pith. sign in

arxiv: 2606.24237 · v1 · pith:RLERMHVEnew · submitted 2026-06-23 · 💻 cs.AI

Towards Federated Long-Tailed Graph Learning: An Energy-Guided Dual Decoupling Approach

Lianshuai Guo , Zhongzheng Yuan , Xunkai Li , Meixia Qu , Wenyu Wang This is my paper

Pith reviewed 2026-06-26 00:11 UTC · model grok-4.3

classification 💻 cs.AI
keywords federated graph learninglong-tailed distributionDirichlet energy pruninggraph neural networksNon-IID dataminority class performanceheterophilic graphs
0
0 comments X

The pith

FedEPD separates topological edge pruning from prototype injection to fix bias and isolation in federated long-tailed graphs.

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

The paper claims that long-tailed distributions in federated graph learning create two problems: the global model favors majority classes, and minority nodes get trapped in neighborhoods dominated by heterophilic edges from head classes. Existing fixes that ignore topology often overfit this noise instead of recovering tail representations. FedEPD counters both issues through a dual decoupling process that first prunes heterophilic edges using distribution-aware Dirichlet energy and then injects robust global prototypes drawn from central nodes. A two-stage optimization keeps majority decision boundaries intact while lifting minority accuracy. The reported gains reach 4.97 percent accuracy and 5.48 percent Macro-F1 on standard long-tailed benchmarks.

Core claim

FedEPD applies distribution-aware Dirichlet energy pruning to remove spatial heterophilic edges, then extracts global prototypes from topologically central nodes and injects them into local representations via spatial low-pass filtering; a two-stage alternating optimization protects majority boundaries while raising minority performance under Non-IID shifts.

What carries the argument

Dual decoupling paradigm that isolates topological purification (Dirichlet energy pruning) from semantic recalibration (central-node prototype injection).

If this is right

  • Minority nodes receive cleaner neighborhoods once heterophilic edges are filtered by energy scores.
  • Global prototypes drawn from central nodes remain usable across clients despite distribution shifts.
  • Majority-class boundaries stay stable because the two-stage optimizer alternates between the two decoupling steps.
  • The framework produces measurable lifts in both Accuracy and Macro-F1 on multiple long-tailed graph benchmarks.

Where Pith is reading between the lines

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

  • The same pruning-plus-prototype pattern could be tested on temporal or heterogeneous graphs where edge noise also follows degree imbalance.
  • If central-node prototypes prove reliable, federated systems might reduce communication rounds by sending only prototype summaries instead of full model updates.
  • The method implicitly assumes static graphs; extensions to dynamic settings would need to recompute energy scores after each edge arrival.

Load-bearing premise

Distribution-aware Dirichlet energy pruning reliably removes heterophilic edges without losing useful structure for minority nodes, and topologically central nodes supply stable global prototypes even under Non-IID client shifts.

What would settle it

A controlled ablation in which removing the energy-pruning step drops minority-node accuracy below the best baseline on the same long-tailed graph datasets, or in which prototype injection from central nodes fails to improve results under stronger Non-IID partitions.

Figures

Figures reproduced from arXiv: 2606.24237 by Lianshuai Guo, Meixia Qu, Wenyu Wang, Xunkai Li, Zhongzheng Yuan.

Figure 1
Figure 1. Figure 1: , manifesting as multiclass long-tailed distributions where a few majority classes dominate the topology, re￾sulting in a tail of minority classes characterized by data scarcity [14]. Within decentralized Non-IID environments, this statistical imbalance is locally amplified, resulting in data sparsity where individual clients lack structural samples for specific tail categories [34]. Furthermore, this prim… view at source ↗
Figure 2
Figure 2. Figure 2: Empirical analysis of class distribution and the Majority-Minority Optimization Conflict in federated graph learning. (a) The extreme multi-class long-tailed distributions within the CoraFull and Amazon-Electronics datasets, where classes are ranked by their local node frequencies. (b)-(d) Comparative Accuracy and F1-score evaluations across the (b) Head, (c) Medium, and (d) Tail partitions for four repres… view at source ↗
Figure 3
Figure 3. Figure 3: The overview of our proposed FedEPD framework. structural interference from adjacent majority classes. Fur￾thermore, while methods like 𝑆 2FGL [24] explore global alignment from a spectral perspective, they are primarily designed for general federated heterogeneity and do not address class imbalance. Consequently, these approaches remain susceptible to representation degradation when ma￾jority classes domi… view at source ↗
Figure 4
Figure 4. Figure 4: Category-level test accuracy across the long-tailed distribution on four datasets. The background gray shaded area illustrates the class sample sizes (right y-axis) sorted in descending order. To clearly visualize the performance trends, the sorted classes are uniformly aggregated into 10 bins along the x-axis. Solid lines track the average test accuracy (left y-axis) of different methods within each bin. … view at source ↗
Figure 5
Figure 5. Figure 5: illustrates the performance variations across differ￾ent pairings of these hyperparameters on the CoraFull and Amazon-Clothing datasets. Overall, FedEPD exhibits low sensitivity to these parameters, demonstrating the robustness of the dual decoupling paradigm. Specifically, 𝜇 controls the intensity of the semantic calibration for tail categories, while 𝛾 regulates the proportion of the robust global pro￾to… view at source ↗
Figure 6
Figure 6. Figure 6: Test accuracy convergence curves of our FedEPD and representative baselines over 200 training rounds on four datasets. (a) Amazon-Electronics 10 2 10 3 Running Time (s) Time Time Time Time (b) ogbn-arxiv 10 2 10 3 Running Time (s) Time Time Time Time 40 50 60 70 80 Test Accuracy (%) Acc Acc Acc Acc 30 40 50 60 70 80 90 Test Accuracy (%) Acc Acc Acc Acc Ours FedSpray GraphFedMig GraphSMOTE Metric: Time (s) … view at source ↗
Figure 7
Figure 7. Figure 7: Performance comparison on the (a) Amazon￾Electronics and (b) ogbn-arxiv datasets. The left vertical axis indicates the running time, and the right vertical axis represents the test accuracy. running time comparable to, or lower than, the baselines with the lowest computational costs. This efficiency advantage is supported by our complexity analysis. Computationally, as established in Section 3.1, the multi… view at source ↗
read the original abstract

Federated Graph Learning facilitates collaborative graph modeling across distributed clients while preserving data privacy. However, real-world data categories frequently exhibit long-tailed distributions. Such statistical scarcity severely degrades performance in two ways: it biases the global model toward majority classes, and it structurally isolates minority nodes by submerging them in heterophilic, head-dominated neighborhoods. While existing methods attempt topology-agnostic statistical compensations, they often fail under data scarcity. Instead of recovering tail nodes, they overfit the structural noise from adjacent dominant classes, leading to representation degradation. To address these limitations, we propose FedEPD, a framework built on a dual decoupling paradigm that separates topological purification from semantic recalibration. Specifically, FedEPD utilizes distribution-aware Dirichlet energy pruning to filter spatial heterophilic edges. It then overcomes Non-IID distribution shifts by extracting robust global prototypes from topologically central nodes, which are incorporated into local representations via a spatial low-pass prototype injection. Furthermore, a two stage alternating optimization strategy strictly protects majority decision boundaries while improving minority accuracy. Extensive experiments demonstrate that FedEPD achieves state-of-the-art performance across diverse long-tailed benchmarks, yielding absolute improvements of up to 4.97% in Accuracy and 5.48% in Macro-F1.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The paper proposes FedEPD, a federated graph learning framework for long-tailed distributions that employs a dual decoupling paradigm: distribution-aware Dirichlet energy pruning to filter heterophilic edges, followed by extraction of global prototypes from topologically central nodes that are injected into local representations via spatial low-pass filtering, protected by a two-stage alternating optimization strategy. The central empirical claim is that FedEPD achieves state-of-the-art performance across diverse long-tailed benchmarks, with absolute improvements of up to 4.97% in Accuracy and 5.48% in Macro-F1.

Significance. If the performance gains are reproducible and the method generalizes beyond the reported benchmarks, the dual-decoupling strategy could meaningfully advance handling of both statistical scarcity and structural heterophily in federated graph neural networks. The separation of topological purification from semantic recalibration addresses a recognized gap in existing topology-agnostic compensations.

major comments (1)
  1. Abstract: The claim that FedEPD achieves state-of-the-art performance with absolute improvements of up to 4.97% in Accuracy and 5.48% in Macro-F1 is presented without any experimental details, baselines, ablation studies, datasets, or error bars. This makes the central performance claim impossible to evaluate from the provided text.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive feedback. We address the single major comment below.

read point-by-point responses

  1. Referee: Abstract: The claim that FedEPD achieves state-of-the-art performance with absolute improvements of up to 4.97% in Accuracy and 5.48% in Macro-F1 is presented without any experimental details, baselines, ablation studies, datasets, or error bars. This makes the central performance claim impossible to evaluate from the provided text.

    Authors: We agree that the abstract presents the performance claim at a high level without accompanying experimental details. The full manuscript (Section 4) provides the requested information: datasets and long-tailed benchmarks, baselines, ablation studies, and results reported with error bars. To improve evaluability from the abstract alone, we will revise it to briefly reference the evaluation setting and direct readers to the experimental section for full details. This constitutes a partial revision given standard abstract length constraints. revision: partial

Circularity Check

0 steps flagged

No significant circularity

full rationale

The abstract and available description present FedEPD as an empirical framework using Dirichlet-energy pruning and prototype injection, with reported benchmark gains. No equations, first-principles derivations, or predictions appear that reduce by construction to fitted parameters, self-definitions, or self-citation chains. The central claims rest on experimental outcomes rather than any load-bearing step that equates to its own inputs. This is the normal case of a self-contained empirical paper with no detectable circularity in the derivation chain.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the unverified effectiveness of the proposed pruning and injection steps; no free parameters, invented entities, or additional axioms are extractable from the abstract alone.

axioms (1)
  • domain assumption Distribution-aware Dirichlet energy pruning can separate heterophilic edges from useful structure in long-tailed graphs.
    Invoked to justify the topological purification step.

pith-pipeline@v0.9.1-grok · 5766 in / 1260 out tokens · 26656 ms · 2026-06-26T00:11:24.526216+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

45 extracted references · 19 canonical work pages

  1. [1]

    Baek,J.,Jeong,W.,Jin,J.,Yoon,J.,Hwang,S.J.,2023. Personalized subgraph federated learning, in: Krause, A., Brunskill, E., Cho, K., Engelhardt,B.,Sabato,S.,Scarlett,J.(Eds.),InternationalConference on Machine Learning, ICML 2023, 23-29 July 2023, Honolulu, Hawaii, USA, PMLR. pp. 1396–1415. URL:https://proceedings. mlr.press/v202/baek23a.html

    2023

  2. [2]

    Deep gaussian embedding of graphs: Unsupervised inductive learning via ranking URL:https: //openreview.net/forum?id=r1ZdKJ-0W

    Bojchevski, A., Günnemann, S., 2018. Deep gaussian embedding of graphs: Unsupervised inductive learning via ranking URL:https: //openreview.net/forum?id=r1ZdKJ-0W

    2018

  3. [3]

    Towards federated long-tailed learning

    Chen, Z., Liu, S., Wang, H., Yang, H.H., Quek, T.Q.S., Liu, Z., 2022. Towards federated long-tailed learning. CoRR abs/2206.14988. URL:https://doi.org/10.48550/arXiv.2206.14988, doi:10.48550/ARXIV.2206.14988,arXiv:2206.14988

    work page doi:10.48550/arxiv.2206.14988 2022

  4. [4]

    Fu, X., Chen, Z., Zhang, B., Chen, C., Li, J., 2024. Federated graph learning with structure proxy alignment, in: Baeza-Yates, R., Bonchi,F.(Eds.),Proceedingsofthe30thACMSIGKDDConference on Knowledge Discovery and Data Mining, KDD 2024, Barcelona, Spain, August 25-29, 2024, ACM. pp. 827–838. URL:https://doi. org/10.1145/3637528.3671717, doi:10.1145/363752...

    work page doi:10.1145/3637528.3671717 2024

  5. [5]

    Open graph benchmark: Datasets for machine learning on graphs URL:https://proceedings.neurips.cc/ paper/2020/hash/fb60d411a5c5b72b2e7d3527cfc84fd0-Abstract.html

    Hu, W., Fey, M., Zitnik, M., Dong, Y., Ren, H., Liu, B., Catasta, M., Leskovec, J., 2020. Open graph benchmark: Datasets for machine learning on graphs URL:https://proceedings.neurips.cc/ paper/2020/hash/fb60d411a5c5b72b2e7d3527cfc84fd0-Abstract.html

    2020

  6. [6]

    Decoupling representation and classifier for long-tailed recognition URL:https://openreview.net/forum?id= r1gRTCVFvB

    Kang, B., Xie, S., Rohrbach, M., Yan, Z., Gordo, A., Feng, J., Kalantidis, Y., 2020. Decoupling representation and classifier for long-tailed recognition URL:https://openreview.net/forum?id= r1gRTCVFvB

    2020

  7. [7]

    Kelvinius, F.E., Georgiev, D., Toshev, A.P., Gasteiger, J., 2023. Accelerating molecular graph neural networks via knowledge dis- tillation URL:http://papers.nips.cc/paper_files/paper/2023/hash/ 51ec452ca04d8ec7160e5bbaf76153f6-Abstract-Conference.html

    2023

  8. [8]

    Li, J., Wang, J., Deng, R., Yan, D., Li, Q., 2024a. Fedgac: Graph federated learning with gradients aggregation calibration for non-iid and long-tailed data, in: IEEE International Symposium on Parallel and Distributed Processing with Applications, ISPA 2024, Kaifeng, China,October30-Nov.2,2024,IEEE.pp.410–417. URL:https:// doi.org/10.1109/ISPA63168.2024....

    work page doi:10.1109/ispa63168.2024.00059 2024

  9. [9]

    Federated optimization in heterogeneous networks URL:https://proceedings.mlsys.org/paper_files/paper/2020/hash/ 1f5fe83998a09396ebe6477d9475ba0c-Abstract.html

    Li, T., Sahu, A.K., Zaheer, M., Sanjabi, M., Talwalkar, A., Smith, V., 2020. Federated optimization in heterogeneous networks URL:https://proceedings.mlsys.org/paper_files/paper/2020/hash/ 1f5fe83998a09396ebe6477d9475ba0c-Abstract.html

    2020

  10. [10]

    Graphfedmig: Tackling class imbalance in federated graph learning via mutual information-guided generation

    Li, X., Fan, Q., Wang, T., Wei, K., Yu, K., Zhang, X., 2025a. Graphfedmig: Tackling class imbalance in federated graph learning via mutual information-guided generation. CoRR abs/2508.10471. URL:https://doi.org/10.48550/arXiv.2508.10471, doi:10.48550/ARXIV.2508.10471,arXiv:2508.10471

    work page doi:10.48550/arxiv.2508.10471

  11. [11]

    Li,X.,Wu,Z.,Zhang,W.,Sun,H.,Li,R.,Wang,G.,2024b.Adafgl:A new paradigm for federated node classification with topology hetero- geneity,in:40thIEEEInternationalConferenceonDataEngineering, ICDE 2024, Utrecht, The Netherlands, May 13-16, 2024, IEEE. pp. 2517–2530. URL:https://doi.org/10.1109/ICDE60146.2024.00198, doi:10.1109/ICDE60146.2024.00198

    work page doi:10.1109/icde60146.2024.00198 2024

  12. [12]

    Fedgta: Topology-awareaveragingforfederatedgraphlearning

    Li, X., Wu, Z., Zhang, W., Zhu, Y., Li, R., Wang, G., 2023. Fedgta: Topology-awareaveragingforfederatedgraphlearning. Proc.VLDB Endow. 17, 41–50. URL:https://www.vldb.org/pvldb/vol17/p41-li. pdf, doi:10.14778/3617838.3617842

    work page doi:10.14778/3617838.3617842 2023

  13. [13]

    Openfgl: A comprehensive benchmark for federated graph learning

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

    arXiv

  14. [15]

    (Eds.), KDD ’21: The 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Virtual Event, Singapore, August 14-18, 2021, ACM

    Liu,Z.,Nguyen,T.,Fang,Y.,2021b.Tail-gnn:Tail-nodegraphneural networks, in: Zhu, F., Ooi, B.C., Miao, C. (Eds.), KDD ’21: The 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, Virtual Event, Singapore, August 14-18, 2021, ACM. pp. 1109–1119. URL:https://doi.org/10.1145/3447548.3467276,doi:10. 1145/3447548.3467276

    work page doi:10.1145/3447548.3467276 2021

  15. [16]

    Fedlf: Adaptive logit adjustment and featureoptimizationinfederatedlong-tailedlearning,303–318URL: https://proceedings.mlr.press/v260/lu25a.html

    Lu, X., Li, P., Jiang, X., 2024. Fedlf: Adaptive logit adjustment and featureoptimizationinfederatedlong-tailedlearning,303–318URL: https://proceedings.mlr.press/v260/lu25a.html

    2024

  16. [17]

    Revisiting heterophily for graph neural networks, in:Koyejo,S.,Mohamed,S.,Agarwal,A.,Belgrave,D.,Cho,K.,Oh, A

    Luan, S., Hua, C., Lu, Q., Zhu, J., Zhao, M., Zhang, S., Chang, X., Precup, D., 2022. Revisiting heterophily for graph neural networks, in:Koyejo,S.,Mohamed,S.,Agarwal,A.,Belgrave,D.,Cho,K.,Oh, A. (Eds.), Advances in Neural Information Processing Systems 35: AnnualConferenceonNeuralInformationProcessingSystems2022, NeurIPS 2022, New Orleans, LA, USA, Nove...

    2022

  17. [18]

    Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V ., Goyal, N., K¨uttler, H., Lewis, M., Yih, W.-t., Rockt¨aschel, T., Riedel, S., and Kiela, D

    McAuley, J.J., Pandey, R., Leskovec, J., 2015. Inferring networks of substitutable and complementary products, in: Cao, L., Zhang, C., Joachims, T., Webb, G.I., Margineantu, D.D., Williams, G. (Eds.), Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Sydney, NSW, Australia, August 10-13, 2015, ACM. pp. 785...

    work page doi:10.1145/2783258.2783381 2015

  18. [19]

    McMahan, B., Moore, E., Ramage, D., Hampson, S., y Arcas, B.A.,

  19. [20]

    Communication-efficient learning of deep networks from decentralizeddata,in:Singh,A.,Zhu,X.J.(Eds.),Proceedingsofthe 20th International Conference on Artificial Intelligence and Statis- tics, AISTATS 2017, 20-22 April 2017, Fort Lauderdale, FL, USA, PMLR. pp. 1273–1282. URL:http://proceedings.mlr.press/v54/ mcmahan17a.html

    2017

  20. [21]

    A critical look at the evaluation of gnns under heterophily: Are we really making progress? URL: https://openreview.net/forum?id=tJbbQfw-5wv

    Platonov, O., Kuznedelev, D., Diskin, M., Babenko, A., Prokhorenkova, L., 2023. A critical look at the evaluation of gnns under heterophily: Are we really making progress? URL: https://openreview.net/forum?id=tJbbQfw-5wv. Lianshuai Guo et al.:Preprint submitted to ElsevierPage 12 of 13

    2023

  21. [22]

    Shang, X., Lu, Y., Huang, G., Wang, H., 2022. Federated learn- ing on heterogeneous and long-tailed data via classifier re-training with federated features , 2218–2224URL:https://doi.org/10.24963/ ijcai.2022/308, doi:10.24963/IJCAI.2022/308

    work page doi:10.24963/ijcai.2022/308 2022

  22. [23]

    Classifying long-tailed and label-noise data via disen- tangling and unlearning

    Shu, C., Li, M., Zhang, Y., Lu, Y., Han, B., Cheung, Y., Wang, H., 2025. Classifying long-tailed and label-noise data via disen- tangling and unlearning. CoRR abs/2503.11414. URL:https:// doi.org/10.48550/arXiv.2503.11414, doi:10.48550/ARXIV.2503.11414, arXiv:2503.11414

    work page doi:10.48550/arxiv.2503.11414 2025

  23. [24]

    Tan, Y., Long, G., Liu, L., Zhou, T., Lu, Q., Jiang, J., Zhang, C.,

  24. [25]

    Fedproto: Federated prototype learning across heterogeneous clients, in: Thirty-Sixth AAAI Conference on Artificial Intelligence, AAAI 2022, Thirty-Fourth Conference on Innovative Applications of Artificial Intelligence, IAAI 2022, The Twelveth Symposium on Educational Advances in Artificial Intelligence, EAAI 2022 Virtual Event, February 22 - March 1, 20...

    work page doi:10.1609/aaai.v36i8.20819 2022

  25. [26]

    Tan,Z.,Huang,S.,Wan,G.,Huang,W.,Li,H.,Ye,M.,2025. S2FGL: spatial spectral federated graph learning, in: Singh, A., Fazel, M., Hsu, D., Lacoste-Julien, S., Berkenkamp, F., Maharaj, T., Wagstaff, K.,Zhu,J.(Eds.),Forty-secondInternationalConferenceonMachine Learning, ICML 2025, Vancouver, BC, Canada, July 13-19, 2025, PMLR / OpenReview.net. URL:https://proce...

    2025

  26. [27]

    Tang, T., Han, Z., Cai, Z., Yu, S., Zhou, X., Oseni, T., Das, S.K.,

  27. [28]

    IEEETrans.NeuralNetworksLearn.Syst.35,11843– 11856

    Personalized federated graph learning on non-iid electronic healthrecords. IEEETrans.NeuralNetworksLearn.Syst.35,11843– 11856. URL:https://doi.org/10.1109/TNNLS.2024.3370297, doi:10. 1109/TNNLS.2024.3370297

    work page doi:10.1109/tnnls.2024.3370297 2024

  28. [29]

    Credit card fraud detection based on federated graph learning

    Tang, Y., Liang, Y., 2024. Credit card fraud detection based on federated graph learning. Expert Syst. Appl. 256, 124979. URL:https://doi.org/10.1016/j.eswa.2024.124979, doi:10.1016/J. ESWA.2024.124979

    work page doi:10.1016/j.eswa.2024.124979 2024

  29. [30]

    Wang, H., Jing, B., Ding, K., Zhu, Y., Cheng, W., Zhang, S., Fan, Y., Zhang,L.,Zhou,D.,2024. Masteringlong-tailcomplexityongraphs: Characterization, learning, and generalization, in: Baeza-Yates, R., Bonchi,F.(Eds.),Proceedingsofthe30thACMSIGKDDConference on Knowledge Discovery and Data Mining, KDD 2024, Barcelona, Spain, August 25-29, 2024, ACM. pp. 3045...

    work page doi:10.1145/3637528.3671880 2024

  30. [31]

    Fedgnn: Feder- ated graph neural network for privacy-preserving recommendation

    Wu, C., Wu, F., Cao, Y., Huang, Y., Xie, X., 2021. Fedgnn: Feder- ated graph neural network for privacy-preserving recommendation. CoRR abs/2102.04925. URL:https://arxiv.org/abs/2102.04925, arXiv:2102.04925

    arXiv 2021

  31. [32]

    Xiao, Z., Chen, Z., Liu, S., Wang, H., Feng, Y., Hao, J., Zhou, J.T., Wu, J., Yang, H.H., Liu, Z., 2023. Fed-grab: Federated long-tailed learning with self-adjusting gradient balancer URL:http://papers.nips.cc/paper_files/paper/2023/hash/ f4b8ddb9b1aa3cb11462d64a70b84db2-Abstract-Conference.html

    2023

  32. [33]

    Yan, S., Li, Z., Wu, C., Pang, M., Lu, Y., Yan, Y., Wang, H.,

  33. [34]

    URL:https://doi.org/10.48550/arXiv.2503.06916, doi:10.48550/ARXIV.2503.06916,arXiv:2503.06916

    You are your own best teacher: Achieving centralized- level performance in federated learning under heterogeneous and long-tailed data. URL:https://doi.org/10.48550/arXiv.2503.06916, doi:10.48550/ARXIV.2503.06916,arXiv:2503.06916

    work page doi:10.48550/arxiv.2503.06916

  34. [35]

    Yin,H.,Benson,A.R.,Leskovec,J.,Gleich,D.F.,2017. Localhigher- order graph clustering, in: Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Min- ing, Halifax, NS, Canada, August 13 - 17, 2017, ACM. pp. 555–

    2017

  35. [36]

    URL:https://doi.org/10.1145/3097983.3098069, doi:10.1145/ 3097983.3098069

    work page doi:10.1145/3097983.3098069

  36. [37]

    Yun, S., Kim, K., Yoon, K., Park, C., 2022. LTE4G: long-tail expertsforgraphneuralnetworks,in:Hasan,M.A.,Xiong,L.(Eds.), Proceedings of the 31st ACM International Conference on Informa- tion & Knowledge Management, Atlanta, GA, USA, October 17- 21, 2022, ACM. pp. 2434–2443. URL:https://doi.org/10.1145/ 3511808.3557381, doi:10.1145/3511808.3557381

    work page doi:10.1145/3511808.3557381 2022

  37. [38]

    Scalable diffusion models with transformers

    Zeng,Y.,Liu,L.,Liu,L.,Shen,L.,Liu,S.,Wu,B.,2023. Globalbal- anced experts for federated long-tailed learning, in: IEEE/CVF Inter- national Conference on Computer Vision, ICCV 2023, Paris, France, October1-6,2023,IEEE.pp.4792–4802. URL:https://doi.org/10. 1109/ICCV51070.2023.00444, doi:10.1109/ICCV51070.2023.00444

    work page doi:10.1109/iccv51070.2023.00444 2023

  38. [39]

    Federated learning with label distribution skew via logits calibration, in: Chaudhuri, K., Jegelka, S., Song, L., Szepesvári, C., Niu, G., Sabato, S

    Zhang, J., Li, Z., Li, B., Xu, J., Wu, S., Ding, S., Wu, C., 2022. Federated learning with label distribution skew via logits calibration, in: Chaudhuri, K., Jegelka, S., Song, L., Szepesvári, C., Niu, G., Sabato, S. (Eds.), International Conference on Machine Learning, ICML2022,17-23July2022,Baltimore,Maryland,USA,PMLR.pp. 26311–26329. URL:https://procee...

    2022

  39. [40]

    Sub- graph federated learning with missing neighbor generation , 6671–6682URL:https://proceedings.neurips.cc/paper/2021/hash/ 34adeb8e3242824038aa65460a47c29e-Abstract.html

    Zhang, K., Yang, C., Li, X., Sun, L., Yiu, S., 2021. Sub- graph federated learning with missing neighbor generation , 6671–6682URL:https://proceedings.neurips.cc/paper/2021/hash/ 34adeb8e3242824038aa65460a47c29e-Abstract.html

    2021

  40. [41]

    A federated graph learning method to realize multi-party collaboration for molecular discovery

    Zhang, L., Zhang, J., Huang, R., Wang, Y., Liu, L., 2026. A federated graph learning method to realize multi-party collaboration for molecular discovery. ResearchGate Preprint URL:https://www. researchgate.net/publication/400648352

    arXiv 2026

  41. [42]

    Pop:Apipeline-oriented paradigm for long-tailed learning survey

    Zhang, R., Yang, J., Liu, Y., Shang, C., Li, M., Shi, J.X., Wang, P., Ma,Y.,Han,B.,Cheung,Y.m.,etal.,2025. Pop:Apipeline-oriented paradigm for long-tailed learning survey

    2025

  42. [43]

    Graphsmote: Imbalanced node classificationongraphswithgraphneuralnetworks,in:Lewin-Eytan, L., Carmel, D., Yom-Tov, E., Agichtein, E., Gabrilovich, E

    Zhao, T., Zhang, X., Wang, S., 2021. Graphsmote: Imbalanced node classificationongraphswithgraphneuralnetworks,in:Lewin-Eytan, L., Carmel, D., Yom-Tov, E., Agichtein, E., Gabrilovich, E. (Eds.), WSDM ’21, The Fourteenth ACM International Conference on Web Search and Data Mining, Virtual Event, Israel, March 8-12, 2021, ACM.pp.833–841. URL:https://doi.org/...

    work page doi:10.1145/3437963.3441720 2021

  43. [44]

    Zhu, J., Yan, Y., Zhao, L., Heimann, M., Akoglu, L., Koutra, D.,

  44. [45]

    (Eds.), Advances in Neural Informa- tion Processing Systems 33: Annual Conference on Neural Infor- mation Processing Systems 2020, NeurIPS 2020, December 6-12, 2020, virtual

    Beyond homophily in graph neural networks: Current limi- tations and effective designs, in: Larochelle, H., Ranzato, M., Had- sell, R., Balcan, M., Lin, H. (Eds.), Advances in Neural Informa- tion Processing Systems 33: Annual Conference on Neural Infor- mation Processing Systems 2020, NeurIPS 2020, December 6-12, 2020, virtual. URL:https://proceedings.ne...

    2020

  45. [46]

    Fedtad: Topology-aware data-free knowledge distillation for subgraph feder- ated learning , 5716–5724URL:https://www.ijcai.org/proceedings/ 2024/632

    Zhu, Y., Li, X., Wu, Z., Wu, D., Hu, M., Li, R., 2024. Fedtad: Topology-aware data-free knowledge distillation for subgraph feder- ated learning , 5716–5724URL:https://www.ijcai.org/proceedings/ 2024/632. Lianshuai Guo et al.:Preprint submitted to ElsevierPage 13 of 13

    2024