Pith. sign in

REVIEW 4 major objections 5 minor 39 references

CoRe-GNN: Multilevel Message passing on Coarsened graphs

T0 review · 4 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read The paper claims that running coarsened long-range and local intra-cluster message passing in parallel at every layer fixes both the accuracy ceiling of graph coarsening and the long-range blindness of Cluster-GCN, with a spectral approxima

desk verdict New two-branch coarsened/intra-cluster architecture with strong results on long-range graphs, but the batched training proxy and an overbroad 'consistently outperforms' claim need attention before publication. read the letter →

arxiv 2608.02128 v1 pith:YDRWBLJW submitted 2026-08-03 cs.LG

Antonin Joly , Nicolas Keriven , Aline Roumy This is my paper
classification cs.LG MSC 68T0768R10
keywords graphneuralnetworkscoarseningCluster-GCNmessagepassinglong-rangedependenciesnodeclassificationbatchingspectralapproximation
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

Training graph neural networks on large graphs is memory-heavy because each layer stores all node representations. The paper shows that two standard fixes—graph coarsening and Cluster-GCN—are complementary corruptions of the same propagation matrix: coarsening replaces it with a low-rank projection that gives spectral guarantees but makes all nodes in a cluster share one representation, while Cluster-GCN keeps only intra-cluster edges, enabling batching but cutting long-range information. CoRe-GNN performs both propagations in parallel each layer, adding a coarsened inter-cluster term to the local intra-cluster term, and proves an approximation guarantee for the combined propagation with the same restricted-spectral-approximation structure as graph coarsening. A cluster-based batching scheme keeps memory independent of total node count, and experiments show it beats both baselines on homophilic, heterophilic, large-scale, and long-range benchmarks. The upshot: if correct, the two scalability tricks are not an either/or—they can be combined without losing either property.

What carries the argument

The carrying mechanism is the matrix identity S ≈ S_intra + Q S_inter^c P, with P a well-partitioned reduction matrix (Moore-Penrose pseudoinverse of lifting Q), Q a lifting matrix with one nonzero per row, and S_inter^c = P S_inter Q the coarsened inter-cluster propagation. It decomposes one message-passing step into a local branch in original node space and a long-range branch in super-node space, coupled each layer. The proof machinery is Loukas's Restricted Spectral Approximation constant, which measures how well smooth signals survive reduce-lift cycles; Theorem 2 applies the same bound to S_inter. The batching extension uses a self-loop in coarsened space for super-nodes not in the cur

What would settle it

Train CoRe-GNN in full-batch and batched mode on a graph that fits in GPU memory (e.g., Cora or PubMed) with identical hyperparameters and measure test accuracy and training loss trajectories; if the batched variant's accuracy is substantially lower or its loss fails to track the full-batch run, the self-loop proxy is not harmless. A more direct check: compute the per-layer difference between the proxy update and the exact batched update for random cluster batches; if the error grows without bound as layers increase, the proxy does not preserve propagation.

Watch

Extended reading notes

Core claim

The central claim is that the GCN propagation matrix S can be replaced by S_intra + Q S_inter^c P without losing the approximation guarantees that make coarsening attractive, while restoring per-node discriminability. Here P is the reduction to super-nodes, Q the lifting back, S_intra the intra-cluster block-diagonal part, and S_inter^c = P S_inter Q the coarsened inter-cluster part. The paper proves (Theorem 2) that for smooth signals the error of this combined propagation is bounded by the restricted-spectral-approximation constant times constants that now depend on S_inter, mirroring Theorem 1 for pure coarsening; empirically the mean propagation error is lower than Coarsen-GNN's across t

Load-bearing premise

The batched training procedure replaces the intra-cluster update for super-nodes outside the current batch with a self-loop in coarsened space; this proxy has no approximation guarantee, and Theorem 2 covers only the full, unbatched propagation, so the scalability claim rests on that proxy not corrupting learned representations.

Editorial extensions

If this is right

  • Setting the inter-cluster weights to zero recovers Cluster-GCN exactly, so CoRe-GNN is a strict generalization; the intra branch adds per-node detail that coarsening alone flattens.
  • Because the theorem bounds propagation error by constants depending on the inter-cluster part, coarsening ratios can be pushed higher without the uniform-representation failure of Coarsen-GNN.
  • The batched memory cost is O(K(n+b)d), independent of the number of nodes N, so graphs that exhaust GPU memory for full GCN can be trained on a single GPU.
  • On long-range road networks (Paris, London), CoRe-GNN and its batched variant obtain the highest accuracies among the compared architectures, indicating the coarsened branch actually conveys long-range information.
  • On heterophilic graphs, CoRe-GNN outperforms Coarsen-GNN, consistent with the claim that intra-cluster propagation restores discriminability lost when neighboring nodes carry different labels.

Reading between the lines

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

  • The self-loop proxy used for non-batch super-nodes at training time is an approximation without a stated guarantee; a natural test is to compare batched against full-batch CoRe-GNN on a mid-size graph where both fit, and if accuracy drops materially the scalability claim depends on an unvalidated proxy.
  • The two-branch structure is formally a template for asynchronous federated learning on graphs: the coarsened graph is a central aggregator and each cluster a client; the self-loop proxy is exactly a 'not reported back' placeholder. The paper gestures at this; making it concrete would require analyzing convergence under delayed updates.
  • Since Theorem 2's constants involve only S_inter, one can imagine designing coarsenings that explicitly minimize the inter-cluster RSA constant rather than the full coarsening, potentially tightening the bound further—the paper leaves this optimization as future work.
  • The architecture is agnostic to the coarsening algorithm and could be combined with learned or overlapping partitions; whether overlapping clusters would improve long-range reach without losing batching is untested.
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 / 5 minor

Summary. The paper proposes CoRe-GNN, a GNN architecture that runs two propagations in parallel at every layer: an inter-cluster propagation on a coarsened graph (capturing long-range structure) and an intra-cluster propagation on the original graph (preserving per-node discriminability). The authors give a unified matrix view of coarsening-based and cluster-based scalable GNNs, introduce a cluster batching scheme for training on graphs with millions of nodes, prove a propagation-approximation guarantee analogous to the one for coarsened GNNs (Theorem 2), and report extensive node-classification experiments on homophilic, heterophilic, large-scale, and long-range benchmarks. The paper is clearly written, the architecture is simple and appealing, and the experimental study is broad, with standard deviations and hyperparameter tables.

Significance. If the empirical claims hold, CoRe-GNN is a useful and conceptually clean combination of two complementary compression strategies, and its batched variant offers a practical memory-accuracy trade-off for large graphs. The unified propagation-matrix perspective is a nice didactic contribution, and the proof of Theorem 2 is a correct, though short, adaptation of the authors' earlier Theorem 1. The paper does not ship code or machine-checked proofs, but the experimental detail is relatively thorough. The main caveats are that the headline 'consistently outperforms' claim is false as stated, and the scalability of the batched variant rests on a self-loop proxy that has no accompanying approximation analysis.

major comments (4)
  1. [Abstract / §1 / §5] The abstract and conclusion claim that CoRe-GNN 'consistently outperforms' Coarsen-GNN and Cluster-GCN. This is contradicted by the paper's own tables: Table 1 shows Coarsen-GNN beats CoRe-GNN on Cora (82.03 vs 81.67) and CiteSeer (73.86 vs 73.62); Table 2 shows Cluster-GCN beats CoRe-GNN and CoRe-GNN(b) on ogbn-arxiv (70.22 vs 70.03 and 69.56); on Amazon-ratings, CoRe-GNN(b) (45.48) is below both Coarsen-GNN (46.99) and Cluster-GCN (47.02). The body text at the homophilic paragraph acknowledges that Coarsen-GNN is best on Cora/CiteSeer, so the abstract and conclusion should be softened to 'competitive or best on most benchmarks' or supplemented with statistical significance tests.
  2. [§3.1 / Algorithm 1 and Algorithm 4] The batched training forward pass replaces the intra-cluster update for non-batch super-nodes by the self-loop proxy H_c^{(l)}[\bar B] <- σ(H_c^{(l-1)}[\bar B] θ^{(l)}) + H_c^{(l)}[\bar B]. This proxy is the only mechanism that makes memory O(K(n+b)d) and independent of N, but no theorem or rigorous analysis bounds its error. Theorem 2 and Algorithm 3 apply only to the full, unbatched propagation. The proxy's error can accumulate across layers and enter the next layer's inter-cluster term through H_c, so the million-node results are evidence for the batched heuristic, not for the exact CoRe-GNN architecture. I ask the authors to either prove an error bound for the proxy (e.g., under Lipschitz activations, in terms of the coarsening RSA constant and intra-cluster spectral properties) or add a systematic ablation on graphs where full-batch training fits, measuring both accuracy gap and rep
  3. [§3.3 / Appendix A.1] Theorem 2 requires that Π and S_inter are both ker(L)-preserving and that S_inter is R-preserving. Unlike S, S_inter = S - S_intra is not a normalized adjacency-like matrix, and it is not clear when these assumptions hold; the paper refers to [16] for a discussion of the analogous assumptions for S, but does not discuss S_inter. If the assumptions are not satisfied, the guarantee may be vacuous. The authors should either give sufficient conditions under which S_inter inherits these properties from S, or provide empirical validation of the assumptions on the datasets used in the paper.
  4. [§4 / Tables 1–2] The large-scale comparisons are made only against the two compressed-graph baselines and full GCN. Since the paper's central claim is about scalability and long-range accuracy, the absence of standard scalable baselines such as GraphSAGE, GraphSAINT, or other long-range-oriented GNNs makes the relative significance hard to judge. At minimum, the authors should state whether these comparisons were omitted deliberately, and ideally include a few published reference numbers on the large-scale datasets where full GCN is OOM.
minor comments (5)
  1. [Abstract] Typo: 'thesamedecomposition' should be 'the same decomposition'.
  2. [Tables 1 and 2] The label 'CoRE-GNN' is internally inconsistent with the body's 'CoRe-GNN'; unify capitalization.
  3. [Figure 3 caption] The captions 'CΠ terms' and 'CS terms' are unclear; the first row appears to show the two bound constants separately, so reword to something like 'C_S and C_Π terms (left two panels)'.
  4. [Appendix C.1, Table 3] The CoRe-GNN time complexity expression O(K((E_c+E_intra)d + (n+N)d^2 + 2Nd)) has the 2Nd term inside the outer K factor; this is linear in N but the d is not squared. If the 2Nd comes from two projection operations, the expression is acceptable but should be clarified to avoid confusion with the d^2 terms.
  5. [§3.3] The sentence 'if we consider Frobenius norms instead we would easily have ||S_inter||_Fro ≤ ||S||_Fro, hinting that the constants are indeed improved' is not a proof for the L-operator norms used in the theorems. The authors should mark this clearly as an informal motivation, not a claim.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: Theorem 2 is a direct corollary of the authors' prior theorem [16], which is an independent published result; the batched proxy is an unquantified approximation, not a fitted-input prediction.

full rationale

We walked the derivation chain. Theorem 2 (Sec. 3.3, Eq. 9) is proven in Appendix A.1 by the identity ||Sx-(S_intra x+Q S_inter_c P x)||_L = ||S_inter x - Π S_inter Π x||_L and then 'applies Theorem 1 to S_inter in place of S'. This is a legitimate reduction to the authors' earlier theorem [16], quoted as Theorem 1. That theorem is parameter-free, carries explicit assumptions (Π and S ker(L)-preserving, S R-preserving), and does not assume the CoRe-GNN conclusion; per the review rules it counts as independent mathematical support even though it is a self-citation. Section 3.3 explicitly says 'we adapt the results in [16]' and notes the constants change, which is honest. The Coarsen-GNN baseline [16,17] is also the authors' method, but using it as a comparator is not circular. The main gap is the batched training proxy (Algorithm 1, line 6): non-batch super-nodes get a self-loop proxy, and the paper itself says backprop 'is precisely why the self-loop proxy of Algorithm 1 is needed at training time', while Theorem 2 covers only the unbatched Eq. 6/Algorithm 3. This is an unquantified approximation and a scalability-correctness risk, but it is not a circular step: no fitted parameter is renamed a prediction, and no definition covertly contains the target result. We also note the tables contradict 'consistently outperforms' (Coarsen-GNN wins on Cora/CiteSeer; Cluster-GCN wins on ogbn-arxiv), an honesty issue outside circularity. Overall: self-citation at the core of the theory, but independent and non-circular; score 2.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

No new physical or mathematical entities are introduced; the self-loop proxy and the two branches are defined purely from existing matrices P, Q, S. The main free parameters are hyperparameters governing the coarsening, which directly affect the theoretical constants in Theorem 2.

free parameters (2)
  • coarsening ratio r = varies per dataset/model (e.g., 0.3 to 0.999)
    Tunable hyperparameter controlling the trade-off between coarsened graph size and cluster size; chosen per dataset and model, affects both theoretical constants C'_S, C'_Pi and empirical accuracy.
  • preserved space dimension K (Loukas) = 100
    The Loukas coarsening preserves the first K eigenvectors of the Laplacian; K=100 is set by default in Appendix G.
assumptions (4)
  • standard math Theorem 1 of [16] (propagation guarantee for Coarsen-GNN)
    Used as a black box in the proof of Theorem 2 (Appendix A.1); proven in the authors' prior work, so it is a cited theorem rather than a new proof.
  • domain assumption S_inter is R-preserving, and Pi and S_inter are ker(L)-preserving
    Assumptions of Theorem 2 (Section 3.3); they are stated but not verified on real graphs, and the paper defers discussion to [16].
  • domain assumption Q is well-partitioned and L_c = Q^T L Q is a well-defined Laplacian
    Adopted from [17]; required for the coarsening construction (Section 2.1).
  • domain assumption Activation sigma is positively homogeneous (e.g., ReLU)
    Lemma 1's commutativity of Q with sigma requires homogeneity; Remark 3 notes sigmoid/tanh fail.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CoRe-GNN: Multilevel Message passing on Coarsened graphs." pith.science (2026). https://pith.science/paper/YDRWBLJW

@misc{pith2026260802128,
  author       = {Pith},
  title        = {Pith review of: CoRe-GNN: Multilevel Message passing on Coarsened graphs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YDRWBLJW}},
  note         = {Machine review of arXiv:2608.02128}
}
read the original abstract

Training Graph Neural Networks on large graphs is challenged by the memory cost of storing all node representations across layers. We show that several existing scalable approaches can be written as structured modifications of the GNN propagation matrix, providing a unified perspective that exposes their respective limitations. In particular, graph coarsening replaces it by a low-rank approximation that enables spectral guarantees but assigns uniform representations to clustered nodes, while Cluster-GCN restricts the propagation matrix to intra-cluster connections that allow efficient batching but sever long-range information. These are complementary failures of the \emph{same} decomposition of the graph into groups of nodes. To obtain the best of both worlds, we propose \textbf{CoRe-GNN}, which performs both propagations in parallel at each layer: a coarsened inter-cluster term capturing long-range structure, and a local intra-cluster term preserving per-node discriminability. We prove that CoRe-GNN inherits analogous approximation guarantees to those of graph coarsening, and introduce a natural cluster-based \emph{batching scheme} that scales to graphs with millions of nodes. On node classification benchmarks spanning homophilic, heterophilic, large-scale, and long-range graphs, CoRe-GNN outperforms both graph coarsening and Cluster-GCN baselines. Notably, CoRe-GNN reaches competitive accuracy on \emph{long-range} tasks, while remaining memory-efficient through batching.

Figures

Figures reproduced from arXiv: 2608.02128 by the authors.

Figure 1
Figure 1. One CoRe-GNN batched layer. H(l) flows through two parallel branches. Top: intra-cluster message passing, where only the active cluster (red) is processed per step while others are handled in separate batches (in the full-batch setting, all clusters are processed simultaneously). Bottom: inter-cluster propagation on the coarsened graph over all n super-nodes at once, reduce (P), propagate (S inter c ), activate (σ),… view at source ↗
Figure 2
Figure 2. Propagation matrices for N = 20 nodes across 5 clusters. CoRe-GNN (batched): active clusters (red, violet) use full intra-cluster propagation; inactive clusters use a coarsened proxy (small square). The representation of the graph and its coarsened counterpart can be found in [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Comparison of the Coarsen-GNN bound (Theorem 1, [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Accuracy as a function of the coarsening ratio r on Cora, with fixed hyperparameters (K = 2, d = 256, lr=0.01, wd=0.0005, Loukas coarsening algorithm). Influence of coarsening ratio. While in the main tables we select the best coarsening ratio for each model, [PITH_FU…
Figure 5
Figure 5. Figure 5: SBM-1 (pin = 0.08, pout = 0.030) Graph structure 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0.40 0.45 0.50 0.55 0.60 0.65 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0.48 0.50 0.52 0.54 0.56 0.58 0.60 0.62 0.64 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0.7 0.8 0.9 1.0 …
Figure 6
Figure 6. Figure 6: SBM-2 (pin = 0.15, pout = 0.020) Graph structure 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0.400 0.425 0.450 0.475 0.500 0.525 0.550 0.575 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0.51 0.52 0.53 0.54 0.55 0.56 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0.6 0.7 0.8 0…
Figure 7
Figure 7. Figure 7: SBM-3 (pin = 0.70, pout = 0.10) Graph structure 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0.14 0.16 0.18 0.20 0.22 0.24 0.26 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0.240 0.245 0.250 0.255 0.260 0.265 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0.275 0.300 0.325 0.3…
Figure 8
Figure 8. Figure 8: SBM-4 (pin = 0.50, pout = 0.20) B.2 Empirical analysis on real graphs We repeat the same analysis on four real datasets: Cora, Citeseer, Chameleon and Squirrel. 14 [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: Cora Dataset 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 1 2 3 4 5 6 7 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 1 2 3 4 5 6 7 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0 10 20 30 40 50 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0.0 0.2 0.4 0.6 0.8 1.0 Coarsen-GNN CoR…
Figure 10
Figure 10. Figure 10: CiteSeer Dataset 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0 2 4 6 8 10 12 0.0 0.2 0.4 0.6 …
Figure 11
Figure 11. Figure 11: Chameleon Dataset 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0.95 1.00 1.05 1.10 1.15 1.20 1.25 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0 1 2 3 4 0.0 0.2 0.4 0.6 0.8 1.0 Coarsening ratio r 0…
Figure 12
Figure 12. Figure 12: Squirrel Dataset B.3 Discussion Across all graphs and coarsening ratios, the mean propagation error averaged over smooth signals is consistently lower for CoRe-GNN than for Coarsen-GNN, even when the theoretical bound constants are not strictly smaller. This gap betwe…
Figure 13
Figure 13. Figure 13: Efficiency condition Ec + Eintra + 2N + nd ≤ E (normalised by E) on ogbn-arxiv with METIS coarsening, for d ∈ {64, 128, 256}. Shaded area: regime where CoRe-GNN is computationally cheaper than classical GCN [PITH_FULL_IMAGE:figures/full_fig_p018_13.png]
Figure 14
Figure 14. Figure 14: Illustration of the coarsening on the same 5-cluster partition as Figure 2. Left: original graph [PITH_FULL_IMAGE:figures/full_fig_p025_14.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

39 extracted references · 6 linked inside Pith

  1. [16]

    Graph Coarsening with Message-Passing Guarantees.Advances in Neural Information Processing Systems, 37:114902–114927, 2024

    Antonin Joly and Nicolas Keriven. Graph Coarsening with Message-Passing Guarantees.Advances in Neural Information Processing Systems, 37:114902–114927, 2024

    2024

  2. [1]

    Spectral clustering with graph neural networks for graph pooling

    Filippo Maria Bianchi, Daniele Grattarola, and Cesare Alippi. Spectral clustering with graph neural networks for graph pooling. InInternational conference on machine learning, pages 874–883. PMLR, 2020

    2020

  3. [2]

    Bronstein, Joan Bruna, Taco Cohen, and Petar Veličković

    Michael M. Bronstein, Joan Bruna, Taco Cohen, and Petar Veličković. Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges.arXiv:2104.13478, 2021

    arXiv 2021

  4. [3]

    Fastgcn: fast learning with graph convolutional networks via importance sampling

    Jie Chen, Tengfei Ma, and Cao Xiao. Fastgcn: fast learning with graph convolutional networks via importance sampling. arXiv preprint arXiv:1801.10247, 2018

    arXiv 2018

  5. [4]

    Cluster-gcn: An efficient algorithm for training deep and large graph convolutional networks

    Wei-Lin Chiang, Xuanqing Liu, Si Si, Yang Li, Samy Bengio, and Cho-Jui Hsieh. Cluster-gcn: An efficient algorithm for training deep and large graph convolutional networks. InProceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining, pages 257–266, 2019

    2019

  6. [5]

    Spectral graph coarsening using inner product preservation and the grassmann manifold

    Ido Cohen and Ronen Talmon. Spectral graph coarsening using inner product preservation and the grassmann manifold. In The Thirty-ninth Annual Conference on Neural Information Processing Systems, 2025

    2025

  7. [6]

    Weighted graph cuts without eigenvectors a multilevel approach.IEEE transactions on pattern analysis and machine intelligence, 29(11):1944–1957, 2007

    Inderjit S Dhillon, Yuqiang Guan, and Brian Kulis. Weighted graph cuts without eigenvectors a multilevel approach.IEEE transactions on pattern analysis and machine intelligence, 29(11):1944–1957, 2007

    1944

  8. [7]

    Graph coarsening via convolution matching for scalable graph neural network training

    Charles Dickens, Edward Huang, Aishwarya Reganti, Jiong Zhu, Karthik Subbian, and Danai Koutra. Graph coarsening via convolution matching for scalable graph neural network training. InCompanion Proceedings of the ACM on Web Conference, pages 1502–1510, 2024

    2024

Show all 39 references
  1. [8]

    Vq-gnn: A universal framework to scale up graph neural networks using vector quantization.Advances in Neural Information Processing Systems, 34:6733–6746, 2021

    Mucong Ding, Kezhi Kong, Jingling Li, Chen Zhu, John Dickerson, Furong Huang, and Tom Goldstein. Vq-gnn: A universal framework to scale up graph neural networks using vector quantization.Advances in Neural Information Processing Systems, 34:6733–6746, 2021

    2021

  2. [9]

    Learning on large graphs using intersecting communities.Advances in Neural Information Processing Systems, 37:57349–57388, 2024

    Ben Finkelshtein, İsmail İ Ceylan, Michael Bronstein, and Ron Levie. Learning on large graphs using intersecting communities.Advances in Neural Information Processing Systems, 37:57349–57388, 2024. 10

    2024

  3. [10]

    Graph u-nets

    Hongyang Gao and Shuiwang Ji. Graph u-nets. Ininternational conference on machine learning, pages 2083–2092. PMLR, 2019

  4. [11]

    Influence-based mini-batching for graph neural networks

    Johannes Gasteiger, Chendi Qian, and Stephan Günnemann. Influence-based mini-batching for graph neural networks. In Learning on Graphs Conference, pages 9–1. PMLR, 2022

    2022

  5. [12]

    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 (ICML), pages 1–14, 2017. ISBN 978-1-4577-0079-8

    2017

  6. [13]

    Inductive representation learning on large graphs.Advances in neural information processing systems, 30, 2017

    Will Hamilton, Zhitao Ying, and Jure Leskovec. Inductive representation learning on large graphs.Advances in neural information processing systems, 30, 2017

    2017

  7. [14]

    Open graph benchmark: Datasets for machine learning on graphs.Advances in neural information processing systems, 33: 22118–22133, 2020

    Weihua Hu, Matthias Fey, Marinka Zitnik, Yuxiao Dong, Hongyu Ren, Bowen Liu, Michele Catasta, and Jure Leskovec. Open graph benchmark: Datasets for machine learning on graphs.Advances in neural information processing systems, 33: 22118–22133, 2020

    2020

  8. [15]

    Condensing graphs via one-step gradient matching

    Wei Jin, Xianfeng Tang, Haoming Jiang, Zheng Li, Danqing Zhang, Jiliang Tang, and Bing Yin. Condensing graphs via one-step gradient matching. InProceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 720–730, 2022

    2022

  9. [17]

    Taxonomy of reduction matrices for graph coarsening.Advances in Neural Information Processing Systems, 38:96058–96083, 2026

    Antonin Joly, Nicolas Keriven, and Aline Roumy. Taxonomy of reduction matrices for graph coarsening.Advances in Neural Information Processing Systems, 38:96058–96083, 2026

    2026

  10. [18]

    Metis: A software package for partitioning unstructured graphs, partitioning meshes, and computing fill-reducing orderings of sparse matrices

    George Karypis and Vipin Kumar. Metis: A software package for partitioning unstructured graphs, partitioning meshes, and computing fill-reducing orderings of sparse matrices. 1997

    1997

  11. [19]

    Ugc: Universal graph coarsening.Advances in Neural Information Processing Systems, 37:63057–63081, 2024

    Mohit Kataria, Sandeep Kumar, et al. Ugc: Universal graph coarsening.Advances in Neural Information Processing Systems, 37:63057–63081, 2024

    2024

  12. [20]

    Semi-Supervised Classification with Graph Convolutional Networks

    Thomas N Kipf and Max Welling. Semi-Supervised Classification with Graph Convolutional Networks. InInternational Conference on Learning Representations, 2017

    2017

  13. [21]

    Efficient learning on large graphs using a densifying regularity lemma

    Jonathan Kouchly, Ben Finkelshtein, Michael Bronstein, and Ron Levie. Efficient learning on large graphs using a densifying regularity lemma. InInternational Conference on Learning Representations, volume 2026, pages 5772–5826, 2026

    2026

  14. [22]

    Featured graph coarsening with similarity guarantees

    Manoj Kumar, Anurag Sharma, Shashwat Saxena, and Sandeep Kumar. Featured graph coarsening with similarity guarantees. InInternational Conference on Machine Learning, pages 17953–17975. PMLR, 2023

    2023

  15. [23]

    Partition-wise graph filtering: A unified perspective through the lens of graph coarsening

    Guoming Li, Jian Yang, and Yifan Chen. Partition-wise graph filtering: A unified perspective through the lens of graph coarsening. InProceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2, pages 1353–1364, 2025

    2025

  16. [24]

    Towards quantifying long-range interactions in graph machine learning: a large graph dataset and a measurement

    Huidong Liang, Haitz Sáez de Ocáriz Borde, Baskaran Sripathmanathan, Michael Bronstein, and Xiaowen Dong. Towards quantifying long-range interactions in graph machine learning: a large graph dataset and a measurement. InInternational Conference on Learning Representations, vol...

    2026

  17. [25]

    Graph reduction with spectral and cut guarantees.Journal of Machine Learning Research, 20(116):1–42, 2019

    Andreas Loukas. Graph reduction with spectral and cut guarantees.Journal of Machine Learning Research, 20(116):1–42, 2019

    2019

  18. [26]

    Classic gnns are strong baselines: Reassessing gnns for node classification

    Yuankai Luo, Lei Shi, and Xiao-Ming Wu. Classic gnns are strong baselines: Reassessing gnns for node classification. Advances in Neural Information Processing Systems, 37:97650–97669, 2024

    2024

  19. [27]

    A critical look at the evaluation of gnns under heterophily: Are we really making progress? InThe Eleventh International Conference on Learning Representations, 2023

    Oleg Platonov, Denis Kuznedelev, Michael Diskin, Artem Babenko, and Liudmila Prokhorenkova. A critical look at the evaluation of gnns under heterophily: Are we really making progress? InThe Eleventh International Conference on Learning Representations, 2023

    2023

  20. [28]

    Fit-gnn: Faster inference time for gnns that’fit’in memory using coarsening.arXiv preprint arXiv:2410.15001, 2024

    Shubhajit Roy, Hrriday Ruparel, Kishan Ved, and Anirban Dasgupta. Fit-gnn: Faster inference time for gnns that’fit’in memory using coarsening.arXiv preprint arXiv:2410.15001, 2024

    2024 arXiv

  21. [29]

    Multi-scale attributed node embedding.Journal of Complex Networks, 9(2), 2021

    Benedek Rozemberczki, Carl Allen, and Rik Sarkar. Multi-scale attributed node embedding.Journal of Complex Networks, 9(2), 2021

    2021

  22. [30]

    The graph neural network model.IEEE transactions on neural networks, 20(1):61–80, 2008

    Franco Scarselli, Marco Gori, Ah Chung Tsoi, Markus Hagenbuchner, and Gabriele Monfardini. The graph neural network model.IEEE transactions on neural networks, 20(1):61–80, 2008

    2008

  23. [31]

    Graph attention networks.arXiv preprint arXiv:1710.10903, 2017

    Petar Veličković, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, and Yoshua Bengio. Graph attention networks.arXiv preprint arXiv:1710.10903, 2017

    2017 arXiv

  24. [32]

    Next level message-passing with hierarchical support graphs

    Carlos Vonessen, Florian Grötschla, and Roger Wattenhofer. Next level message-passing with hierarchical support graphs. arXiv preprint arXiv:2406.15852, 2024. 11

    2024 arXiv

  25. [33]

    Simplifying graph convolutional networks

    Felix Wu, Amauri Souza, Tianyi Zhang, Christopher Fifty, Tao Yu, and Kilian Weinberger. Simplifying graph convolutional networks. InInternational conference on machine learning, pages 6861–6871. PMLR, 2019

    2019

  26. [34]

    Revisiting semi-supervised learning with graph embeddings

    Zhilin Yang, William Cohen, and Ruslan Salakhudinov. Revisiting semi-supervised learning with graph embeddings. In International conference on machine learning, pages 40–48. PMLR, 2016

    2016

  27. [35]

    Hierarchical graph representation learning with differentiable pooling.Advances in neural information processing systems, 31, 2018

    Zhitao Ying, Jiaxuan You, Christopher Morris, Xiang Ren, Will Hamilton, and Jure Leskovec. Hierarchical graph representation learning with differentiable pooling.Advances in neural information processing systems, 31, 2018

    2018

  28. [36]

    Graphsaint: Graph sampling based inductive learning method.arXiv preprint arXiv:1907.04931, 2019

    Hanqing Zeng, Hongkuan Zhou, Ajitesh Srivastava, Rajgopal Kannan, and Viktor Prasanna. Graphsaint: Graph sampling based inductive learning method.arXiv preprint arXiv:1907.04931, 2019

    1907 arXiv

  29. [37]

    Decoupling the depth and scope of graph neural networks.Advances in neural information processing systems, 34:19665–19679, 2021

    Hanqing Zeng, Muhan Zhang, Yinglong Xia, Ajitesh Srivastava, Andrey Malevich, Rajgopal Kannan, Viktor Prasanna, Long Jin, and Ren Chen. Decoupling the depth and scope of graph neural networks.Advances in neural information processing systems, 34:19665–19679, 2021

    2021

  30. [38]

    Hierarchical message-passing graph neural networks: Z

    Zhiqiang Zhong, Cheng-Te Li, and Jun Pang. Hierarchical message-passing graph neural networks: Z. zhong et al.Data Mining and Knowledge Discovery, 37(1):381–408, 2023

    2023

  31. [39]

    Layer-dependent importance sampling for training deep and large graph convolutional networks.Advances in neural information processing systems, 32, 2019

    Difan Zou, Ziniu Hu, Yewen Wang, Song Jiang, Yizhou Sun, and Quanquan Gu. Layer-dependent importance sampling for training deep and large graph convolutional networks.Advances in neural information processing systems, 32, 2019. A Proofs A.1 Proof of Theorem 2 ∥Sx−(S intrax+QS ...

    2019

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.