Pith. sign in

REVIEW 5 major objections 6 minor 41 references

Overcoming Challenges of Partial Client Participation in Federated Learning : A Comprehensive Review

T0 review · 5 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A survey of federated learning argues that skipped client rounds deserve their own toolbox and maps it.

desk verdict A useful taxonomy of partial-participation FL methods, but with several demonstrable citation errors that undermine its reliability as an entry-point reference. read the letter →

arxiv 2506.02887 v2 pith:GDIYS7MI submitted 2025-06-03 cs.LG cs.DC

Mrinmay Sen , Shruti Aparna , Rohit Agarwal , Chalavadi Krishna Mohan This is my paper
classification cs.LGcs.DC
keywords federatedlearningpartialclientparticipationselectioncommunicationefficiencyvariancereductiontaxonomysurveynon-IIDdata
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 learning normally assumes every client joins every training round, but real deployments have devices that drop out, run out of battery, or miss rounds. This survey argues that partial client participation is a distinct, under-addressed challenge and that the scattered solutions can be organized into a usable map: client selection, aggregation strategies, communication optimization, and resource-aware scheduling. Across eleven representative frameworks—from FedAvg baselines to variance-reduction, memory, momentum, and server-aided methods—the survey finds no universally best option; gains depend on data distribution, participation rate, and resource constraints. If the mapping is accurate, the paper gives practitioners a structured entry point for choosing a method and researchers a common frame for comparing future work.

What carries the argument

The central organizing device is a taxonomy of partial-participation remedies, anchored to a formal problem statement in which the global objective is minimized over the participating subset $\tilde{\mathcal{C}}$ rather than all $M$ clients (Eq. 2). The taxonomy groups methods by mechanism—selective layer training, cyclic or anchor-based client sampling, gradient compression with error feedback, gradient memory, server-side auxiliary data, variance-reduction control variates, and global momentum—and the comparative tables (I through VI) translate each framework into a common format of objective, sampling, convergence guarantee, aggregation rule, update rule, and communication efficiency. This common format is what lets a reader see trade-offs across methods side by side.

What would settle it

Check Tables I through VI against their cited sources: if reference [15] does not describe error-feedback gradient compression, or if reference [4] does not contain the reported $O(1/\sqrt{nT})$ convergence rate for generalized FedAvg, the survey's comparative tables cannot be trusted. A reader could also re-run one method from each of the four taxonomy families under a fixed heterogeneity and participation protocol and see whether the qualitative rankings in Table VI reproduce.

Watch

Extended reading notes

Core claim

The paper's central claim is that partial client participation—where a random subset $\tilde{\mathcal{C}}$ of $m < M$ clients updates the global model each round—creates distinct convergence, fairness, and resource problems, and that existing remedies form a coherent landscape rather than isolated tricks. The survey organizes the landscape into four technical families: client selection, aggregation strategies, communication optimization, and resource-aware scheduling. It then profiles eleven frameworks, including FedProx, SCAFFOLD, FedVARP, MIFA, and FedCM, giving each one's objective, sampling rule, aggregation rule, convergence guarantee, and reported performance in Tables I through VI. The paper's overall conclusion is that no single framework dominates: memory-based and variance-reduction methods usually beat FedAvg under low participation, but only at added storage or communication cost, so deployment constraints must decide the choice.

Load-bearing premise

The survey's usefulness depends on each method description and table entry faithfully matching the cited source; in this text that premise is threatened at least where the Fed-EF section cites a distributed K-FAC paper ([15]), the generalized FedAvg convergence rate cites a general survey ([4]), and Table I's Embracing FL entry cites a different paper ([18]) than the section text does ([13]). If those mappings are wrong, the comparative tables cannot be trusted without checking the originals.

Editorial extensions

If this is right

  • A practitioner choosing among FedProx, SCAFFOLD, FedVARP, MIFA, FedCM, and the other profiled methods can use Tables I through VI to shortlist by their dominant constraint: bandwidth, non-IID drift, dropout resilience, or server memory.
  • The comparison implies that variance-reduction and memory-based methods gain the most over FedAvg when participation rates are low, and that these gains shrink as more clients join each round.
  • Because most evaluated frameworks concentrate on MNIST, CIFAR-10, and FEMNIST, the survey supports conclusions about image classification more strongly than about language or time-series tasks.
  • The survey's finding that no single framework dominates across all criteria implies that reported accuracy numbers are meaningful only alongside participation rates, heterogeneity settings, and resource budgets.

Reading between the lines

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

  • One consequence the paper leaves implicit is that Table VI's accuracy numbers are descriptive, not comparative: the figures come from different models, client counts, and training budgets, so a reader should not rank methods by those numbers alone.
  • The four-family taxonomy could be stress-tested by applying it to partial-participation methods published after this survey, such as long-horizon client-selection strategies, to see whether a fifth family is needed.
  • A natural test of the survey's organizational claim would be a head-to-head benchmark that runs one representative from each family under identical heterogeneity, participation, and communication budgets, converting the qualitative trade-offs into quantitative curves.
  • Because the survey's value depends on accurate citation mapping, the first check before relying on any table entry is to verify that the cited paper actually contains the described mechanism.
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

5 major / 6 minor

Summary. This paper is a survey of federated learning (FL) methods that address partial client participation, a setting where only a subset of clients join each training round. The survey organizes existing solutions into a taxonomy of client selection, aggregation strategies, communication optimization, and resource-aware scheduling, and provides six comparative tables covering mathematical formulations, advantages/disadvantages, benchmark settings, research objectives, technical categories, and reported performance. Twelve methods are reviewed, including FedAvg, FedProx, SCAFFOLD, FedVARP, MIFA, FedCM, GradMA, FedAMD, EmbracingFL, CyCP, Fed-EF, SAFARI, and a generalized FedAvg variant, each with a description, a claimed theoretical analysis, experiment summary, and limitations.

Significance. If the survey's descriptions and citations were accurate, it would be a useful entry point for researchers selecting methods for partial-participation FL, a practically important and growing area. The paper itself contributes no new algorithms or experiments; its value lies entirely in faithful curation and comparison. That value is currently compromised by multiple demonstrable citation and representation errors in the body and in Tables I, III, and VI. Because these are local, fixable problems rather than irreparable flaws in the survey's scope, the appropriate outcome is a major revision after a thorough citation audit. Should the authors correct the misattributions and add sources to Table VI, the survey could serve its intended purpose.

major comments (5)
  1. [Section III-C, Table I] Fed-EF is attributed to reference [15], which is Pauloski et al., 'Convolutional neural network training with distributed K-FAC' (SC 2020). That paper is about distributed second-order optimization and does not present the error-feedback federated algorithm described in Section III-C. The described SGD and AMSGrad variants with per-client error accumulators must be cited to the actual Fed-EF paper; as written, the Fed-EF rows in Tables I and VI are unverifiable.
  2. [Section III-F, Table I] The 'Generalized FedAvg' algorithm and its convergence rate O(sqrt(K/(nT)) + 1/T) are attributed to reference [4], which is Li, Wen & He, 'Federated learning with decentralized data: A survey' (ACM CSUR 2020). A survey cannot be the source of a specific algorithm with two-sided learning rates and a specific convergence guarantee. The original research paper must be cited, otherwise the claimed rate and the algorithm description in Section III-F cannot be checked.
  3. [Section III-A vs. Table I] The body text describes 'Embracing FL' with reference [13] (Lee et al., 'Embracing federated learning: Enabling weak client participation via partial model training'), and the experiments and model descriptions in Section III-A and Table III match that paper. Table I, however, cites reference [18] (Yang et al., 'Embracing system and statistical heterogeneity for federated learning with partial model personalization'), which is a different paper with a different objective. At least one of these citations is wrong, and the table currently disagrees with the text.
  4. [Sections III-E and III-J] Two additional methods show citation mismatches. Section III-E describes FedAMD's anchor-miner sampling and attributes it to [9], but [9] is a different FedAMD paper (Li et al., CIKM 2021) about heterogeneous client sampling; the described anchor-miner mechanism matches reference [20] (Wu et al., 'Anchor sampling for federated learning with partial client participation'), which is never cited in the body. Similarly, Section III-J's FedVARP and ClusterFedVARP description matches reference [19] (Jhunjhunwala et al., UAI 2022), while the text cites reference [8]; [19] is never cited. These errors are load-bearing because they corrupt the method-to-source mapping that the survey's comparative tables depend on.
  5. [Table VI] Table VI lists specific accuracy numbers (e.g., Fed-EF 90-92% non-IID, SCAFFOLD 80% non-IID) with no source column, even though the survey does not run its own experiments. Without per-row citations, these numbers cannot be verified or attributed, which defeats the comparative purpose of the table. In addition, the table lists FedCM with ResNet-20, while Section III-I's experiment description uses ResNet-18; this internal inconsistency makes the table unreliable even as a summary.
minor comments (6)
  1. [Various sections] The manuscript contains numerous typos and grammatical errors, including 'categories' for 'categorize', 'emphasis' for 'emphasize', 'THe' for 'The', 'Intialization', and 'Aggregration'. A thorough proofread is needed.
  2. [Sections III-B, III-H, etc.] Several experimental subsections are written in the first person as though the survey authors ran the experiments (e.g., 'We evaluate the performance of Cyclic Client Participation (CyCP) in FedAvg' in Section III-B). Since this is a survey, the results should be attributed to the original papers, e.g., 'In [14], the authors evaluate...'.
  3. [References [32] and [36]] References [32] and [36] share the same journal, volume, pages, and overlapping author lists, yet have different titles. This may be a duplicate entry or a metadata error; please verify.
  4. [Table I, Convergence Guarantee column] The 'Convergence Guarantee' entries are vague and sometimes inconsistent with the text: for example, Fed-EF is listed as O(1/T), but Section III-C only claims rates comparable to full-precision FL, and the conditions are not stated. The column should either give the exact theorem statements or be removed.
  5. [Table IV] Table IV groups FedCM under 'Reduce Communication Cost', but Section III-I's limitation paragraph states that FedCM introduces additional communication overhead from server to clients. This contradiction should be resolved.
  6. [Section I-C and Related Work] The introduction claims partial participation is under-addressed in existing surveys, but the reference list includes several recent surveys. A dedicated related-work paragraph distinguishing this survey from [1], [2], and [4] would make the claimed gap precise.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular reasoning found: the survey's conclusions are imported from external sources, and its internal empirical claims are not derived from fitted inputs or self-cited premises.

full rationale

This paper is a literature survey, not a derivation. Its central claims are taxonomic and comparative: it organizes existing partial-client-participation methods into categories and summarizes their properties, convergence rates, and experimental results. None of these claims are obtained by fitting a parameter to a subset of data and then predicting a closely related quantity. The problem definition in Eqs. (1) and (2) merely restates the standard federated averaging objective with a subsampled client set, which is the subject of the survey rather than a result derived from the survey's own inputs. Cited convergence guarantees such as O(1/sqrt(T)) or O(1/sqrt(nT)) are attributed to the original external method papers (e.g., [4], [14]) and are not re-derived here; even if some of those attributions are incorrect, the error is one of source mapping or scholarly accuracy, not a circular loop. There are no load-bearing self-citations: the reference list consists of external works by other research groups, and no uniqueness theorem or ansatz from the present authors' prior work is invoked to force a conclusion. The internal 'Experiment' passages (e.g., the CyCP discussion in Section III-B) report empirical observations presented as evidence, not predictions derived by construction from the survey's own equations. Consequently, none of the enumerated circularity patterns apply, and the appropriate finding is no significant circularity.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

For a survey, the free-parameter and invented-entity ledgers are empty: the paper contributes no fitted constants and postulates no new entities. The load-bearing premises are representational: method descriptions must match their citations (violated for Fed-EF [15] and generalized FedAvg [4]), and comparison numbers must come from a defined protocol (violated by Table VI). These premises are the real axioms a reader pays for upstream.

assumptions (3)
  • domain assumption The description of each method accurately reflects its cited source.
    A survey's comparisons inherit the accuracy of its citations. This fails for Fed-EF (Section III-C vs. reference [15], a distributed K-FAC paper) and for the generalized FedAvg algorithm (Section III-F vs. reference [4], a survey).
  • domain assumption Table VI's accuracy numbers are drawn from comparable, correctly attributed experimental protocols.
    The table reports unified per-method accuracy (e.g., FedAvg 72/78%, SCAFFOLD 80/83%) with no protocol definition, no error bars, and no citations, so the premise of comparability is unsupported.
  • domain assumption The taxonomy categories (client selection, aggregation, communication optimization, resource-aware scheduling) are exhaustive over the partial-participation literature.
    The survey selects twelve methods without an inclusion protocol and without comparing against client-selection survey taxonomies, so exhaustiveness is asserted, not demonstrated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Overcoming Challenges of Partial Client Participation in Federated Learning : A Comprehensive Review." pith.science (2026). https://pith.science/paper/GDIYS7MI

@misc{pith2026250602887,
  author       = {Pith},
  title        = {Pith review of: Overcoming Challenges of Partial Client Participation in Federated Learning : A Comprehensive Review},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GDIYS7MI}},
  note         = {Machine review of arXiv:2506.02887}
}
read the original abstract

Federated Learning (FL) is a learning mechanism that falls under the distributed training umbrella, which collaboratively trains a shared global model without disclosing the raw data from different clients. This paper presents an extensive survey on the impact of partial client participation in federated learning. While much of the existing research focuses on addressing issues such as generalization, robustness, and fairness caused by data heterogeneity under the assumption of full client participation, limited attention has been given to the practical and theoretical challenges arising from partial client participation, which is common in real-world scenarios. This survey provides an in-depth review of existing FL methods designed to cope with partial client participation. We offer a comprehensive analysis supported by theoretical insights and empirical findings, along with a structured categorization of these methods, highlighting their respective advantages and disadvantages.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

41 extracted references · 35 canonical work pages

  1. [15]

    Convolu- tional neural network training with distributed K-FAC,

    J. G. Pauloski, Z. Zhang, L. Huang, W. Xu, and I. T. Foster, “Convolu- tional neural network training with distributed K-FAC,” in Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis (SC) . IEEE/ACM, 2020, p. 94

    work page 2020

  2. [4]

    Federated learning with decentralized data: A survey,

    Q. Li, Z. Wen, and B. He, “Federated learning with decentralized data: A survey,” ACM Computing Surveys (CSUR) , vol. 53, no. 2, pp. 1–37, 2020

    work page 2020

  3. [18]

    Embracing system and statistical heterogeneity for federated learning with partial model personalization,

    Y . Yang, M. Ding, J. Li, L. Song, J. Liu, and H. V . Poor, “Embracing system and statistical heterogeneity for federated learning with partial model personalization,” IEEE Transactions on Neural Networks and Learning Systems, pp. 1–15, 2022

    work page 2022

  4. [13]

    Embracing Federated Learning: Enabling Weak Client Participation via Partial Model Training

    S. Lee, T. Zhang, S. Prakash, Y . Niu, and S. Avestimehr, “Embracing federated learning: Enabling weak client participation via partial model training,” arXiv preprint arXiv:2406.15125 , 2024. [Online]. Available: https://arxiv.org/abs/2406.15125

    work page Pith review arXiv 2024

  5. [9]

    Fedamd: Efficient federated learning via heterogeneous client sampling and adaptive model updates,

    L. Li, B. He, L. Fan, and J. Chen, “Fedamd: Efficient federated learning via heterogeneous client sampling and adaptive model updates,” in Proceedings of the 30th ACM International Conference on Information and Knowledge Management (CIKM) . ACM, 2021, pp. 385–394

    work page 2021

  6. [20]

    Anchor Sampling for Federated Learning with Partial Client Participation

    F. Wu, S. Guo, Z. Qu, S. He, Z. Liu, and J. Gao, “Anchor sampling for federated learning with partial client participation,” arXiv preprint arXiv:2206.05891 , 2022. [Online]. Available: https: //arxiv.org/abs/2206.05891

    work page Pith review arXiv 2022

  7. [19]

    Fedvarp: Tackling the variance due to partial client participation in federated learning,

    D. Jhunjhunwala, P. Sharma, A. Nagarkatti, and G. Joshi, “Fedvarp: Tackling the variance due to partial client participation in federated learning,” in Proceedings of the Thirty-Eighth Conference on Uncer- tainty in Artificial Intelligence . PMLR, 2022, pp. 906–916. [Online]. Available: https://proceedings.mlr.press/v180/jhunjhunwala22a.html

    work page 2022

  8. [8]

    Fedvarp: Variance reduced partial client participation in federated learning,

    P. Li, C. Han, P. Zhou, Y . Zhang, and S. Ma, “Fedvarp: Variance reduced partial client participation in federated learning,” IEEE Transactions on Neural Networks and Learning Systems , pp. 1–15, 2022

    work page 2022

Show all 41 references
  1. [1]

    Recent advances on federated learning: A systematic survey,

    B. Liu, N. Lv, Y . Guo, and Y . Li, “Recent advances on federated learning: A systematic survey,” Neurocomputing, vol. 597, p. 128019, 2024

    2024

  2. [2]

    A survey on federated learning: Challenges and applications,

    J. Wen, Z. Zhang, Y . Lan, Z. Cui, J. Cai, and W. Zhang, “A survey on federated learning: Challenges and applications,” International Journal of Machine Learning and Cybernetics , vol. 14, pp. 513–535, 2023

    2023

  3. [3]

    A survey on efficient federated learning methods for foundation model training,

    H. Woisetschla ¨ager, A. Erben, S. Wang, R. Mayer, and H.-A. Jacobsen, “A survey on efficient federated learning methods for foundation model training,” in Proceedings of the International Joint Conference on Artificial Intelligence (IJCAI), 2024, accepted for publication. Ava...

    2024 doi

  4. [5]

    A survey on federated learning systems: Vision, hype and reality for data privacy and protection,

    Q. Li, Z. Wen, Z. Wu, S. Hu, N. Wang, Y . Li, X. Liu, and B. He, “A survey on federated learning systems: Vision, hype and reality for data privacy and protection,” arXiv preprint arXiv:1907.09693v7 , 2021

    1907 arXiv

  5. [6]

    Communication-efficient learning of deep networks from decentralized data,

    B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arcas, “Communication-efficient learning of deep networks from decentralized data,” in Proceedings of the 20th International Conference on Artificial Intelligence and Statistics (AISTATS), vol. 54. PMLR, 2017, pp. 1273– 1282

    2017

  6. [7]

    Safari: Server- aided federated learning on incomplete and heterogeneous data,

    M. Diao, H. Zhang, Y . Li, X. Liu, and H. Huang, “Safari: Server- aided federated learning on incomplete and heterogeneous data,” IEEE Transactions on Mobile Computing, vol. 22, no. 6, pp. 3083–3095, 2023

    2023

  7. [10]

    Fedcm: Federated learning via contextualized client momentum,

    J. Zhang, X. Meng, Q. Wang, and J. Xu, “Fedcm: Federated learning via contextualized client momentum,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 36, no. 3, 2022, pp. 2997– 3006

    2022

  8. [11]

    Mifa: Memory- augmented federated learning on incomplete client participation,

    J. Chen, Y . Xu, Y . Liu, Q. Liu, X. Li, and B. He, “Mifa: Memory- augmented federated learning on incomplete client participation,” IEEE Transactions on Parallel and Distributed Systems , vol. 34, no. 3, pp. 882–895, 2023

    2023

  9. [12]

    Scaffold: Stochastic controlled averaging for federated learning,

    S. P. Karimireddy, S. Kale, M. Mohri, S. J. Reddi, S. Stich, and A. T. Suresh, “Scaffold: Stochastic controlled averaging for federated learning,” in Proceedings of the 37th International Conference on Machine Learning (ICML) , vol. 119, 2020, pp. 5132–5143

    2020

  10. [14]

    On the convergence of federated averaging with cyclic client participation,

    Y . J. Cho, P. Sharma, G. Joshi, Z. Xu, S. Kale, and T. Zhang, “On the convergence of federated averaging with cyclic client participation,” JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 15 in International Conference on Machine Learning (ICML) , vol. 202. PMLR, 20...

    2021

  11. [16]

    Gradma: Memory-augmented federated learning with server- side quadratic programming,

    D. Duan, Y . Wang, H. Yin, H. Wang, J. Zhou, C. Hu, and J. Huai, “Gradma: Memory-augmented federated learning with server- side quadratic programming,” IEEE Transactions on Neural Networks and Learning Systems , vol. 34, no. 1, pp. 75–89, 2023

    2023

  12. [17]

    Federated optimization in heterogeneous networks,

    T. Li, A. K. Sahu, M. Zaheer, M. Sanjabi, A. Talwalkar, and V . Smith, “Federated optimization in heterogeneous networks,” in Proceedings of Machine Learning and Systems (MLSys) , vol. 2, 2020, pp. 429–450

    2020

  13. [21]

    Understanding server-assisted federated learning in the presence of incomplete client participation,

    H. Yang, P. Qiu, P. Khanduri, M. Fang, and J. Liu, “Understanding server-assisted federated learning in the presence of incomplete client participation,” in Proceedings of the 41st International Conference on Machine Learning . PMLR, 2024, pp. 56 298–56 318. [Online]. Availabl...

    2024

  14. [22]

    Semi-variance reduction for fair federated learning,

    S. Malekmohammadi, “Semi-variance reduction for fair federated learning,” arXiv preprint arXiv:2406.16193 , 2024. [Online]. Available: https://arxiv.org/abs/2406.16193

    2024 arXiv

  15. [23]

    Federated learning for non-iid data via client variance reduction and adaptive server update,

    H. Nguyen, L. Phan, H. Warrier, and Y . Gupta, “Federated learning for non-iid data via client variance reduction and adaptive server update,” arXiv preprint arXiv:2207.08391 , 2022. [Online]. Available: https://arxiv.org/abs/2207.08391

    2022 arXiv

  16. [24]

    Non-convex optimization in federated learning via variance reduction and adaptive learning,

    D. Thakur, A. Guzzo, G. Fortino, and S. K. Das, “Non-convex optimization in federated learning via variance reduction and adaptive learning,” arXiv preprint arXiv:2412.11660 , 2024. [Online]. Available: https://arxiv.org/abs/2412.11660

    2024 arXiv

  17. [25]

    Beyond admm: A unified client-variance-reduced adaptive federated learning framework,

    S. Wang, Y . Xu, Z. Wang, T.-H. Chang, T. Q. Quek, and D. Sun, “Beyond admm: A unified client-variance-reduced adaptive federated learning framework,” arXiv preprint arXiv:2212.01519 , 2022. [Online]. Available: https://arxiv.org/abs/2212.01519

    2022 arXiv

  18. [26]

    Emulating full client participation: A long-term client selection strategy for federated learning,

    Q. Li, J. Miao, P. Zhao, L. Zhou, S. Ji, B. Zhou, and F. Liu, “Emulating full client participation: A long-term client selection strategy for federated learning,” arXiv preprint arXiv:2405.13584 , 2024. [Online]. Available: https://arxiv.org/abs/2405.13584

    2024 arXiv

  19. [27]

    Fedcada: Adaptive client-side optimization for accelerated and stable federated learning,

    L. Zhou, Y . He, K. Zhai, X. Liu, S. Liu, X. Ma, G. Ye, Y .-G. Jiang, and H. Chai, “Fedcada: Adaptive client-side optimization for accelerated and stable federated learning,” arXiv preprint arXiv:2405.11811 , 2024. [Online]. Available: https://arxiv.org/abs/2405.11811

    2024 arXiv

  20. [28]

    Efficient client selection in federated learning,

    W. Marfo, D. K. Tosh, and S. V . Moore, “Efficient client selection in federated learning,” arXiv preprint arXiv:2502.00036 , 2025. [Online]. Available: https://arxiv.org/abs/2502.00036

    2025 arXiv

  21. [29]

    Delta: Diverse client sampling for faster federated learning,

    L. Wang, Y . Guo, T. Lin, and X. Tang, “Delta: Diverse client sampling for faster federated learning,” arXiv preprint arXiv:2205.13925 , 2022. [Online]. Available: https://arxiv.org/abs/2205.13925

    2022 arXiv

  22. [30]

    Variance-reduced heterogeneous federated learning via stratified client selection,

    G. Shen, D. Gao, L. Yang, F. Zhou, D. Song, W. Lou, and S. Pan, “Variance-reduced heterogeneous federated learning via stratified client selection,” arXiv preprint arXiv:2201.05762 , 2022. [Online]. Available: https://arxiv.org/abs/2201.05762

    2022 arXiv

  23. [31]

    Heterogeneity-guided client sampling: Towards fast and efficient non-iid federated learning,

    H. Chen and H. Vikalo, “Heterogeneity-guided client sampling: Towards fast and efficient non-iid federated learning,” arXiv preprint arXiv:2310.00198, 2023. [Online]. Available: https://arxiv.org/abs/2310. 00198

    2023 arXiv

  24. [33]

    Clustered sampling: Low-variance and improved representativity for clients selection in federated learning,

    Y . Fraboni, R. Vidal, and E. M. El Mhamdi, “Clustered sampling: Low-variance and improved representativity for clients selection in federated learning,” arXiv preprint arXiv:2103.00174 , 2021. [Online]. Available: https://arxiv.org/abs/2103.00174

    2021 arXiv

  25. [34]

    Feddc: Federated learning with non-iid data via local drift decoupling and correction,

    L. Gao, H. Fu, L. Li, Y . Chen, M. Xu, and C.-Z. Xu, “Feddc: Federated learning with non-iid data via local drift decoupling and correction,” arXiv preprint arXiv:2203.11751, 2022. [Online]. Available: https://arxiv.org/abs/2203.11751

    2022 arXiv

  26. [35]

    Combating client dropout in federated learning via friend model substitution,

    H. Wang and J. Xu, “Combating client dropout in federated learning via friend model substitution,” arXiv preprint arXiv:2205.13222 , 2022. [Online]. Available: https://arxiv.org/abs/2205.13222

    2022 arXiv

  27. [36]

    A novel client sampling scheme for unbalanced data distribution under federated learning,

    X. Li, Y . Zhang, W. Wang, Y . Liu, and Y . Wang, “A novel client sampling scheme for unbalanced data distribution under federated learning,” Complex & Intelligent Systems , vol. 9, pp. 2081–2099, 2023. [Online]. Available: https://link.springer.com/article/ 10.1007/s40747-022-00895-3

    2023 doi

  28. [37]

    Fedrtid: An efficient shuffle federated learning via random participation and adaptive time con- straint,

    W. Zhang, X. Liu, Y . Wang, and M. Li, “Fedrtid: An efficient shuffle federated learning via random participation and adaptive time con- straint,” Cybersecurity, vol. 7, no. 1, pp. 1–15, 2024. [Online]. Available: https://link.springer.com/article/10.1186/s42400-024-00293-x

    2024 doi

  29. [38]

    Federated learning with regularized client participation,

    G. Malinovsky, S. Horv ´ath, K. Burlachenko, and P. Richt ´arik, “Federated learning with regularized client participation,” arXiv preprint arXiv:2302.03662, 2023. [Online]. Available: https://arxiv.org/ abs/2302.03662

    2023 arXiv

  30. [40]

    Adaptive client sampling in federated learning via online learning with bandit feedback,

    B. Zhao, L. Wang, Z. Liu, Z. Zhang, J. Zhou, C. Chen, and M. Kolar, “Adaptive client sampling in federated learning via online learning with bandit feedback,” Journal of Machine Learning Research, vol. 26, no. 8, pp. 1–67, 2025. [Online]. Available: https://www.jmlr.org/papers...

    2025

  31. [41]

    Fedccea: A practical approach of client contribution evaluation for federated learning,

    S. K. Shyn, D. Kim, and K. Kim, “Fedccea: A practical approach of client contribution evaluation for federated learning,” arXiv preprint arXiv:2106.02310, 2021. [Online]. Available: https://arxiv.org/abs/2106. 02310

    2021 arXiv

  32. [2023]

    Available: https://arxiv.org/abs/2312.17430

    [Online]. Available: https://arxiv.org/abs/2312.17430

    arXiv

  33. [2025]

    Available: https://arxiv.org/abs/2503.11151

    [Online]. Available: https://arxiv.org/abs/2503.11151

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.