Pith. sign in

REVIEW 3 major objections 5 minor 48 references

FedSA-GCL: A Semi-Asynchronous Federated Graph Learning Framework with Personalized Aggregation and Cluster-Aware Broadcasting

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read FedSA-GCL claims to be the first asynchronous federated graph learning method explicitly targeting non-IID graph data, reporting average gains of 1.9% to 3.0% over ten baselines.

desk verdict The mechanism is genuinely new and the experiments are broad, but the convergence theorem analyzes a different algorithm than the one implemented. read the letter →

arxiv 2507.18219 v3 pith:XV62GLMG submitted 2025-07-24 cs.LG cs.AI

Zhongzheng Yuan , Lianshuai Guo , Xunkai Li , Yinlin Zhu , Wenyu Wang , Meixia Qu This is my paper
classification cs.LGcs.AI
keywords federatedgraphlearningasynchronousnon-IIDgraphssoftlabelfeaturematrixstaleness-awareaggregationclusterbroadcastingnodeclassification
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

Federated graph learning typically assumes synchronous rounds, which is impractical when clients are uneven or drop out. Existing asynchronous federated learning ignores graph topology and label skew, so applying it to graphs causes semantic drift. This paper proposes FedSA-GCL, a semi-asynchronous framework that groups clients by the similarity of their soft-label distributions, weights each client's update by both structural confidence and staleness, and proactively broadcasts fresh models to inactive clients in the same group. The paper argues that this is the first asynchronous FGL method that explicitly targets graph non-IID, and reports average accuracy gains of 1.9% over the best baseline with Louvain splits and 3.0% with Metis splits across eight datasets.

What carries the argument

Two coupled mechanisms carry the argument. The Soft Label Feature Matrix (SFM, Eq. (6)) compresses each client's label distribution into a degree-weighted $|\mathcal{C}| \times |\mathcal{C}|$ matrix of edge-wise soft-label outer products, and clients are clustered by cosine similarity on these matrices. The aggregation weight in Eq. (10) couples a spatial filter, the local smoothness confidence LSC (Eq. (9), an entropy-based measure of neighborhood prediction consistency), with a temporal filter, the staleness penalty $(d_{j,t})^{-\alpha}$. ClusterCast (Eq. (11)) then broadcasts each cluster's aggregated model to inactive members, which cache it and mix it with their local model proportionally to LSC. The convergence theorem is carried by the effective staleness $\tau_{\mathrm{eff}} = \sup_t \sum_{i\in\mathcal{U}_t} q_{i,t} d_{i,t}$, whose boundedness the proof ties to $\alpha\ge 1$.

What would settle it

Measure the empirical squared aggregation bias $\mathbb{E}\|\sum_{i\in\mathcal{U}_t} q_{i,t}\nabla F_i(\omega) - \nabla F_c(\omega)\|^2$ across rounds on a highly skewed Metis split; if it grows with non-IID severity or training time rather than remaining bounded, the constant $\epsilon_w^2$ in Theorem 1 fails to hold and the bound does not cover the evaluated configuration. Track $\sum_{i\in\mathcal{U}_t} q_{i,t} d_{i,t}$ under $\alpha=0.4$ to check whether effective staleness stays bounded as delays grow.

Watch

Extended reading notes

Core claim

FedSA-GCL's central claim is that semi-asynchronous federated graph learning can beat both synchronous and asynchronous baselines on non-IID graph data without sacrificing convergence stability. The method builds a soft label feature matrix (SFM) per client from degree-weighted products of neighboring node soft labels, clusters clients by cosine similarity on those matrices, and then aggregates each client's personalized model from its cluster using weights proportional to local smoothness confidence (LSC), an entropy-based measure of neighborhood prediction consistency, divided by a staleness penalty. The ClusterCast mechanism further sends each aggregated model to non-participating clients in the same semantic cluster, which cache and confidence-mix it before the next local training round. The paper proves an ergodic convergence bound for the cluster objective and supports the claim with experiments on eight datasets against ten baselines, including robustness and efficiency evaluations.

Load-bearing premise

The convergence proof depends on Assumption 4, which states that the aggregation bias from partial participation and non-uniform weights is bounded by a constant independent of training dynamics; the proof's bounded-drift guarantee is shown for $\alpha\ge 1$, whereas the experiments use $\alpha=0.4$.

Editorial extensions

If this is right

  • Semi-asynchronous FGL with cluster-aware aggregation can remain robust to stragglers and intermittent client participation while matching or beating synchronous methods on non-IID graph data.
  • With ClusterCast, inactive clients receive model updates without uploading, so per-round upstream communication drops relative to strictly bidirectional FedAvg-style aggregation.
  • The convergence analysis indicates that forcing aggregation weights to decay with staleness keeps asynchronous drift bounded when the decay exponent is at least one.
  • The method maintains accuracy under 10% label masking and 10% edge dropping on the benchmark splits, improving over the best baseline by 3.7% and 2.8% respectively.
  • The added SFM and LSC computation is linear in the number of edges and classes, keeping the method's complexity below that of the GNN backbone even for dense graphs.

Reading between the lines

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

  • A testable extension is to replace the static similarity threshold $\theta$ with a server-side adaptive threshold that updates from already-uploaded SFMs, avoiding the global synchronization the paper warns against.
  • Because the proof's bounded-drift argument holds for $\alpha\ge 1$ while experiments use $\alpha=0.4$, evaluating the method in the proven regime would show whether accuracy persists there or whether the empirical choice needs a separate justification.
  • The privacy analysis acknowledges class-distribution leakage via SFM; an immediate empirical follow-up would measure how accurately an honest-but-curious server can recover label proportions from SFMs under different local differential privacy noise levels.
  • ClusterCast's broadcast pattern suggests a natural adaptation to dynamic graphs, where periodic SFM recomputation could absorb topological drift while still avoiding full synchronization.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes FedSA-GCL, a semi-asynchronous federated graph learning framework that clusters clients using a Soft Label Feature Matrix (SFM), aggregates per-client models with weights combining Local Smoothness Confidence (LSC) and staleness penalties, and introduces a ClusterCast mechanism that broadcasts personalized models to inactive but semantically similar clients. The authors claim this is the first asynchronous FGL method that explicitly addresses the graph non-IID problem, and they report average accuracy gains of 1.9% under Louvain partitioning and 3.0% under Metis partitioning across eight datasets and ten baselines. The paper also includes a convergence analysis (Theorem 1), ablations, robustness tests, hyperparameter sensitivity studies, and communication-efficiency evaluations.

Significance. If the empirical claims hold, the paper addresses a genuine gap: asynchronous federated graph learning that explicitly handles non-IID graphs. The experimental study is broad in scope, covering eight datasets, ten baselines, two graph partitioners, ablations, robustness perturbations, varying client counts, and communication-overhead analysis, and the authors are transparent about the downstream bandwidth trade-off of ClusterCast. However, the theoretical convergence guarantee as stated does not apply to the algorithm that is implemented and evaluated, and the empirical superiority claim is not supported by significance testing. These issues bear directly on the paper's central claims of a novel, guaranteed-convergent, state-of-the-art asynchronous FGL method.

major comments (3)
  1. [5.2 and Algorithm 1, Eq. (10)-(11), (14)-(16)] The convergence analysis in Section 5.2 defines a single global model \tilde\omega^t updated by the pseudo-gradient step in Eq. (14), and Theorem 1 bounds the gradient norm of that model. Algorithm 1, however, never constructs or updates such a global model: lines 10-18 compute a separate personalized aggregate \tilde\omega_i for each uploaded client i via Eq. (10) and send it to client i and to inactive peers in C_i, while the server knowledge base stores per-client models, SFM, LSC, and timestamps. In addition, Algorithm 2 line 11 applies Eq. (11) to mix the downloaded model with the client's current local model before training, which changes the initialization point of local updates for broadcast recipients. Consequently, the descent inequality (16) and the ergodic bound (15) are derived for an algorithm different from the proposed FedSA-GCL, so Theorem 1 does not constitute a convergence guarantee for the method whose results are reported in Section 6.
  2. [5.3 (proof sketch) and 6.1 (hyperparameters)] The proof sketch in Section 5.3 asserts that the effective staleness \tau_eff is bounded because the inner summation is proportional to d_i,t^{1-\alpha}, and the text states that this requires \alpha \ge 1. The experiments in Section 6.1 set \alpha = 0.4, so the configuration used for all main results lies outside the regime where the proof's staleness argument holds. Moreover, the bound in Eq. (15) contains an additive \epsilon_w^2 term that is not derived from K, \alpha, or the clustering threshold; it is simply assumed in Assumption 4 to be independent of training dynamics. The theorem is therefore conditional on an unverified assumption and does not cover the hyperparameter setting actually evaluated.
  3. [Tables 3-6 and Section 6.2] The central empirical claim that FedSA-GCL 'outperforms the baselines' is supported only by aggregate mean accuracies. The paper reports 95% confidence intervals but no significance tests, and per-dataset differences are often within one another's intervals or even reversed; for example, Table 3 on CiteSeer shows FedSA-GCL at 71.8±0.5 versus SWESALT at 73.1±0.3 (where SWESALT is higher), and Table 4 on CS shows FedSA-GCL at 89.2±0.3 versus FedGTA at 89.3±0.2 (where FedGTA is higher). Without paired significance testing across repeated runs or across datasets, the claimed average margins of 1.9% and 3.0% are not statistically established, which weakens the paper's core empirical contribution.
minor comments (5)
  1. [Section 4] The privacy analysis states that reverse-engineering individual node features or edge connectivity from SFM/LSC is 'computationally prohibitive,' but this claim is not quantified or supported by any information-theoretic bound; adding a formal argument or a citation to prior art would strengthen this section.
  2. [Section 6.1] The adaptation of SWESALT from self-supervised to supervised cross-entropy loss is a non-trivial modification that may affect its performance; the authors should provide more details on how the adaptation preserves the method's intended behavior, or alternatively discuss the potential impact on fairness of the comparison.
  3. [Section 1] The claim that FedSA-GCL is 'the first asynchronous FGL study that explicitly addresses the graph non-IID problem' is difficult to verify unconditionally; consider softening this claim or providing a more exhaustive literature comparison to substantiate it.
  4. [Data availability] The paper states 'Data will be made available on request' but does not provide a code repository; making the implementation available would substantially improve reproducibility, especially because the asynchronous simulation uses custom threading logic.
  5. [Section 6.6] The statement that 'the total number of client-to-server uploads remains mathematically fixed across all algorithms' is only true if all methods use the same number of local epochs per client trip; the sensitivity analysis in Section 6.4 shows that local epoch count varies and affects results, so this claim should be qualified.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the performance claims are measured against external baselines and the theoretical bound is an explicit conditional statement rather than a recycled input.

full rationale

No step in the derivation chain reduces a claimed prediction to a fitted input or to an equivalent self-citation. The central empirical claims, average margins of 1.9% under Louvain and 3.0% under Metis over ten baselines, are evaluated externally on eight public datasets with ablations, robustness tests, and hyperparameter sweeps, so they are not forced by construction. The convergence analysis in Section 5 rests on explicit assumptions, including Assumption 4 bounding the aggregation bias by epsilon_w^2; this is a stated condition rather than a hidden reuse of the conclusion, and no equation in the proof is identical to an input by construction. The LSC metric is adopted from FedGTA [13], whose authors overlap with the present paper, but the adoption is a design component; the paper's contribution, staleness-aware coupling and ClusterCast, is tested against baselines that include FedGTA itself, so the self-citation is not load-bearing for the reported gains. The proof analyzes a single global pseudo-gradient model in Eq. (14) while Algorithm 1 emits per-client personalized aggregates in Eq. (10), and the proof's alpha >= 1 condition does not match the experimental alpha = 0.4; these are correctness gaps rather than circularity. No renaming of a known result or imported uniqueness theorem appears in the manuscript.

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

The method introduces five tunable hyperparameters, the most consequential being the staleness exponent alpha, whose experimental value conflicts with the proof's alpha >=1 requirement. The convergence theorem rests on four assumptions, of which Assumption 4 is the most fragile because it assumes the central bias property instead of deriving it. No new physical or formal entities are introduced.

free parameters (5)
  • Similarity threshold theta = 0.5
    Eq. (7) cosine threshold for client clustering; set to 0.5 as the 'unbiased mathematical midpoint' and confirmed by sensitivity analysis. It directly controls which clients enter each peer set and receive ClusterCast broadcasts.
  • Staleness attenuation alpha = 0.4
    Eq. (10) exponent controlling the stale-update penalty. Chosen by sensitivity analysis, but the convergence proof in Section 5.3 requires alpha >=1 to bound effective staleness, so the evaluated value is outside the proven regime.
  • Propagation balance lambda = 0.5
    Eq. (8) interpolation coefficient in the Non-param LP used to compute LSC; empirically set and fixed.
  • Non-param LP steps k = 5
    Number of propagation steps in Eq. (8) for LSC; empirically set and fixed.
  • Semi-asynchronous buffer threshold K = 5
    Number of uploaded client updates that triggers a server aggregation, per Algorithm 1 and Section 6.1; it shapes the staleness distribution and convergence behavior.
assumptions (5)
  • standard math The cluster objective F_c(omega) is continuously differentiable and L-smooth (Assumption 1).
    Standard smoothness condition used in the descent inequality of Theorem 1.
  • domain assumption The variance of stochastic gradients within each cluster is bounded by sigma_c^2 (Assumption 2).
    This depends on SFM clustering producing semantically coherent clusters; no bound in terms of theta or data distributions is derived.
  • domain assumption The gradient norm is bounded by G^2 and staleness delay is bounded by tau_max (Assumption 3).
    Standard asynchronous FL assumptions; the protocol does not enforce a hard staleness bound.
  • ad hoc to paper Partial participation and non-uniform weighting cause an aggregation bias bounded by epsilon_w^2 independent of training dynamics (Assumption 4).
    This is the key property needed for the convergence neighborhood; it is assumed rather than derived from K, alpha, or LSC.
  • domain assumption Louvain and Metis subgraph partitions into 20 clients simulate realistic non-IID federated graph deployments.
    The empirical claim transfers from these simulated partitions to real distributed graph settings.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FedSA-GCL: A Semi-Asynchronous Federated Graph Learning Framework with Personalized Aggregation and Cluster-Aware Broadcasting." pith.science (2026). https://pith.science/paper/XV62GLMG

@misc{pith2026250718219,
  author       = {Pith},
  title        = {Pith review of: FedSA-GCL: A Semi-Asynchronous Federated Graph Learning Framework with Personalized Aggregation and Cluster-Aware Broadcasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XV62GLMG}},
  note         = {Machine review of arXiv:2507.18219}
}
read the original abstract

Federated Graph Learning (FGL) is a distributed learning paradigm that enables collaborative training over large-scale subgraphs located on multiple local systems. However, most existing FGL approaches rely on synchronous communication, which leads to inefficiencies and is often impractical in real-world deployments. Meanwhile, current asynchronous federated learning (AFL) methods are primarily designed for conventional tasks such as image classification and natural language processing, consequently failing to account for the unique topological properties of graph data. Directly applying these methods to graph learning frequently results in semantic drift and representational inconsistency within the global model. To address these challenges, we propose FedSA-GCL, a semi-asynchronous federated framework that leverages both inter-client label distribution divergence and graph topological characteristics through a novel ClusterCast mechanism for efficient training. We evaluate FedSA-GCL on multiple real-world graph datasets using the Louvain and Metis algorithms and conduct comparative analysis against 10 baselines. Extensive experiments demonstrate that our method achieves superior robustness and outstanding efficiency, outperforming the baselines by an average margin of 1.9% with Louvain and 3.0% with Metis.

Figures

Figures reproduced from arXiv: 2507.18219 by the authors.

Figure 1
Figure 1. Overview of the FedSA-GCL framework. 𝐘̂ 𝑖 = Softmax ( 𝑓( ̃𝜔𝑖 ; 𝐺𝑖 (𝑖 , 𝑖 ))) . (5) where 𝐺𝑖 (𝑖 , 𝑖 ) denotes the local subgraph of client 𝑖, and 𝑓( ̃𝜔𝑖 ; ⋅) is the predictive model parameterized by the aggregated weights ̃𝜔𝑖 . Since connected nodes often exhibit similarity in both feature distributions and labels, while different clients present distinct label distribution patterns, we construct a Soft Label Fea… view at source ↗
Figure 2
Figure 2. Comparison of communication mechanisms. ClusterCast proactively pushes model updates to both currently uploaded clients and inactive ones that share sim￾ilar graph structures. As shown in [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Comparison of per-round communication over [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: Client-wise label distributions for selected datasets split via Metis and Louvain, demonstrating the resulting non-IID [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: Performance comparison of different methods [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]
Figure 6
Figure 6. Figure 6: Client Similarity Evolution Across Client Trips on Cora and PubMed. [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Ablation Performance Comparison on Cora and [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]
Figure 8
Figure 8. Figure 8: Impact of local epochs on model performance [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: Sensitivity analysis of hyperparameters 𝛼 and 𝜃. The Computers dataset ( [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: Performance comparison under varying numbers [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 11
Figure 11. Figure 11: 3D Visualization of Client Trips across Methods. [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]
Figure 12
Figure 12. Figure 12: Total communication overhead breakdown against test accuracy. [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

48 extracted references · 42 canonical work pages

  1. [1]

    Z.Guo,H.Wang,Adeepgraphneuralnetwork-basedmechanismfor socialrecommendations,IEEETransactionsonIndustrialInformatics 17 (4) (2020) 2776–2783

    work page 2020

  2. [2]

    X. Liu, C. Song, F. Huang, H. Fu, W. Xiao, W. Zhang, Graphcdr: a graph neural network method with contrastive learning for cancer drug response prediction, Briefings in Bioinformatics 23 (1) (2022) bbab457

    work page 2022

  3. [3]

    H. Wang, C. Dai, Y. Wen, X. Wang, W. Liu, S. He, X. Bo, S. Peng, Gadrp: graph convolutional networks and autoencoders for cancer drug response prediction, Briefings in Bioinformatics 24 (1) (2023) bbac501

    work page 2023

  4. [4]

    L.He,X.Wang,D.Wang,H.Zou,H.Yin,G.Xu,Simplifyinggraph- based collaborative filtering for recommendation, in: Proceedings of the sixteenth ACM international conference on web search and data mining, 2023, pp. 60–68

    work page 2023

  5. [5]

    J. Yu, H. Yin, X. Xia, T. Chen, L. Cui, Q. V. H. Nguyen, Are graph augmentations necessary?, in: Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Informa- tion Retrieval, ACM, 2022, pp. 1294–1303

    work page 2022

  6. [6]

    X. Fu, B. Zhang, Y. Dong, C. Chen, J. Li, Federated graph machine learning: A survey of concepts, techniques, and applications, ACM SIGKDD Explorations Newsletter 24 (2) (2022) 32–47

    work page 2022

  7. [7]

    35, 2021, pp

    X.Zhang,C.Huang,Y.Xu,L.Xia,P.Dai,L.Bo,J.Zhang,Y.Zheng, Traffic flow forecasting with spatial-temporal graph diffusion net- work, in: Proceedings of the AAAI conference on artificial intelli- gence, Vol. 35, 2021, pp. 15008–15015

    work page 2021

  8. [8]

    M. Li, Z. Zhu, Spatial-temporal fusion graph neural networks for traffic flow forecasting, in: Proceedings of the AAAI conference on artificial intelligence, Vol. 35, 2021, pp. 4189–4196

    work page 2021

Show all 48 references
  1. [9]

    S. Yang, Z. Zhang, J. Zhou, Y. Wang, W. Sun, X. Zhong, Y. Fang, Q. Yu, Y. Qi, Financial risk analysis for smes with graph-based supplychainmining,in:Proceedingsofthetwenty-ninthinternational conferenceoninternationaljointconferencesonartificialintelligence, 2021, pp. 4661–4667

    2021

  2. [10]

    1396–1415

    J.Baek,W.Jeong,J.Jin,J.Yoon,S.J.Hwang,Personalizedsubgraph federated learning, in: International conference on machine learning, PMLR, 2023, pp. 1396–1415

    2023

  3. [11]

    H.Xie,J.Ma,L.Xiong,C.Yang,Federatedgraphclassificationover non-iid graphs, Advances in neural information processing systems 34 (2021) 18839–18852

    2021

  4. [12]

    Zhang, C

    K. Zhang, C. Yang, X. Li, L. Sun, S. M. Yiu, Subgraph federated learning with missing neighbor generation, Advances in Neural In- formation Processing Systems 34 (2021) 6671–6682

    2021

  5. [13]

    X. Li, Z. Wu, W. Zhang, Y. Zhu, R. Li, G. Wang, Fedgta: Topology- aware averaging for federated graph learning, Proc. VLDB Endow. 17 (1) (2023) 41–50.doi:10.14778/3617838.3617842. URLhttps://www.vldb.org/pvldb/vol17/p41-li.pdf

    2023

  6. [14]

    36, 2022, pp

    Y.Tan,G.Long,L.Liu,T.Zhou,Q.Lu,J.Jiang,C.Zhang,Fedproto: Federated prototype learning across heterogeneous clients, in: Pro- ceedings of the AAAI conference on artificial intelligence, Vol. 36, 2022, pp. 8432–8440

    2022

  7. [15]

    X. Li, Z. Wu, W. Zhang, H. Sun, R.-H. Li, G. Wang, Adafgl: A new paradigm for federated node classification with topology heterogene- ity,in:2024IEEE40thInternationalConferenceonDataEngineering (ICDE), IEEE, 2024, pp. 2517–2530

    2024

  8. [16]

    Z.Jiang,W.Wang,B.Li,B.Li,Pisces:Efficientfederatedlearningvia guidedasynchronoustraining,in:Proceedingsofthe13thSymposium on Cloud Computing, 2022, pp. 370–385

    2022

  9. [17]

    Y. Zhu, X. Li, Z. Wu, D. Wu, M. Hu, R. Li, Fedtad: Topology-aware data-free knowledge distillation for subgraph federated learning, in: Proceedings of the Thirty-Third International Joint Conference on Artificial Intelligence, IJCAI 2024, Jeju, South Korea, August 3-9, 2024, i...

    2024

  10. [18]

    Y. Chen, X. Sun, Y. Jin, Communication-efficient federated deep learning with layerwise asynchronous model update and temporally weighted aggregation, IEEE transactions on neural networks and learning systems 31 (10) (2019) 4229–4238

    2019

  11. [19]

    C. Xie, S. Koyejo, I. Gupta, Asynchronous federated optimization, arXiv preprint arXiv:1903.03934 (2019)

    2019 arXiv

  12. [20]

    C. Xu, Y. Qu, Y. Xiang, L. Gao, Asynchronous federated learning on heterogeneous devices: A survey, Computer Science Review 50 (2023) 100595

    2023

  13. [21]

    Q. Ma, Y. Xu, H. Xu, Z. Jiang, L. Huang, H. Huang, Fedsa: A semi-asynchronous federated learning mechanism in heterogeneous edgecomputing,IEEEJournalonSelectedAreasinCommunications 39 (12) (2021) 3654–3672

    2021

  14. [22]

    3581–3607

    J.Nguyen,K.Malik,H.Zhan,A.Yousefpour,M.Rabbat,M.Malek, D.Huba,Federatedlearningwithbufferedasynchronousaggregation, in: International conference on artificial intelligence and statistics, PMLR, 2022, pp. 3581–3607

    2022

  15. [23]

    Y. Liao, D. Wu, P. Lin, K. Guo, Accelerating unsupervised federated graph neural networks via semi-asynchronous communication, in: CCF Conference on Computer Supported Cooperative Work and Social Computing, Springer, 2023, pp. 378–392

    2023

  16. [24]

    C. Chen, Z. Xu, W. Hu, Z. Zheng, J. Zhang, Fedgl: Federated graph learning framework with global self-supervision, Information Sciences 657 (2024) 119976

    2024

  17. [25]

    Karypis, V

    G. Karypis, V. Kumar, A fast and high quality multilevel scheme for partitioning irregular graphs, SIAM Journal on scientific Computing 20 (1) (1998) 359–392

    1998

  18. [26]

    V. D. Blondel, J.-L. Guillaume, R. Lambiotte, E. Lefebvre, Fast unfolding of communities in large networks, Journal of statistical mechanics: theory and experiment 2008 (10) (2008) P10008

    2008

  19. [27]

    T. N. Kipf, M. Welling, Semi-supervised classification with graph convolutionalnetworks,in:5thInternationalConferenceonLearning Representations, ICLR 2017, Toulon, France, April 24-26, 2017, Conference Track Proceedings, OpenReview.net, 2017. URLhttps://openreview.net/forum?id...

    2017

  20. [28]

    W. L. Hamilton, Z. Ying, J. Leskovec, Inductive representation learning on large graphs, in: I. Guyon, U. von Luxburg, S. Bengio, H. M. Wallach, R. Fergus, S. V. N. Vishwanathan, R. Garnett Accepted byKnowledge-Based Systems, DOI: 10.1016/j.knosys.2026.116373Page 19 of 20 (Eds...

  21. [29]

    Velickovic, G

    P. Velickovic, G. Cucurull, A. Casanova, A. Romero, P. Liò, Y. Ben- gio, Graph attention networks, in: 6th International Conference on LearningRepresentations,ICLR2018,Vancouver,BC,Canada,April 30 - May 3, 2018, Conference Track Proceedings, OpenReview.net, 2018. URLhttps://op...

    2018

  22. [30]

    Precup, Y

    J.Gilmer,S.S.Schoenholz,P.F.Riley,O.Vinyals,G.E.Dahl,Neural message passing for quantum chemistry, in: D. Precup, Y. W. Teh (Eds.), Proceedings of the 34th International Conference on Machine Learning, ICML 2017, Sydney, NSW, Australia, 6-11 August 2017, ProceedingsofMachineLe...

    2017

  23. [31]

    K. Xu, W. Hu, J. Leskovec, S. Jegelka, How powerful are graph neural networks?, in: 7th International Conference on Learning Rep- resentations, ICLR 2019, New Orleans, LA, USA, May 6-9, 2019, OpenReview.net, 2019. URLhttps://openreview.net/forum?id=ryGs6iA5Km

    2019

  24. [32]

    J. Chen, T. Ma, C. Xiao, Fastgcn: Fast learning with graph con- volutional networks via importance sampling, in: 6th International Conference on Learning Representations, ICLR 2018, Vancouver, BC, Canada, April 30 - May 3, 2018, Conference Track Proceedings, OpenReview.net, 20...

    2018

  25. [33]

    Chiang, X

    W. Chiang, X. Liu, S. Si, Y. Li, S. Bengio, C. Hsieh, Cluster-gcn: Anefficientalgorithmfortrainingdeepandlargegraphconvolutional networks, in: A. Teredesai, V. Kumar, Y. Li, R. Rosales, E. Terzi, G. Karypis (Eds.), Proceedings of the 25th ACM SIGKDD Interna- tional Conference ...

    2019

  26. [34]

    6861–6871

    F.Wu,A.H.S.Jr.,T.Zhang,C.Fifty,T.Yu,K.Q.Weinberger,Sim- plifyinggraphconvolutionalnetworks,in:K.Chaudhuri,R.Salakhut- dinov (Eds.), Proceedings of the 36th International Conference on MachineLearning,ICML2019,9-15June2019,LongBeach,Califor- nia,USA,ProceedingsofMachineLearning...

    2019

  27. [35]

    Rossi, F

    E. Rossi, F. Frasca, B. Chamberlain, D. Eynard, M. M. Bronstein, F. Monti, SIGN: scalable inception graph neural networks, CoRR abs/2004.11198 (2020).arXiv:2004.11198. URLhttps://arxiv.org/abs/2004.11198

    2020 arXiv

  28. [36]

    H. Zhu, P. Koniusz, Simple spectral graph convolution, in: 9th In- ternational Conference on Learning Representations, ICLR 2021, Virtual Event, Austria, May 3-7, 2021, OpenReview.net, 2021. URLhttps://openreview.net/forum?id=CYO5T-YjWZV

    2021

  29. [37]

    M. Chen, Z. Wei, B. Ding, Y. Li, Y. Yuan, X. Du, J. Wen, Scalable graph neural networks via bidirectional propagation, in: H. Larochelle, M. Ranzato, R. Hadsell, M. Balcan, H. Lin (Eds.), Advances in Neural Information Processing Systems 33: Annual Conference on Neural Informa...

    2020

  30. [38]

    Zhang, Z

    W. Zhang, Z. Yin, Z. Sheng, Y. Li, W. Ouyang, X. Li, Y. Tao, Z.Yang,B.Cui,Graphattentionmulti-layerperceptron,in:A.Zhang, H. Rangwala (Eds.), KDD ’22: The 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Washington, DC, USA, August 14 - 18, 2022, ACM, 2022, p...

    2022

  31. [39]

    McMahan, E

    B. McMahan, E. Moore, D. Ramage, S. Hampson, B. A. y Arcas, Communication-efficient learning of deep networks from decentral- ized data, in: Artificial intelligence and statistics, PMLR, 2017, pp. 1273–1282

    2017

  32. [40]

    C. He, K. Balasubramanian, E. Ceyani, Y. Rong, P. Zhao, J. Huang, M. Annavaram, S. Avestimehr, Fedgraphnn: A federated learning system and benchmark for graph neural networks, in: ICLR 2021 Workshop on Distributed and Private Machine Learning (DPML), 2021

    2021

  33. [41]

    Y. Yao, Y. Li, X. Fan, J. Li, K. Liu, W. Jin, S. Ravi, P. S. Yu, C. Joe- Wong, Fedgraph: A research library and benchmark for federated graph learning, CoRR abs/2410.06340 (2024).arXiv:2410.06340, doi:10.48550/ARXIV.2410.06340. URLhttps://doi.org/10.48550/arXiv.2410.06340

    2024 doi arXiv

  34. [42]

    Zhang, L

    T. Zhang, L. Gao, S. Lee, M. Zhang, S. Avestimehr, Timelyfl: Heterogeneity-aware asynchronous federated learning with adaptive partial training, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 5064–5073

    2023

  35. [43]

    Klicpera, A

    J. Klicpera, A. Bojchevski, S. Günnemann, Predict then propagate: Graph neural networks meet personalized pagerank, in: 7th Inter- national Conference on Learning Representations, ICLR 2019, New Orleans, LA, USA, May 6-9, 2019, OpenReview.net, 2019. URLhttps://openreview.net/f...

    2019

  36. [44]

    Z. Yang, W. Cohen, R. Salakhudinov, Revisiting semi-supervised learning with graph embeddings, in: International conference on machine learning, PMLR, 2016, pp. 40–48

    2016

  37. [45]

    Shchur, M

    O. Shchur, M. Mumme, A. Bojchevski, S. Günnemann, Pitfalls of graph neural network evaluation, arXiv preprint arXiv:1811.05868 (2018)

    2018 arXiv

  38. [46]

    T. Li, A. K. Sahu, M. Zaheer, M. Sanjabi, A. Talwalkar, V. Smith, Federated optimization in heterogeneous networks, Proceedings of Machine learning and systems 2 (2020) 429–450

    2020

  39. [47]

    Q. Li, B. He, D. Song, Model-contrastive federated learning, in: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 10713–10722

    2021

  40. [48]

    X. Li, Y. Zhu, B. Pang, G. Yan, Y. Yan, Z. Li, Z. Wu, W. Zhang, R. Li, G. Wang, Openfgl: A comprehensive benchmark for federated graph learning, Proc. VLDB Endow. 18 (5) (2025) 1305–1320.doi: 10.14778/3718057.3718061. URLhttps://www.vldb.org/pvldb/vol18/p1305-li.pdf Accepted b...

    2025

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.