tasks. The authors ship code and use public datasets, and the ablations help isolate the contribution of bi-directional message passing and Ego-IDs. However, the current proofs of both central theorems are sketches, and the empirical significance testing is incomplete, so the contribution is not yet fully substantiated.

major comments (4)
  1. [A.2, Theorem 3.1] The proof of Theorem 3.1 misstates the defining property of a permutation-invariant function: it asserts ψ(ρ·E)=ρ·ψ(E) and similarly for φ, but for invariant functions the right-hand side should be ψ(E), not ρ·ψ(E). The theorem is nevertheless plausibly correct, but the proof needs to be rewritten as an equivariance argument for the two-stage layer, showing that h_{π(i)π(j)}, a_{π(j)}, and the updated edge features transform correctly under node and edge permutations. As written, the proof is not valid.
  2. [A.3, Lemma 3.1 / Theorem 3.2] The proof of Lemma 3.1 is a sketch that defers to Egressy et al. and does not establish that the two-stage layer defined by Eqs. (4)-(6) can simulate Algorithm 1. In particular, Algorithm 1 sends messages only from active nodes and prevents finished nodes from being updated, whereas the MEGA-GNN layer has no such conditional: all nodes send messages in every layer, and a node with initial feature 0 would send a 0-prefixed message that dominates min-aggregation. The proof needs to specify an explicit encoding (e.g., a sentinel for inactive nodes and a masking mechanism for finished nodes) and show that the layer functions f, g, and EdgeAgg can implement it. Without this, Theorem 3.2 is not established.
  3. [Lemma 3.1 and Theorem 3.2] Lemma 3.1 is stated only for connected multigraphs with a chosen root, but Theorem 3.2 claims universality without these qualifications. The proof does not explain how the root r is selected in the GNN or how disconnected graphs are handled. Please either add the connectedness and root assumptions to Theorem 3.2 or extend the proof to disconnected graphs (e.g., per-component BFS with component identifiers).
  4. [Tables 2-3, Section 4.2] The empirical claims of 'significantly outperforms' and 'surpassing' are not supported by statistical tests. On ETH (Table 3), MEGA-PNA (64.84±1.73) is within one standard deviation of Multi-PNA (64.61±1.40), yet Section 4.2 states that MEGA-PNA 'surpasses' Multi-PNA and 'achieves the highest F1 score'. The abstract more cautiously says 'on par' for ETH. Please add significance tests (e.g., paired tests across seeds) and make the claims in Sections 4.2 and 6 consistent with the abstract.
minor comments (6)
  1. [Abstract vs. Section 4.2/Conclusion] The ETH result is described as 'on par' in the abstract, 'surpassing' in Section 4.2, and 'slightly improving' in the conclusion; please use consistent wording.
  2. [Definition 3.1] The sentence defining the strict total order on edges via 'e(0)_ijp < e(0)_{i'j'p'}' is incomplete; it should specify how feature vectors are compared (e.g., lexicographic order) and should state that all edges have distinct feature vectors.
  3. [Section 6] The claim 'first message-passing framework explicitly designed for multigraphs' overlooks Multi-GNN and ADAMM; the intended novelty is the two-stage aggregation within message passing, so please rephrase.
  4. [Appendix A.1, Eq. (14)] The displayed chain 'X(l) = f(G(X,p) ≠ f(G(X,pσ)) = X̂(l)' is malformed; fix the parentheses and the equality chain.
  5. [Appendix B.2, Eq. (21)] GenAgg is applied to a SUM in the displayed formula, which is inconsistent with the text describing GenAgg as a scalar-valued aggregator; clarify whether the intended input is the multiset of edge features.
  6. [Tables 6-7] Tables 6 and 7 use commas as decimal separators (e.g., '43,66±0.54') while the main tables use periods; standardize the decimal separator.

Circularity Check

1 steps flagged · score 4.0 of 10

Universality proof imports its BFS-simulation premise from overlapping-author prior work rather than deriving it from Eqs. (4)-(6).

  1. self citation load bearing [Appendix A.3 (proof of Lemma 3.1), relied on by Theorem 3.2 in Section 3.5]
    "Egressy et al. (2024) showed that a GNN can mimic a Breadth-First Search (BFS) algorithm to compute unique node IDs given pre-computed port numbers for the edges. We follow the same BFS-based approach... We are not going to reiterate the setup of the entire proof and focus on the differences."

    Theorem 3.2's universality rests on Lemma 3.1, whose proof's load-bearing claim is that the MEGA-GNN layer 'mimics Algorithm 1' (BFS node-ID assignment). That claim is not derived from Eqs. (4)-(6); it is inherited from Egressy et al. (2024), which shares author K. Atasu with this paper. The appendix explicitly declines to reproduce the transfer, and Algorithm 1's active/finished sets and 'send only if v in X' rule are not expressible in Eqs. (4)-(6) as written. So the uniqueness/universality conclusion is supported by the self-citation chain rather than by a self-contained derivation, making the cited prior work load-bearing for the paper's central theoretical claim.

full rationale

The paper is not circular in the input-output sense: no fitted parameter is relabeled as a prediction, and the AML/ETH results are external benchmarks against published baselines. Theorem 3.1's permutation-equivariance proof, though compressed and partly resting on a standard composition argument (Bronstein et al.), is a self-contained argument about permutation-invariant aggregations. The one load-bearing circularity is in the universality chain: Lemma 3.1/Theorem 3.2 assert that a MEGA-GNN layer can simulate the BFS algorithm of Egressy et al., and the proof supports this only by citing that overlapping-author paper and saying 'We are not going to reiterate the setup of the entire proof.' Because Algorithm 1 requires conditional active/finished sets that Eqs. (4)-(6) do not obviously implement, the simulation is the very thing to be proven; deferring it to a self-citation makes the universality claim inherited rather than independently established. This raises the score to 4: the architectural contribution and experiments have independent content, but a central theoretical claim reduces to a self-citation chain. The gap could also be framed as a correctness/completeness issue rather than definitional circularity; it is not a fitted-input 'prediction'.

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

The central theoretical claims rely on standard approximation theorems, the Loukas universality criteria, and a BFS simulation lifted from Egressy et al. (2024). The central empirical claims rely on per-dataset hyperparameter choices and reported benchmark splits. The only invented entities are the artificial aggregation nodes.

free parameters (1)
  • Per-dataset hyperparameter configuration = lr 0.0008 to 0.006, hidden 20 to 64, batch 4096 to 8192, dropout 0.1 to 0.28, class weight ratios 1:3 to 1:7
    Table 5 gives separate settings for each dataset and base model; the reported gains depend on these choices and no fixed validation/search protocol is described.
assumptions (5)
  • standard math Universal Approximation Theorem for multilayer perceptrons
    Used in Appendix A.3 to assert MEGA-GNN layers can mimic the BFS ID-assignment algorithm.
  • domain assumption Loukas (2020) conditions are sufficient for universality of message passing GNNs
    Theorem 3.2 cites these conditions (enough layers, width, expressiveness, unique node IDs) as sufficient without proving them for two-stage aggregation.
  • domain assumption Egressy et al. (2024) BFS-to-GNN simulation transfers to MEGA-GNN
    Appendix A.3 says the proof follows Egressy et al. and highlights only differences; the construction is not re-derived.
  • domain assumption A strict total ordering of edges exists via initial edge features
    Lemma 3.1 and Theorem 3.2 require an edge labeling L:E to [1,m]; real datasets are not shown to satisfy this, for example if timestamps contain ties.
  • domain assumption Minimum aggregation over edge labels is expressible by the chosen EdgeAgg functions
    The unique-ID proof uses min over parallel edges and elementwise MLP followed by min at node level; this assumes the model instances used in experiments include these operations.
invented entities (1)
  • Artificial nodes V_art and reverse artificial nodes \hat{V}_art
    purpose: Intermediate aggregation points between unique node pairs that summarize parallel edges before node-level aggregation.
    These are computational constructs introduced by the architecture. They do not correspond to observed entities and make no falsifiable prediction outside the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multigraph Message Passing with Bi-Directional Multi-Edge Aggregations." pith.science (2026). https://pith.science/paper/5C6Z2OLI

@misc{pith2026241200241,
  author       = {Pith},
  title        = {Pith review of: Multigraph Message Passing with Bi-Directional Multi-Edge Aggregations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5C6Z2OLI}},
  note         = {Machine review of arXiv:2412.00241}
}
read the original abstract

Graph Neural Networks (GNNs) have seen significant advances in recent years, yet their application to multigraphs, where parallel edges exist between the same pair of nodes, remains under-explored. Standard GNNs, designed for simple graphs, compute node representations by combining all connected edges at once, without distinguishing between edges from different neighbors. There are some GNN architectures proposed specifically for multigraphs, yet these architectures perform only node-level aggregation in their message passing layers, which limits their expressive power. Furthermore, these approaches either lack permutation equivariance when a strict total edge ordering is absent, or fail to preserve the topological structure of the multigraph. To address all these shortcomings, we propose MEGA-GNN, a unified framework for message passing on multigraphs that can effectively perform diverse graph learning tasks. Our approach introduces a two-stage aggregation process in the message passing layers: first, parallel edges are aggregated, followed by a node-level aggregation of messages from distinct neighbors. We show that MEGA-GNN is not only permutation equivariant but also universal given a strict total ordering on the edges. Experiments show that MEGA-GNN significantly outperforms state-of-the-art solutions by up to 13\% on Anti-Money Laundering datasets and is on par with their accuracy on real-world phishing classification datasets in terms of minority class F1 score.

Figures

Figures reproduced from arXiv: 2412.00241 by the authors.

Figure 1
Figure 1. Directed multi￾graph port numbering of Multi-GNN (Egressy et al. (2024)). In the literature, two key works specifically address multigraphs: Multi￾GNN Egressy et al. (2024) and ADAMM Sotiropoulos et al. (2023). Multi-GNN introduced a provably powerful GNN architecture with sim￾ple adaptations for directed multigraphs. A notable contribution of Multi-GNN is the multigraph port numbering on edges, which allows the mod… view at source ↗
Figure 2
Figure 2. Illustration of ADAMM Sotiropou￾los et al. (2023) ADAMM proposed aggregating all parallel edges between two nodes to a single undirected super-edge, thereby transforming the multigraph into a simple graph before message passing. An initial features for this super-edge are computed using DeepSet (Zaheer et al. (2017)) and the subsequent message passing layers operate on these features. However, this approach results … view at source ↗
Figure 3
Figure 3. Comparison of two-stage and single-stage aggregation strategies for multigraphs. The left panel illustrates a sample multigraph. The center panel depicts the two-stage aggregation process, wherein the first stage aggregates parallel edges, followed by a second stage node-level aggregation. The right panel shows the single-stage node-level aggregation scheme commonly em￾ployed by standard GNNs, which aggregates all e… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Illustration of Multi-Edge Aggregation using artificial nodes in a multigraph. (a) A multigraph with parallel edges eij1, eij2, eij3 between the nodes i and j. (b) Artificial nodes positioned between adjacent pairs of nodes to handle the aggregation of parallel edges. …
Figure 5
Figure 5. Figure 5: Throughput (transactions/sec) and minority-class F1 score (right) comparisons between [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Counterexample for the permutation equivariance of Multi-GNN Egressy et al. (2024). [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]

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. SALT-GNN: Handling Dense Neighborhoods in Anti-Money Laundering Graphs via Statistics-Aware Attention

    cs.LG 2026-07 conditional novelty 6.0 of 10

    Layer-wise fusion of degree-aware statistical aggregation and attention (SALT-GNN) fixes dense-recipient AML degradation that aggregate F1 scores hide.

Reference graph

Works this paper leans on

39 extracted references · 18 canonical work pages · cited by 1 Pith paper

  1. [1]

    The surprising power of graph neural networks with random node initialization, 2021

    Ralph Abboud, İsmail İlkan Ceylan, Martin Grohe, and Thomas Lukasiewicz. The surprising power of graph neural networks with random node initialization, 2021. URL https://arxiv.org/abs/2010.01179

    arXiv 2021

  2. [2]

    Realistic synthetic financial transactions for anti-money laundering models

    Erik Altman, Jovan Blanu s a, Luc Von Niederh \"a usern, Beni Egressy, Andreea Anghel, and Kubilay Atasu. Realistic synthetic financial transactions for anti-money laundering models. In Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track, 2023

    work page 2023

  3. [3]

    Graph neural networks with local graph parameters

    Pablo Barcel\' o , Floris Geerts, Juan Reutter, and Maksimilian Ryschkov. Graph neural networks with local graph parameters. In M. Ranzato, A. Beygelzimer, Y. Dauphin, P.S. Liang, and J. Wortman Vaughan (eds.), Advances in Neural Information Processing Systems, volume 34, pp.\ 25280--25293. Curran Associates, Inc., 2021. URL https://proceedings.neurips.cc...

    work page 2021

  4. [4]

    Bronstein, and Haggai Maron

    Beatrice Bevilacqua, Fabrizio Frasca, Derek Lim, Balasubramaniam Srinivasan, Chen Cai, Gopinath Balamurugan, Michael M. Bronstein, and Haggai Maron. Equivariant subgraph aggregation networks. In International Conference on Learning Representations, 2022. URL https://openreview.net/forum?id=dFbKQaRk15w

    2022

  5. [5]

    Efficient subgraph GNN s by learning effective selection policies

    Beatrice Bevilacqua, Moshe Eliasof, Eli Meirom, Bruno Ribeiro, and Haggai Maron. Efficient subgraph GNN s by learning effective selection policies. In The Twelfth International Conference on Learning Representations, 2024. URL https://openreview.net/forum?id=gppLqZLQeY

    work page 2024

  6. [6]

    Graph feature preprocessor: Real-time extraction of subgraph-based features from transaction graphs, 2024

    Jovan Blanuša, Maximo Cravero Baraja, Andreea Anghel, Luc von Niederhäusern, Erik Altman, Haris Pozidis, and Kubilay Atasu. Graph feature preprocessor: Real-time extraction of subgraph-based features from transaction graphs, 2024. URL https://arxiv.org/abs/2402.08593

    arXiv 2024

  7. [7]

    Bronstein

    Giorgos Bouritsas, Fabrizio Frasca, Stefanos Zafeiriou, and Michael M. Bronstein. Improving graph neural network expressivity via subgraph isomorphism counting. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45 0 (1): 0 657--668, 2023. doi:10.1109/TPAMI.2022.3154319

    arXiv 2023

  8. [8]

    Geometric deep learning: Grids, groups, graphs, geodesics, and gauges

    Michael M Bronstein, Joan Bruna, Taco Cohen, and Petar Veličković. Geometric deep learning: Grids, groups, graphs, geodesics, and gauges. arXiv preprint arXiv:2104.13478, 2021

    arXiv 2021

Show all 39 references
  1. [9]

    Phishing scams detection in Ethereum transaction network

    Liang Chen, Jiaying Peng, Yang Liu, Jintang Li, Fenfang Xie, and Zibin Zheng. Phishing scams detection in Ethereum transaction network . ACM Trans. Internet Technol., 21 0 (1): 0 1--16, Feb. 2021. doi:10.1145/3398071

    2021 doi

  2. [10]

    Principal neighbourhood aggregation for graph nets

    Gabriele Corso, Luca Cavalleri, Dominique Beaini, Pietro Li\` o , and Petar Veli c kovi\' c . Principal neighbourhood aggregation for graph nets. In H. Larochelle, M. Ranzato, R. Hadsell, M.F. Balcan, and H. Lin (eds.), Advances in Neural Information Processing Systems, volume...

    2020

  3. [11]

    Provably powerful graph neural networks for directed multigraphs

    Béni Egressy, Luc von Niederhäusern, Jovan Blanusa, Erik Altman, Roger Wattenhofer, and Kubilay Atasu. Provably powerful graph neural networks for directed multigraphs. In Proceedings of the AAAI Conference on Artificial Intelligence (AAAI). Underline Science Inc., 2024. doi:1...

    2024 doi

  4. [12]

    Hypergraph neural networks

    Yifan Feng, Haoxuan You, Zizhao Zhang, Rongrong Ji, and Yue Gao. Hypergraph neural networks. Proceedings of the AAAI Conference on Artificial Intelligence, 33 0 (01): 0 3558--3565, Jul. 2019. doi:10.1609/aaai.v33i01.33013558. URL https://ojs.aaai.org/index.php/AAAI/article/view/4235

    2019 doi

  5. [13]

    Matthias Fey and Jan E. Lenssen. Fast graph representation learning with PyTorch Geometric . In ICLR Workshop on Representation Learning on Graphs and Manifolds, 2019

    2019

  6. [14]

    Bronstein, and Haggai Maron

    Fabrizio Frasca, Beatrice Bevilacqua, Michael M. Bronstein, and Haggai Maron. Understanding and extending subgraph GNN s by rethinking their symmetries. In Alice H. Oh, Alekh Agarwal, Danielle Belgrave, and Kyunghyun Cho (eds.), Advances in Neural Information Processing System...

    2022

  7. [15]

    Fuchs* and Petar Veličković*

    Fabian B. Fuchs* and Petar Veličković*. Universality of neural networks on sets vs. graphs. In ICLR Blogposts 2023, 2023. URL https://iclr-blogposts.github.io/2023/blog/2023/sets-and-graphs/. https://iclr-blogposts.github.io/2023/blog/2023/sets-and-graphs/

    2023

  8. [16]

    Schoenholz, Patrick F

    Justin Gilmer, Samuel S. Schoenholz, Patrick F. Riley, Oriol Vinyals, and George E. Dahl. Neural message passing for quantum chemistry. In Proceedings of the 34th International Conference on Machine Learning - Volume 70, ICML'17, pp.\ 1263–1272. JMLR.org, 2017

    2017

  9. [17]

    Inductive representation learning on large graphs

    Will Hamilton, Zhitao Ying, and Jure Leskovec. Inductive representation learning on large graphs. In I. Guyon, U. Von Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett (eds.), Advances in Neural Information Processing Systems, volume 30. Curran Associa...

    2017

  10. [18]

    Multilayer feedforward networks are universal approximators

    Kurt Hornik, Maxwell Stinchcombe, and Halbert White. Multilayer feedforward networks are universal approximators. Neural Networks, 2 0 (5): 0 359--366, 1989. ISSN 0893-6080. doi:https://doi.org/10.1016/0893-6080(89)90020-8. URL https://www.sciencedirect.com/science/article/pii...

    1989

  11. [19]

    Unignn: a unified framework for graph and hypergraph neural networks

    Jing Huang and Jie Yang. Unignn: a unified framework for graph and hypergraph neural networks. In Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, IJCAI-21 , pp.\ 2563--2569. International Joint Conferences on Artificial Intelligence Orga...

    2021 doi

  12. [20]

    edgnn: a simple and powerful gnn for directed labeled graphs, 2019

    Guillaume Jaume, An phi Nguyen, María Rodríguez Martínez, Jean-Philippe Thiran, and Maria Gabrani. edgnn: a simple and powerful gnn for directed labeled graphs, 2019. URL https://arxiv.org/abs/1904.08745

    2019 arXiv

  13. [21]

    Universal invariant and equivariant graph neural networks

    Nicolas Keriven and Gabriel Peyr\' e . Universal invariant and equivariant graph neural networks. In H. Wallach, H. Larochelle, A. Beygelzimer, F. d Alch\' e -Buc, E. Fox, and R. Garnett (eds.), Advances in Neural Information Processing Systems, volume 32. Curran Associates, I...

    2019

  14. [22]

    Kipf and Max Welling

    Thomas N. Kipf and Max Welling. Semi-supervised classification with graph convolutional networks. In International Conference on Learning Representations, 2017. URL https://openreview.net/forum?id=SJU4ayYgl

    2017

  15. [23]

    Generalised f-mean aggregation for graph neural networks

    Ryan Kortvelesy, Steven Morad, and Amanda Prorok. Generalised f-mean aggregation for graph neural networks. In Thirty-seventh Conference on Neural Information Processing Systems, 2023. URL https://openreview.net/forum?id=JMrIeKjTAe

    2023

  16. [24]

    What graph neural networks cannot learn: depth vs width

    Andreas Loukas. What graph neural networks cannot learn: depth vs width. In International Conference on Learning Representations, 2020. URL https://openreview.net/forum?id=B1l2bp4YwS

    2020

  17. [25]

    Provably powerful graph networks

    Haggai Maron, Heli Ben-Hamu, Hadar Serviansky, and Yaron Lipman. Provably powerful graph networks. In H. Wallach, H. Larochelle, A. Beygelzimer, F. d Alch\' e -Buc, E. Fox, and R. Garnett (eds.), Advances in Neural Information Processing Systems, volume 32. Curran Associates, ...

    2019

  18. [26]

    Hamilton, Jan Eric Lenssen, Gaurav Rattan, and Martin Grohe

    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. In Proceedings of the Thirty-Third AAAI Conference on Artificial Intelligence and Thirt...

    2019 doi

  19. [27]

    Topology (2nd edn), 2000

    James R Munkres. Topology (2nd edn), 2000

    2000

  20. [28]

    Random features strengthen graph neural networks

    Ryoma Sato, Makoto Yamada, and Hisashi Kashima. Random features strengthen graph neural networks. In Proceedings of the 2021 SIAM International Conference on Data Mining, SDM , 2021

    2021

  21. [29]

    Modeling relational data with graph convolutional networks

    Michael Schlichtkrull, Thomas N Kipf, Peter Bloem, Rianne Van Den Berg, Ivan Titov, and Max Welling. Modeling relational data with graph convolutional networks. In The semantic web: 15th international conference, ESWC 2018, Heraklion, Crete, Greece, June 3--7, 2018, proceeding...

    2018

  22. [30]

    Adamm: Anomaly detection of attributed multi-graphs with metadata: A unified neural network approach

    Konstantinos Sotiropoulos, Lingxiao Zhao, Pierre Jinghong Liang, and Leman Akoglu. Adamm: Anomaly detection of attributed multi-graphs with metadata: A unified neural network approach. In 2023 IEEE International Conference on Big Data (BigData), 2023. doi:10.1109/BigData59044....

    2023

  23. [31]

    Composition-based multi-relational graph convolutional networks

    Shikhar Vashishth, Soumya Sanyal, Vikram Nitin, and Partha Talukdar. Composition-based multi-relational graph convolutional networks. In International Conference on Learning Representations, 2020. URL https://openreview.net/forum?id=BylA_C4tPr

    2020

  24. [32]

    Veličković, G

    P. Veličković, G. Cucurull, A. Casanova, A. Romero, P. Liò, and Y. Bengio. Graph attention networks. In International Conference on Learning Representations, 2018. URL https://openreview.net/forum?id=rJXMpikCZ

    2018

  25. [33]

    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

  26. [34]

    Identity-aware graph neural networks

    Jiaxuan You, Jonathan Gomes-Selman, Rex Ying, and Jure Leskovec. Identity-aware graph neural networks. In AAAI Conference on Artificial Intelligence (AAAI), 2021

    2021

  27. [35]

    Deep sets

    Manzil Zaheer, Satwik Kottur, Siamak Ravanbakhsh, Barnabas Poczos, Russ R Salakhutdinov, and Alexander J Smola. Deep sets. In I. Guyon, U. Von Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett (eds.), Advances in Neural Information Processing Systems, ...

    2017

  28. [36]

    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 gl...

  29. [37]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  30. [38]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  31. [39]

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.