Pith. sign in

REVIEW 3 major objections 7 minor 1 cited by

Federated Continual Graph Learning

T0 review · 3 major / 7 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read POWER, the first federated continual graph learning framework, claims to beat all baselines on accuracy and forgetting across eight graph datasets by pairing local coverage replay with server-side pseudo-prototype transfer.

desk verdict A real first formulation of federated continual graph learning with a broadly evaluated framework; the central claim mostly survives the noise objection, but the pseudo-prototype module needs stronger validation. read the letter →

arxiv 2411.18919 v3 pith:LLS2QXDZ submitted 2024-11-28 cs.LG cs.AIcs.DBcs.SI

classification cs.LGcs.AIcs.DBcs.SI
keywords federatedcontinualgraphlearningneuralnetworkscatastrophicforgettingexperiencereplaypseudo-prototypereconstructionknowledgetransferclass-incremental
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

Until now, continual graph learning assumed one institution holds the whole evolving graph. This paper argues that real graph data lives across private databases, and proposes federated continual graph learning (FCGL) as the setting where multiple clients train GNNs on their own evolving graphs while a server aggregates their models. The paper reports that even naive federated training beats isolated continual graph learning, but two problems limit it: local graph forgetting within each client, and global expertise conflict, where averaging expert models leaves a global model worse than the best local expert. It then introduces POWER, which replays carefully selected experience nodes locally and performs trajectory-aware knowledge transfer on the server via reconstructed pseudo-prototypes. On eight datasets, POWER claims to consistently beat federated versions of existing CGL methods and vision-centric federated continual learning baselines on both accuracy and forgetting metrics.

What carries the argument

The argument runs on three coupled mechanisms. First, local-global coverage maximization picks experience nodes in embedding space using a weighted blend of local and global GNN embeddings, so replay samples represent a class from both one client's view and the federated view. Second, pseudo-prototype reconstruction lets the server learn class prototypes without raw features: clients upload gradients of class prototypes through a fixed randomly initialized encoding network, and the server recovers pseudo-prototypes by gradient matching, then builds a global buffer graph via K-nearest neighbors. Third, trajectory-aware knowledge transfer uses each client's cumulative label distribution over tasks (a decaying sum that encodes how its expertise evolved) as weights in a KL-divergence loss that pulls the global model toward the local experts most responsible for each class. Together these mechanisms target the two failure modes: replay fights LGF at the client, prototype transfer fights GEC at the server.

What would settle it

Measure the cosine similarity between each reconstructed pseudo-prototype and the true class feature centroid across the eight datasets; low similarity would falsify the claim that the global buffer graph carries genuine class expertise rather than noise from the random encoding network.

Watch

Extended reading notes

Core claim

The central discovery is that federated continual graph learning is not only feasible but necessary: with clients following divergent graph evolution trajectories, even a plain FedAvg combination of a continual graph learner outperforms isolated training, because different clients become experts in different classes at different times. The paper identifies two obstacles—local graph forgetting (LGF), the usual catastrophic forgetting within a client, and global expertise conflict (GEC), the loss of expertise that happens when parameters from differently specialized clients are averaged—and claims POWER resolves both. LGF is addressed by storing one experience node per class per task chosen to maximize combined local-global coverage in embedding space, then replaying those nodes during later tasks. GEC is addressed server-side: clients send gradient patterns of class prototypes rather than raw data, the server reconstructs pseudo-prototypes by gradient matching, builds a KNN graph over them, and uses each client's decaying cumulative label distribution (its evolution trajectory) to weight a KL-divergence distillation from local experts into the global model. The reported result is that POWER consistently beats the ten baselines in average accuracy and in forgetting across eight graph datasets spanning six domains.

Load-bearing premise

The load-bearing premise is that pseudo-prototypes reconstructed by gradient matching from a randomly initialized network are faithful stand-ins for each client's class expertise; if the reconstruction is noisy, the server-side buffer graph and knowledge transfer will spread distorted expertise.

Editorial extensions

If this is right

  • Federated deployment of continual graph learning is viable: the paper's experiments indicate that collective intelligence across clients outweighs isolated training even before any new algorithmic machinery is added.
  • Experience replay for graphs can be made federation-aware by choosing buffer nodes with both local and global embeddings; this yields better retention than selecting with local embeddings alone, as the ablation on Cora shows.
  • Raw class prototypes need not leave the client: gradient matching from a randomly initialized network supplies the server with usable class knowledge, keeping the raw-feature privacy boundary intact.
  • The extra communication cost of the GEC module is one small prototype-gradient upload per task, compared with GNN parameter uploads every round, so conflict resolution is cheap relative to standard federated communication.
  • The framework stays effective under missing features, missing edges, missing labels, and partial client participation, which are exactly the conditions decentralized evolving graphs present.

Reading between the lines

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

  • A direct audit of the reconstruction step—comparing reconstructed pseudo-prototypes against true class centroids in feature space—would tell whether the server's global buffer graph carries genuine class expertise; the paper does not perform this check.
  • The trajectory encoding is label-only: representing client trajectories by cumulative label histograms ignores topology changes, so a natural extension would incorporate graph-structure drift (edge additions, community shifts) into the trajectory weights.
  • The local-global coverage selector could be lifted out of the continual setting and used as a privacy-aware buffer construction for any federated graph learner, since it needs only embeddings from local and global models.
  • The KL-transfer step is server-side, so it should also benefit asynchronous or partially participating clients whose prototypes were stored in earlier rounds, a property the sparsity experiments support but do not fully stress-test.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 7 minor

Summary. The paper introduces Federated Continual Graph Learning (FCGL), a new problem setting in which multiple clients hold private evolving graphs and a central server coordinates GNN training under storage and privacy constraints. Through an empirical study on a Cora-based simulation, the authors identify two challenges: local graph forgetting (LGF), where a client loses prior knowledge when adapting to a new task, and global expertise conflict (GEC), where parameter aggregation across clients with divergent evolution trajectories makes the global GNN suboptimal on both past and current tasks. To address these, the paper proposes POWER, which (1) selects and replays experience nodes with maximum local-global coverage to mitigate LGF, and (2) reconstructs pseudo-prototypes from client gradient information and performs trajectory-aware knowledge transfer on a server-side global buffer graph to mitigate GEC. The method is evaluated on eight graph datasets against ten baselines, with additional ablations, sensitivity analyses, and sparsity experiments.

Significance. If the empirical results are taken at face value, the paper makes a useful contribution to the emerging area of federated continual graph learning. It is the first systematic treatment of the problem, with a wide benchmark (eight datasets across six domains), ten adapted baselines, an open-source implementation, and extensive secondary experiments (ablations, hyperparameter sensitivity, convergence, communication cost, and sparsity robustness). The LGF/GEC decomposition is intuitively grounded in the motivating empirical study, and the proposed two-module remedy is a reasonable first solution. The release of code and a technical report supports reproducibility. The main weaknesses are statistical: the headline claim of consistent superiority is not backed by significance testing, and the central pseudo-prototype mechanism is not directly validated for fidelity. These issues are fixable within the manuscript's scope rather than fatal.

major comments (3)
  1. [Sec. 5.2, Table 2] The central claim that POWER "consistently outperforms all baselines in both AM and FM" is based solely on means and standard deviations over ten runs, with no significance tests, confidence intervals, or paired comparisons. Several AM gaps are modest relative to the reported variance, for example Computers (43.26±2.94 vs. Fed-DSLR 40.32±5.30) and CiteSeer (54.47±5.21, where the reported standard deviation is larger than the gap to the best baseline). With eight datasets and ten baselines, the multiple-comparison issue is nontrivial. I recommend adding paired significance tests (e.g., Wilcoxon signed-rank over datasets or paired bootstrap over runs) and reporting effect sizes or confidence intervals for the AM and FM differences. The directional consistency across all eight datasets is suggestive but does not by itself establish the claim.
  2. [Sec. 4.2, Eqs. (8)-(11)] The pseudo-prototype reconstruction mechanism is never validated for fidelity. The only evidence for the GEC module is the full-module ablation in Table 3 ("w/o GEC"), which removes both pseudo-prototype reconstruction and trajectory-aware knowledge transfer together; the intra-module ablation only swaps the cumulative trajectory for a non-cumulative one. Because the global buffer graph in Eq. (13) and the knowledge-transfer loss in Eq. (14) are built on these reconstructed pseudo-prototypes, noisy reconstructions would propagate directly into the server-side training signal. Please add a direct fidelity check (e.g., cosine similarity or nearest-class accuracy between reconstructed pseudo-prototypes and true class-mean prototypes) and a sensitivity analysis over reconstruction hyperparameters (LBFGS iterations, random seeds), or an ablation that replaces pseudo-prototypes with oracle prototypes. Without this, the mechanism's contribution to GEC is not isolated.
  3. [Sec. 4.2, 'Pseudo Prototype Reconstruction Mechanism'] The privacy-preservation claim for prototype gradients is not supported. The text asserts that gradients from a randomly initialized network make it hard to extract meaningful information, but the paper itself shows that class labels are recoverable from the gradient sign (c = argmin_i ...), and gradient inversion is known to leak input information even through random networks. No formal privacy analysis, differential-privacy guarantee, or empirical attack evaluation is provided. Since privacy is one of the two motivating constraints of FCGL, this claim should be either substantiated or substantially qualified.
minor comments (7)
  1. [A.4 vs. Eq. (3)] The hyperparameter list in A.4 refers to a threshold distance η, while Eq. (3) defines the same quantity as ε; the notation should be unified.
  2. [Eq. (5)] Eq. (5) uses a binary cross-entropy form with a (1-y_i) log(1-ŷ_i) term for what is described as a multi-class one-hot label. If the classifier output is a softmax over classes, this loss is nonstandard; please clarify the intended output activation and loss formulation.
  3. [Table 2] In the FedSage+ row for Squirrel, the AM and FM columns report identical values (29.52±2.18 and 79.82±3.04); this looks like a typo and should be checked.
  4. [Eq. (12)] The notation p_k^i ∝ { sum ... } is ambiguous: please state explicitly whether p_k^i is normalized over classes or over samples, since the normalization affects the interpretation of the decayed cumulative trajectory q_t^k.
  5. [Sec. 5.4, Fig. 5] The sentence "a small φ may fail to capture it effectively" is confusing in light of Eq. (12), where smaller φ causes earlier tasks to decay more quickly; please rephrase to describe the effect of φ on the trajectory encoding.
  6. [Sec. 1, Contributions] The text contains a grammatical typo: "an novel and effective FCGL training framework" should read "a novel and effective FCGL training framework."
  7. [Algorithm 1] Line 36 of Algorithm 1 says "Reconstruct P̂ via ∇Γ_k^{t,c}" but the details of the LBFGS optimization and its hyperparameters appear only in A.4; a brief pointer in the main text would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: POWER is an empirical framework validated against external baselines; no equation reduces by construction to the reported result.

full rationale

The paper's central claim is an empirical comparison: 'Experiments on various graph datasets demonstrate POWER's superiority over federated adaptations of CGL baselines and vision-centric federated continual learning approaches.' This is evaluated against external datasets (Cora, CiteSeer, OGB-arxiv, Computers, Physics, Squirrel, Flickr, Roman-empire) and 10 adapted baselines (Table 2), with ablations (Table 3). The method's components — local-global coverage replay (Eqs. 2-7), pseudo-prototype reconstruction via gradient matching (Eqs. 9-11), and trajectory-aware knowledge transfer (Eqs. 12-14) — are constructive training objectives, not definitions of the reported AM/FM metrics. No fitted parameter is renamed as a prediction; the improvement is not encoded by construction in the equations. The two self-citations ([22] FedGTA and [65] FedTAD) appear as background references in federated graph learning and as baselines; they are not load-bearing for the central claim of superiority, which rests on external benchmark comparisons. The absence of significance testing in Table 2 is a statistical-evidence concern, not a circularity. Overall, no equation or citation chain makes the target result equivalent to its inputs.

Assumptions & free parameters 7 free parameters · 6 assumptions · 1 invented entities

The framework depends on several hand-chosen hyperparameters (alpha, beta, phi, epsilon, b, k_bar) that trade off components of the method, and on the core modeling assumption that gradient-matched pseudo-prototypes are meaningful enough to support server-side knowledge distillation. These are reasonable for an empirical ML paper, but they are not derived or externally benchmarked.

free parameters (7)
  • alpha = 0.5
    Trade-off in Eq. (2) balancing local and global embeddings for coverage-based node selection; fixed by hand.
  • beta = tuned near optimum (Fig. 5)
    Trade-off between new-task loss and replay loss in Eq. (7); exploratory hyperparameter.
  • phi = tuned near optimum (Fig. 5)
    Decay coefficient for past-task contributions in evolution trajectory Eq. (12); exploratory hyperparameter.
  • epsilon = searched in {0.01, 0.1, 0.5}
    Threshold distance in coverage Eq. (3); exploratory hyperparameter.
  • b = 1
    Number of experience nodes selected per task and class in Eq. (4); default buffer size.
  • k_bar = 1
    K-Nearest-Neighbor parameter for global buffer graph construction in Eq. (13).
  • LBFGS iterations and learning rate = 300 iterations, lr=1.0
    Optimization settings for pseudo-prototype reconstruction in Eq. (11).
assumptions (6)
  • domain assumption Class-incremental CGL setting where each task introduces new node classes and inter-task edges are removed.
    Standard CGL setup inherited from [8, 7, 25, 63], used to simulate client tasks in Section 3 and 5.1.
  • domain assumption GNNs experience catastrophic forgetting on streaming graph data.
    Standard continual learning premise, cited in Section 1 and used to define LGF.
  • domain assumption Louvain-based graph partitioning creates client subgraphs with divergent but complementary class trajectories.
    The FCGL simulation in Section 3 relies on Louvain partitioning to generate meaningful cross-client diversity; real data may not follow this structure.
  • ad hoc to paper Gradient matching on a randomly initialized encoding network yields pseudo-prototypes that capture class expertise.
    Eqs. (9)-(11) assume that matching prototype gradients is sufficient to reconstruct useful class prototypes; no formal guarantee or direct fidelity check is provided.
  • domain assumption A KNN graph over reconstructed pseudo-prototypes represents the global input space well enough for knowledge transfer.
    Eq. (13) builds the global buffer graph from pseudo-prototypes; the paper assumes this graph supports meaningful GNN predictions for the distillation loss in Eq. (14).
  • domain assumption A decayed cumulative label distribution is a good proxy for each client's evolving expertise.
    Eq. (12) encodes evolution trajectories using class-count history with decay; this is a modeling choice not derived from data properties.
invented entities (1)
  • Pseudo-prototype
    purpose: A trainable vector at the server reconstructed from client prototype gradients, used to populate the global buffer graph for knowledge transfer without sharing raw features.
    The paper provides no falsifiable handle for whether pseudo-prototypes truly match true class prototypes; validation is only indirect via final task accuracy.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Federated Continual Graph Learning." pith.science (2026). https://pith.science/paper/LLS2QXDZ

@misc{pith2026241118919,
  author       = {Pith},
  title        = {Pith review of: Federated Continual Graph Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LLS2QXDZ}},
  note         = {Machine review of arXiv:2411.18919}
}
read the original abstract

Managing evolving graph data presents substantial challenges in storage and privacy, and training graph neural networks (GNNs) on such data often leads to catastrophic forgetting, impairing performance on earlier tasks. Despite existing continual graph learning (CGL) methods mitigating this to some extent, they rely on centralized architectures and ignore the potential of distributed graph databases to leverage collective intelligence. To this end, we propose Federated Continual Graph Learning (FCGL) to adapt GNNs across multiple evolving graphs under storage and privacy constraints. Our empirical study highlights two core challenges: local graph forgetting (LGF), where clients lose prior knowledge when adapting to new tasks, and global expertise conflict (GEC), where the global GNN exhibits sub-optimal performance in both adapting to new tasks and retaining old ones, arising from inconsistent client expertise during server-side parameter aggregation. To address these, we introduce POWER, a framework that preserves experience nodes with maximum local-global coverage locally to mitigate LGF, and leverages pseudo-prototype reconstruction with trajectory-aware knowledge transfer to resolve GEC. Experiments on various graph datasets demonstrate POWER's superiority over federated adaptations of CGL baselines and vision-centric federated continual learning approaches.

Figures

Figures reproduced from arXiv: 2411.18919 by the authors.

Figure 1
Figure 1. Illustration of our FCGL paradigm, which achieves collabo [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Experimental results of our empirical study. (a) Node label distribution of each client’s class-incremental tasks, each exhibiting [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Overview of our proposed POWER framework. We follow a class-incremental setting, where each client collects an evolving graph [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Sensitivity analysis for the replay buffer size [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Sensitivity analysis for various combinations of the trade [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Convergence curves for the proposed POWER framework and six baseline methods across eight benchmark graph datasets observed [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: Performance under sparse settings for the proposed POWER framework and six baseline methods on the Physics dataset. [PITH_FULL_IMAGE:figures/full_fig_p014_7.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. A Comprehensive Data-centric Overview of Federated Graph Learning

    cs.LG 2025-07 conditional novelty 5.0 of 10

    A data-centric taxonomy for Federated Graph Learning that classifies 79 studies by data characteristics and data utilization, plus a discussion of integration with pre-trained large models.

Reference graph

Works this paper leans on

74 extracted references · 44 canonical work pages · cited by 1 Pith paper

  1. [1]

    Takuya Akiba, Shotaro Sano, Toshihiko Yanase, Takeru Ohta, and Masanori Koyama. 2019. Optuna: A next-generation hyperparameter optimization frame- work. In Proceedings of the 25th ACM SIGKDD international conference on knowl- edge discovery & data mining, KDD . 2623–2631

  2. [2]

    Jinheon Baek, Wonyong Jeong, Jiongdao Jin, Jaehong Yoon, and Sung Ju Hwang

  3. [3]

    Vicente Balmaseda, María Coronado, and Gonzalo de Cadenas-Santiagoc. 2023. Predicting Systemic Risk in Financial Systems Using Deep Graph Learning. In- telligent Systems with Applications (2023), 200240

  4. [4]

    Dongmin Bang, Sangsoo Lim, Sangseon Lee, and Sun Kim. 2023. Biomedical knowledge graph learning for drug repurposing by extending guilt-by-association to multiple layers. Nature Communications 14, 1 (2023), 3570

  5. [5]

    Vincent D Blondel, Jean-Loup Guillaume, Renaud Lambiotte, and Etienne Lefeb- vre. 2008. Fast unfolding of communities in large networks. Journal of statistical mechanics: theory and experiment 2008, 10 (2008), P10008

  6. [6]

    Chuan Chen, Weibo Hu, Ziyue Xu, and Zibin Zheng. 2021. FedGL: federated graph learning framework with global self-supervision. arXiv preprint arXiv:2105.03170 (2021)

  7. [7]

    Seungyoon Choi, Wonjoong Kim, Sungwon Kim, Yeonjun In, Sein Kim, and Chanyoung Park. 2024. DSLR: Diversity Enhancement and Structure Learning for Rehearsal-based Graph Continual Learning. In Proceedings of the ACM on Web Conference 2024. 733–744

  8. [8]

    Matthias De Lange, Rahaf Aljundi, Marc Masana, Sarah Parisot, Xu Jia, Aleš Leonardis, Gregory Slabaugh, and Tinne Tuytelaars. 2021. A continual learning survey: Defying forgetting in classification tasks. IEEE transactions on pattern analysis and machine intelligence 44, 7 (2021), 3366–3385

Show all 74 references
  1. [9]

    Jiahua Dong, Lixu Wang, Zhen Fang, Gan Sun, Shichao Xu, Xiao Wang, and Qi Zhu. 2022. Federated class-incremental learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 10164–10173

  2. [10]

    Dongqi Fu, Wenxuan Bao, Ross Maciejewski, Hanghang Tong, and Jingrui He

  3. [11]

    Xingbo Fu, Binchi Zhang, Yushun Dong, Chen Chen, and Jundong Li. 2022. Feder- ated graph machine learning: A survey of concepts, techniques, and applications. ACM SIGKDD Explorations Newsletter 24, 2 (2022), 32–47

  4. [12]

    ACM SIGKDD Explorations Newsletter 25, 1 (2023), 54–72

    Privacy-Preserving Graph Machine Learning from Data to Computation: A Survey. ACM SIGKDD Explorations Newsletter 25, 1 (2023), 54–72

  5. [13]

    Xiaoxue Han, Zhuo Feng, and Yue Ning. 2024. A topology-aware graph coars- ening framework for continual graph learning. Advances in Neural Information Processing Systems 37 (2024), 132491–132523

  6. [14]

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

  7. [15]

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

  8. [16]

    Chaoyang He, Keshav Balasubramanian, Emir Ceyani, Carl Yang, Han Xie, Lichao Sun, Lifang He, Liangwei Yang, S Yu Philip, Yu Rong, et al. 2021. FedGraphNN: A Federated Learning Benchmark System for Graph Neural Networks. In Inter- national Conference on Learning Representations...

  9. [17]

    Woochang Hyun, Jaehong Lee, and Bongwon Suh. 2023. Anti-Money Laundering in Cryptocurrency via Multi-Relational Graph Neural Network. In Pacific-Asia Conference on Knowledge Discovery and Data Mining . Springer, 118–130

  10. [18]

    Wenke Huang, Guancheng Wan, Mang Ye, and Bo Du. [n. d.]. Federated Graph Semantic and Structural Learning. ([n. d.])

  11. [19]

    Quentin Lhoest, Albert Villanova Del Moral, Yacine Jernite, Abhishek Thakur, Patrick Von Platen, Suraj Patil, Julien Chaumond, Mariama Drame, Julien Plu, Lewis Tunstall, et al. 2021. Datasets: A community library for natural language processing. arXiv preprint arXiv:2109.02846 (2021)

  12. [20]

    Diederik P Kingma. 2014. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014)

  13. [21]

    Xunkai Li, Zhengyu Wu, Wentao Zhang, Henan Sun, Rong-Hua Li, and Guoren Wang. 2024. AdaFGL: A New Paradigm for Federated Node Classification with Topology Heterogeneity. arXiv preprint arXiv:2401.11750 (2024)

  14. [22]

    Jialu Li, Yu Wang, Pengfei Zhu, Wanyu Lin, and Qinghua Hu. 2024. What matters in graph class incremental learning? An information preservation perspective. Advances in Neural Information Processing Systems 37 (2024), 26195–26223

  15. [23]

    Xunkai Li, Yinlin Zhu, Boyang Pang, Guochen Yan, Yeyu Yan, Zening Li, Zhengyu Wu, Wentao Zhang, Rong-Hua Li, and Guoren Wang. 2024. OpenFGL: A Comprehensive Benchmarks for Federated Graph Learning. arXiv preprint arXiv:2408.16288 (2024)

  16. [24]

    Xunkai Li, Zhengyu Wu, Wentao Zhang, Yinlin Zhu, Rong-Hua Li, and Guoren Wang. 2024. Fedgta: Topology-aware averaging for federated graph learning. arXiv preprint arXiv:2401.11755 (2024)

  17. [25]

    Huihui Liu, Yiding Yang, and Xinchao Wang. 2021. Overcoming catastrophic forgetting in graph neural networks. In Proceedings of the AAAI conference on artificial intelligence, Vol. 35. 8653–8661

  18. [26]

    Dong C Liu and Jorge Nocedal. 1989. On the limited memory BFGS method for large scale optimization. Mathematical programming 45, 1 (1989), 503–528

  19. [27]

    Divyam Madaan, Jaehong Yoon, Yuanchun Li, Yunxin Liu, and Sung Ju Hwang

  20. [28]

    Yilun Liu, Ruihong Qiu, and Zi Huang. 2023. Cat: Balanced continual graph learning with graph condensation. In 2023 IEEE International Conference on Data Mining (ICDM). IEEE, 1157–1162

  21. [29]

    Michael McCloskey and Neal J Cohen. 1989. Catastrophic interference in con- nectionist networks: The sequential learning problem. In Psychology of learning and motivation. Vol. 24. Elsevier, 109–165

  22. [30]

    Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Aguera y Arcas. 2017. Communication-efficient learning of deep networks from decentralized data. Artificial intelligence and statistics (2017), 1273–1282

  23. [31]

    Arun Mallya and Svetlana Lazebnik. 2018. Packnet: Adding multiple tasks to a single network by iterative pruning. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition . 7765–7773

  24. [32]

    Hongbin Pei, Bingzhe Wei, Kevin Chen-Chuan Chang, Yu Lei, and Bo Yang. 2020. Geom-gcn: Geometric graph convolutional networks. In International Conference on Learning Representations, ICLR

  25. [33]

    Oleg Platonov, Denis Kuznedelev, Michael Diskin, Artem Babenko, and Liud- mila Prokhorenkova. 2023. A critical look at the evaluation of GNNs under heterophily: are we really making progress? International Conference on Learning Representations, ICLR (2023)

  26. [34]

    Chaoxi Niu, Guansong Pang, Ling Chen, and Bing Liu. 2024. Replay-and-Forget- Free Graph Class-Incremental Learning: A Task Profiling and Prompting Ap- proach. arXiv preprint arXiv:2410.10341 (2024)

  27. [35]

    Zongshuai Qu, Tao Yao, Xinghui Liu, and Gang Wang. 2023. A Graph Con- volutional Network Based on Univariate Neurodegeneration Biomarker for Alzheimer’s Disease Diagnosis. IEEE Journal of Translational Engineering in Health and Medicine (2023)

  28. [36]

    Appan Rakaraddi, Lam Siew Kei, Mahardhika Pratama, and Marcus De Carvalho

  29. [37]

    Xiao Qin, Nasrullah Sheikh, Chuan Lei, Berthold Reinwald, and Giacomo Domeni- coni. 2023. Seign: A simple and efficient graph neural network for large dynamic graphs. In 2023 IEEE 39th International Conference on Data Engineering (ICDE) . IEEE, 2850–2863

  30. [38]

    Juho Salminen. 2012. Collective intelligence in humans: A literature review.arXiv preprint arXiv:1204.3401 (2012)

  31. [39]

    Oleksandr Shchur, Maximilian Mumme, Aleksandar Bojchevski, and Stephan Günnemann. 2018. Pitfalls of graph neural network evaluation. arXiv preprint arXiv:1811.05868 (2018)

  32. [40]

    Junwei Su, Difan Zou, Zijun Zhang, and Chuan Wu. 2023. Towards robust graph incremental learning on evolving graphs. In International Conference on Machine Learning. PMLR, 32728–32748

  33. [41]

    Roger Ratcliff. 1990. Connectionist models of recognition memory: constraints imposed by learning and forgetting functions. Psychological review 97, 2 (1990), 285

  34. [42]

    Petar Veličković, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio, and Yoshua Bengio. 2018. Graph attention networks. In International Confer- ence on Learning Representations, ICLR

  35. [43]

    Guancheng Wan, Wenke Huang, and Mang Ye. 2024. Federated graph learning under domain shift with generalizable prototypes. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 38. 15429–15437

  36. [44]

    Kuansan Wang, Zhihong Shen, Chiyuan Huang, Chieh-Han Wu, Yuxiao Dong, and Anshul Kanakia. 2020. Microsoft academic graph: When experts are not enough. Quantitative Science Studies 1, 1 (2020), 396–413

  37. [45]

    Minh-Tuan Tran, Trung Le, Xuan-May Le, Mehrtash Harandi, and Dinh Phung

  38. [46]

    Zhen Wang, Weirui Kuang, Yuexiang Xie, Liuyi Yao, Yaliang Li, Bolin Ding, and Jingren Zhou. 2022. FederatedScope-GNN: Towards a Unified, Comprehensive and Efficient Package for Federated Graph Learning. Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and D...

  39. [47]

    Lianghao Xia, Chao Huang, Jiao Shi, and Yong Xu. 2023. Graph-less Collaborative Filtering. In Proceedings of the ACM Web Conference, WWW . 17–27

  40. [48]

    Han Xie, Jing Ma, Li Xiong, and Carl Yang. 2021. Federated graph classification over non-iid graphs. Advances in Neural Information Processing Systems, NeurIPS (2021)

  41. [49]

    Liangwei Yang, Shengjie Wang, Yunzhe Tao, Jiankai Sun, Xiaolong Liu, Philip S Yu, and Taiqing Wang. 2023. DGRec: Graph Neural Network for Recommendation with Diversified Embedding Generation. In Proceedings of the Sixteenth ACM Federated Continual Graph Learning KDD ’25, Augus...

  42. [50]

    Shuzhe Wang, Zakaria Laskar, Iaroslav Melekhov, Xiaotian Li, and Juho Kannala

  43. [51]

    In Proceedings of the IEEE/CVF International Conference on Computer Vision

    Continual learning for image-based camera localization. In Proceedings of the IEEE/CVF International Conference on Computer Vision . 3252–3262

  44. [52]

    Yuhang Yao, Weizhao Jin, Srivatsan Ravi, and Carlee Joe-Wong. 2024. FedGCN: Convergence-communication tradeoffs in federated training of graph convolu- tional networks. Advances in neural information processing systems 36 (2024)

  45. [53]

    Jaehong Yoon, Wonyong Jeong, Giwoong Lee, Eunho Yang, and Sung Ju Hwang

  46. [54]

    Jaehong Yoon, Eunho Yang, Jeongtae Lee, and Sung Ju Hwang. 2017. Lifelong learning with dynamically expandable networks. arXiv preprint arXiv:1708.01547 (2017)

  47. [55]

    Hanqing Zeng, Hongkuan Zhou, Ajitesh Srivastava, Rajgopal Kannan, and Viktor Prasanna. 2020. Graphsaint: Graph sampling based inductive learning method. In International conference on learning representations, ICLR

  48. [56]

    Qiang Yang, Yang Liu, Yong Cheng, Yan Kang, Tianjian Chen, and Han Yu. 2019. Federated learning. Synthesis Lectures on Artificial Intelligence and Machine Learning 13, 3 (2019), 1–207

  49. [57]

    Cohen, and Ruslan Salakhutdinov

    Zhilin Yang, William W. Cohen, and Ruslan Salakhutdinov. 2016. Revisiting Semi-Supervised Learning with Graph Embeddings. In Proceedings of the 33rd International Conference on International Conference on Machine Learning, ICML . 40–48

  50. [58]

    Ke Zhang, Lichao Sun, Bolin Ding, Siu Ming Yiu, and Carl Yang. 2024. Deep Efficient Private Neighbor Generation for Subgraph Federated Learning. In Pro- ceedings of the 2024 SIAM International Conference on Data Mining (SDM) . SIAM, 806–814

  51. [59]

    Ke Zhang, Carl Yang, Xiaoxiao Li, Lichao Sun, and Siu Ming Yiu. 2021. Sub- graph federated learning with missing neighbor generation. Advances in Neural Information Processing Systems, NeurIPS (2021)

  52. [60]

    In Inter- national Conference on Machine Learning

    Federated continual learning with weighted inter-client transfer. In Inter- national Conference on Machine Learning . PMLR, 12073–12086

  53. [61]

    Xikun Zhang, Dongjin Song, and Dacheng Tao. 2022. Sparsified subgraph memory for continual graph representation learning. In2022 IEEE International Conference on Data Mining (ICDM) . IEEE, 1335–1340

  54. [62]

    Xikun Zhang, Dongjin Song, and Dacheng Tao. 2023. Ricci curvature-based graph sparsification for continual graph representation learning. IEEE Transactions on Neural Networks and Learning Systems (2023)

  55. [63]

    Huanding Zhang, Tao Shen, Fei Wu, Mingyang Yin, Hongxia Yang, and Chao Wu

  56. [64]

    arXiv preprint arXiv:2105.11099 (2021)

    Federated graph learning–a position paper. arXiv preprint arXiv:2105.11099 (2021)

  57. [65]

    Jie Zhang, Chen Chen, Weiming Zhuang, and Lingjuan Lyu. 2023. Target: Feder- ated class-continual learning via exemplar-free distillation. In Proceedings of the IEEE/CVF International Conference on Computer Vision . 4782–4793

  58. [68]

    Xikun Zhang, Dongjin Song, Yixin Chen, and Dacheng Tao. 2024. Topology- aware Embedding Memory for Continual Learning on Expanding Networks. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. 4326–4337

  59. [71]

    Fan Zhou and Chengtai Cao. 2021. Overcoming catastrophic forgetting in graph neural networks with experience replay. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 35. 4714–4722

  60. [72]

    Yinlin Zhu, Miao Hu, and Di Wu. 2025. POWER Technical Report. https://github. com/zyl24/FCGL_POWER/blob/main/POWER_Technical_Report.pdf. Accessed: 2025-05-22

  61. [73]

    Yinlin Zhu, Xunkai Li, Zhengyu Wu, Di Wu, Miao Hu, and Rong-Hua Li. 2024. FedTAD: Topology-aware Data-free Knowledge Distillation for Subgraph Feder- ated Learning. arXiv preprint arXiv:2404.14061 (2024). KDD ’25, August 3–7, 2025, Toronto, ON, Canada Yinlin Zhu, Miao Hu, Di W...

  62. [74]

    For all considered baselines, we adopt hyperparameters from the original publica- tions whenever possible

    with hyperparameters set to a learning rate of1× 10−2, weight decay of 5× 10−4, and a dropout rate of 0.5. For all considered baselines, we adopt hyperparameters from the original publica- tions whenever possible. In cases where these are not specified, we employ an automated ...

  63. [2021]

    arXiv preprint arXiv:2110.06976 (2021)

    Representational continuity for unsupervised continual learning. arXiv preprint arXiv:2110.06976 (2021)

  64. [2022]

    In Proceedings of the 31st ACM International Conference on Information & Knowledge Management

    Reinforced continual learning for graphs. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management . 1666–1674

  65. [2023]

    Personalized Subgraph Federated Learning. (2023)

  66. [2024]

    In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Text-enhanced data-free approach for federated class-incremental learn- ing. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 23870–23880

Pith tools

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