Pith. sign in

REVIEW 3 major objections 4 minor 38 references

Cluster-Aware Over-the-Air Federated Learning with Energy-Harvesting Devices: From Global Training to Model Personalization

T0 review · 3 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper shows that user clusters turn over-the-air federated learning with energy-harvesting devices into either fairer global training or personalized per-cluster training.

desk verdict A useful unified cluster-aware framework for OTA FL with energy harvesting; the main soft spot is an unstated identifiability requirement in the LSE-based scheduling that fails for low-energy users. read the letter →

arxiv 2608.01426 v1 pith:DFE4KW7E submitted 2026-08-02 cs.LG cs.DC

Furkan Bagci , Busra Tegin , Mohammad Kazemi , Tolga M. Duman This is my paper · ORCID
classification cs.LGcs.DC
keywords federatedlearningover-the-aircomputationenergyharvestingdiverseclientschedulingclusteredpersonalizationdataheterogeneityblindtransmitters
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

The paper aims to show that the natural cluster structure among users with similar data can serve as one shared mechanism for two different goals in over-the-air federated learning with energy-harvesting devices: making a single global model fairer, and producing personalized models per user group. For the global-model goal, cluster information drives scheduling that picks a representative subset of active users, reducing the gap between the scheduled gradient and the full-participation gradient. For the personalization goal, the same clusters define separate learning targets, and the parameter server recovers all clusters' updates simultaneously from one superimposed wireless signal using MMSE or cluster-wise weighted combining. The convergence analysis decomposes the error into contributions from OTA channel noise, data heterogeneity, and gradient approximation error, which motivates the scheduling and combining designs. Numerical experiments on MNIST, FMNIST, and CIFAR-10 indicate that the framework improves fairness or personalization while lowering communication overhead.

What carries the argument

The carrying mechanism is the user-cluster structure, a grouping of devices by similarity of their label distributions or update directions. In Mode I it enters through cluster-aware scheduling: entropy-based selection when label distributions are known, and least-squares estimation of user representations from the noisy aggregated signal when they are not. In Mode II it is encoded in the cluster-assignment matrix $A$ and its projector $P = A(A^\top A)^{-1}A^\top$; the server's MMSE or cluster-wise weighted combining recovers each cluster's average update from the same received signal. The convergence analysis centers on the gradient-approximation error $\epsilon$, the gap between the full-p

What would settle it

Run the LSE-based scheduler with $M=40$ users, one class per user, an estimation window of $T=100$ iterations, and energy arrival probabilities $p^m_e=0.25$ for most users but near zero for a few. If those low-energy users never appear in the participation matrix $A$, the least-squares estimate $\hat{\Theta}_{\text{rep}}$ is not identifiable and the estimated cosine-similarity clusters will not reflect their data; fairness should then degrade to or below the no-scheduling baseline. Checking the rank of $A$ in that run settles whether the paper's identifiability assumption holds.

Watch

Extended reading notes

Core claim

The paper's central claim is that user-cluster structure is one abstraction that resolves a tension in over-the-air federated learning: data heterogeneity can be treated either as bias to be removed or as structure to be exploited, and the same clusters serve both goals. For a fair global model, the server either uses known label distributions to pick the active-user subset with highest label entropy, or, when distributions are unknown, estimates each user's update representation from $T$ aggregated over-the-air signals by least squares ($\hat{\Theta}_{PS}=A\,\Theta_{\text{rep}}+N^*$) and clusters users by cosine similarity; scheduling then samples across clusters. For personalization, the s

Load-bearing premise

The weakest load-bearing premise is that the $T\times M$ participation matrix $A$ in Eq. (28) has full column rank, so every user's update representation is recoverable from aggregated over-the-air signals during the estimation window; users with very low energy-arrival probabilities may rarely or never participate, making their representations unidentifiable and the diversity-aware scheduler blind to their data.

Editorial extensions

If this is right

  • One parameter server can train multiple cluster-specific models simultaneously over one multiple-access channel, avoiding separate transmissions or multiple servers for personalized FL.
  • In energy-harvesting settings, cluster-aware scheduling makes partial participation more representative, so fairness improves most when data heterogeneity is strongest, as shown by the CIFAR-10 experiments with strong Dirichlet skew.
  • The error-bound decomposition yields concrete design rules: increasing the number of server antennas $K$ or the minimum scheduled-set size $S_{\min}$ suppresses the OTA noise floor, while the heterogeneity floor $2(\tau-1)\Gamma/(\mu\tau)$ cannot be removed by stepsize tuning.
  • Even with no CSI at the transmitters and only cluster-level CSI at the server, the CWC and partial-MMSE combiners still beat a single-global-model baseline, so personalization is feasible under realistic channel knowledge.
  • The LSE representation phase uses only participation patterns and aggregated noisy updates, so scheduling can be made diversity-aware without requiring users to disclose their data distributions.

Reading between the lines

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

  • Beyond the paper: the full-column-rank identifiability requirement is a hidden constraint; an adaptive estimation window that lengthens until every user has participated enough times would be a natural extension the paper does not analyze.
  • Beyond the paper: the same cluster abstraction could carry into hierarchical federated learning or multi-task learning, where clusters become task groups; the paper mentions such directions only as future work.
  • Beyond the paper: a testable prediction is that the gap between LSE-based scheduling and entropy-based scheduling shrinks as the estimation window $T$ grows and as energy arrival probabilities become more balanced across users.
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 / 4 minor

Summary. The paper proposes a unified cluster-aware over-the-air federated learning (OTA-FL) framework for energy-harvesting (EH) mobile devices. In Mode I, cluster structure guides diversity-aware scheduling to train a fair global model; in Mode II, the same cluster structure defines cluster-level objectives and OTA recovery targets for personalized clustered FL. The authors provide convergence analysis for the global training mode, propose entropy-based scheduling for known data distributions and least-squares-based user-representation estimation for unknown distributions, and develop MMSE/CWC combining schemes for personalized clustered OTA-FL under full or partial CSI at the parameter server. Numerical experiments on MNIST, FMNIST, and CIFAR-10 support the claims of improved fairness or personalization with reduced communication overhead.

Significance. If the proposed framework and its analysis hold, the paper offers a useful unified treatment of two usually separate FL objectives—fair global training and personalized clustered training—under practical OTA and EH constraints. The convergence analysis is honest: it explicitly states that the finite-K decaying-stepsize bound does not establish convergence and that the constant-stepsize floor decomposes into OTA noise, data heterogeneity, and participation-bias terms. The manuscript also benefits from extensive simulations across multiple datasets, a clear treatment of different CSI levels, and the absence of fitted numerical constants in the theory. These are genuine strengths. However, two load-bearing technical points need attention: the LSE-based representation estimator implicitly requires a full-column-rank participation matrix, and the convergence bound depends on an unspecified constant c in Lemma 4. Both issues are addressable, but they currently limit the strength of the central claims.

major comments (3)
  1. [Section III.C, Eq. (28)] The LSE estimator for Theta_rep is defined via A in {0,1}^{T x M}. Solving the least-squares problem requires A^T A invertible, i.e., every user must participate at least once during the estimation window and the participation patterns must span R^M. This identifiability condition is never stated. Under the Bernoulli EH model, user m is absent from the entire T-round estimation window with probability (1-p_m^e)^T. For p_m^e=0.01 and T=100 this is about 0.37; for p_m^e=0.05 it is about 0.006, which is not negligible with M=100. In that regime the corresponding user representation is unidentifiable, cosine-similarity clustering is blind to that user, and the diversity-aware scheduler may systematically ignore exactly the users whose data is underrepresented. The simulations use p_m^e(t)=0.25 with T=100 or 200 (expected 25-50 participations per user), which masks this regime. Please add the
  2. [Appendix A, Lemma 4, Eq. (55)] Lemma 4 bounds the cross term as (eta^2 tau(tau-1)LG + eta tau epsilon)c, where c is only described as 'some constant ... related to Gamma, G and mu'. No expression or computable bound for c is provided. Since c appears linearly in B_1 = 2(tau-1)Gamma + tau epsilon-bar c, the asymptotic residual neighborhood in Corollary 1 and the interpretation in Remark 1 depend on an unspecified quantity. As stated, the result establishes the form of the bound but not a concrete quantitative neighborhood. Please make c explicit or replace it with a computable bound, or clearly state that the asymptotic neighborhood is only qualitative.
  3. [Section III.C, Eqs. (24)-(28)] The LSE-based representation estimation is central to the unknown-distribution scheduling mode, but no consistency or error guarantee is provided for Theta_hat_rep. The only evidence is the single MNIST visualization in Fig. 2. Even when A has full column rank, the quality of the estimate depends on T, on the magnitude of Theta_diff,j relative to Theta_rep, and on the statistical properties of N*. The paper should state the assumptions under which the estimated representations are reliable and ideally give an error bound or sample-complexity result. Without this, the claim that LSE-based scheduling 'minimizes' or even 'reduces' the gradient approximation error epsilon in (16) is not supported by the theory.
minor comments (4)
  1. [Eq. (16)] The definition of epsilon appears with a norm whose squared value is then used in Eq. (15). Please clarify whether epsilon is the norm or the squared norm, and use consistent notation.
  2. [Section IV.B.1] In the partial-CSI MMSE subsection, X_c,n in C^H is described as 'the matrix of transmitted cluster updates' but it is a vector per symbol. Please correct the wording.
  3. [Appendix A] Lemmas 3 and 4 are justified by reference to [24, Lemma 1] rather than proved. Given that Lemma 4 contains the unspecified constant c, a full proof or at least a more detailed derivation would improve self-containedness.
  4. [Section III.C] The sentence about clustering 'with the expected number of users per cluster determined by each cluster's energy distribution' is vague. Please specify how the cluster sizes are computed from energy arrival probabilities.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the convergence bound's epsilon term is a standard decomposition, and the LSE/MMSE/CWC components are not fitted predictions or self-citation-forced constructions.

full rationale

The derivation chain is self-contained in the relevant sense. Theorem 1's bound contains epsilon (Eq. 16) defined as the gap between the full-participation gradient and the scheduled-subset gradient. This is a bias-variance-style decomposition, not a fitted parameter renamed as a prediction: epsilon is an input to the theorem that depends on whatever scheduler is used, and the theorem holds for any scheduler. The proposed entropy-based and LSE-based schedulers are motivated by reducing this gap, but their benefit is verified empirically against a no-scheduling baseline rather than asserted by construction. The LSE user representation in Section III.C (Eqs. 26-28) is an unsupervised estimator of update directions; it does not fit any target accuracy, and cosine-similarity clustering is validated against true update similarities in Fig. 2. The personalized CFL section derives standard MMSE and CWC combiners; no constant is fitted to make simulations match theory. The main potential concern is an identifiability gap: solving Eq. (28) for Theta_rep implicitly requires A to have full column rank, which is not stated and can fail for low-energy users who rarely participate. This is a correctness risk in the low-energy regime, not a circularity. Self-citations ([11], [13]) are to published peer-reviewed results whose assumptions do not include the target result, so under the reviewing rules they count as genuine evidence and do not raise the circularity score.

Assumptions & free parameters 0 free parameters · 6 assumptions · 1 invented entities

The central claims rest on standard system-model assumptions (Bernoulli energy arrivals, i.i.d. Gaussian channels, bounded gradients, strong convexity for the theory) rather than on fitted parameters. No constants are tuned to data. The only intrinsic construct is the latent user representation used for clustering, which is estimated from the system and validated only in the paper's own experiments.

assumptions (6)
  • domain assumption Local loss functions F_m are L-smooth and mu-strongly convex for the convergence analysis.
    Invoked in Section III.A.1 before Theorem 1 to obtain the contraction factor A(i); the experiments use CNNs, so the theorem's assumptions do not hold in the evaluation.
  • domain assumption Stochastic gradients have bounded second moment G^2.
    Standard bounded-gradient condition used in Theorem 1 and Corollaries.
  • domain assumption Bernoulli unit-energy arrivals, unit-sized battery, and 1 energy unit consumed per local computation plus transmission; energy arrivals are reported to the PS.
    System model in Section II.A; the scheduling and convergence results depend on this abstraction of energy harvesting.
  • domain assumption Channels are i.i.d. complex Gaussian, no CSIT, PS has K antennas; interference between users diminishes with K as in [11].
    Used in the OTA signal model (Eqs. (5), (10)) and in Lemma 1; the CWC/MMSE recovery relies on this statistical assumption.
  • domain assumption The participation matrix A in the LSE formulation (Eq. (28)) has full column rank over the T-iteration estimation window.
    Required for unique LSE estimation of Theta_rep; the paper does not state or analyze conditions on T and energy arrival probabilities for identifiability.
  • domain assumption For partial CSI, cluster updates are captured by the cluster-sum projector P, and intra-cluster deviations behave as structured noise with known covariance for MMSE.
    Section IV.B.2 Eq. (40)-(41) models H_n(I-P)X_n as an additional noise term; cluster alignment is assumed good enough for recovery.
invented entities (1)
  • User representation Theta_rep (and its LSE estimate)
    purpose: An M x 2N latent matrix representing the typical update direction of each user, inferred from aggregated OTA signals to enable clustering and diversity-aware scheduling without accessing user data.
    Introduced in Section III.C as a statistical latent variable. It is validated only through the paper's own cosine-similarity visualization and downstream simulation gains; it makes no external falsifiable prediction independent of this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Cluster-Aware Over-the-Air Federated Learning with Energy-Harvesting Devices: From Global Training to Model Personalization." pith.science (2026). https://pith.science/paper/DFE4KW7E

@misc{pith2026260801426,
  author       = {Pith},
  title        = {Pith review of: Cluster-Aware Over-the-Air Federated Learning with Energy-Harvesting Devices: From Global Training to Model Personalization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DFE4KW7E}},
  note         = {Machine review of arXiv:2608.01426}
}
read the original abstract

Federated learning (FL) enables distributed optimization and learning across decentralized edge devices while preserving data privacy, but its performance is fundamentally constrained by heterogeneous data distributions, limited communication resources, and energy availability. In practical wireless networks, mobile devices (MDs) often exhibit diverse data and learning objectives, naturally forming clusters of users with jointly trainable models. When devices rely on energy harvesting (EH), stochastic energy arrivals further complicate participation and scheduling under communication constraints. In this work, we study over-the-air (OTA) FL with EH MDs under heterogeneous data distributions, and investigate two closely related learning objectives within a unified framework: one aiming for a more representative global model by reducing data bias, and the other learning more personalized cluster-specific models by exploiting this bias. In the global training mode, cluster information guides energy- and diversity-aware scheduling, ensuring that the scheduled active users provide a more representative aggregate update. In the personalization mode, the same cluster structure defines cluster-level learning objectives and OTA recovery targets, enabling the parameter server to train multiple cluster-specific models through simultaneous transmissions over the wireless multiple-access channel. Numerical results demonstrate that the proposed unified framework improves fairness or personalization, depending on the operating mode, while reducing communication overhead.

Figures

Figures reproduced from arXiv: 2608.01426 by the authors.

Figure 1
Figure 1. Unified cluster-aware OTA-FL framework for EH devices. Heterogeneous EH users communicate with a single PS over an OTA wireless multiple [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. The visualization of cosine similarity on the user clusters. [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Test accuracy of entropy-based scheduling for CIFAR-10 with [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Test accuracy for MNIST with M = 40, |Bm| = 1250, and pme (t) = 0.25. τ = 5, and a mini-batch size of |ξm(t)| = 100 for MNIST and FMNIST, and τ = 3 and |ξm(t)| = 128 for CIFAR-10. To simulate highly non-i.i.d. data, we consider two different distribution scenarios. In …
Figure 5
Figure 5. Figure 5: Test accuracy for MNIST and FMNIST. and estimation phases of T = 100 and T = 200 iterations. In both cases, entropy-based scheduling yields higher and more stable accuracy. For cases with unknown data distributions, the PS estimates local user representations after T i…
Figure 6
Figure 6. Figure 6: Test accuracy for CFL with M = 20, H = 3, and K = 20. 0 20 40 60 80 100 Number of Global Iterations, t 0.6 0.7 0.8 0.9 1.0 Test accuracy Full MMSE - K:20 Partial MMSE - K:20 Partial CWC - K:20 Full MMSE - K:5 Partial MMSE - K:5 Partial CWC - K:5 85 90 95 0.95 0.96 0.97…
Figure 7
Figure 7. Figure 7: Test accuracy for CFL with M = 20, H = 3, and K ∈ {5, 20}. performance under limited-antenna conditions. Nevertheless, they achieve a relatively acceptable accuracy, typically around 5–10 percentage points lower than that of the full CSI MMSE setup with K = 20. Next, w…
Figure 8
Figure 8. Figure 8: shows that the proposed combining methods success￾fully demonstrate the feasibility of delivering distinct person￾alized models to different clusters simultaneously over the air. As expected, the full CSI MMSE method outperforms both the partial MMSE and partial CWC ap…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 34 canonical work pages

  1. [11]

    Blind federated edge learning,

    M. M. Amiri, T. M. Duman, D. G ¨und¨uz, S. R. Kulkarni, and H. V . Poor, “Blind federated edge learning,”IEEE Trans. Wireless Commun., vol. 20, no. 8, pp. 5129–5143, Aug. 2021

    work page 2021

  2. [24]

    Diverse client selection for federated learning via submodular maxi- mization,

    R. Balakrishnan, T. Li, T. Zhou, N. Himayat, V . Smith, and J. Bilmes, “Diverse client selection for federated learning via submodular maxi- mization,” inProc. Int. Conf. Learn. Represent. (ICLR), Apr. 2022, pp. 1–18

    work page 2022

  3. [1]

    Update estimation and scheduling for over-the-air federated learning with energy harvesting devices,

    F. Bagci, B. Tegin, M. Kazemi, and T. M. Duman, “Update estimation and scheduling for over-the-air federated learning with energy harvesting devices,” inProc. IEEE Int. Conf. Commun. Workshops (ICC Work- shops), Montreal, Canada, Jun. 2025, pp. 1435–1440

    work page 2025

  4. [2]

    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,” inProc. Int. Conf. Artif. Intell. Stat. (AISTATS), vol. 54. PMLR, Apr 2017, pp. 1273–1282

    work page 2017

  5. [3]

    Federated learning: Strategies for improving communication efficiency,

    J. Kone ˇcn´y, H. B. McMahan, F. X. Yu, P. Richt ´arik, A. T. Suresh, and D. Bacon, “Federated learning: Strategies for improving communication efficiency,”arXiv preprint arXiv:1610.05492, 2017

    arXiv 2017

  6. [4]

    Federated learning for Internet of Things: A com- prehensive survey,

    D. C. Nguyen, M. Ding, P. N. Pathirana, A. Seneviratne, J. Li, and H. Vincent Poor, “Federated learning for Internet of Things: A com- prehensive survey,”IEEE Commun. Surveys Tuts., vol. 23, no. 3, pp. 1622–1658, Third Quarter 2021

    work page 2021

  7. [5]

    Federated learning via over- the-air computation,

    K. Yang, T. Jiang, Y . Shi, and Z. Ding, “Federated learning via over- the-air computation,”IEEE Trans. Wireless Commun., vol. 19, no. 3, pp. 2022–2035, Mar. 2020

    2022

  8. [6]

    Machine learning at the wireless edge: Distributed stochastic gradient descent over-the-air,

    M. Mohammadi Amiri and D. G ¨und¨uz, “Machine learning at the wireless edge: Distributed stochastic gradient descent over-the-air,”IEEE Trans. Signal Process., vol. 68, pp. 2155–2169, Mar. 2020

    work page 2020

Show all 38 references
  1. [7]

    Dynamic scheduling for over-the-air federated edge learning with energy constraints,

    Y . Sun, S. Zhou, Z. Niu, and D. G ¨und¨uz, “Dynamic scheduling for over-the-air federated edge learning with energy constraints,”IEEE J. Sel. Areas Commun., vol. 40, no. 1, pp. 227–242, Nov. 2022

    2022

  2. [8]

    Learned digital codes for over-the-air federated learning,

    A. Tarizzo, M. Kazemi, and D. G ¨und¨uz, “Learned digital codes for over-the-air federated learning,” 2025. [Online]. Available: https: //arxiv.org/abs/2509.16577

    2025 arXiv

  3. [9]

    Device scheduling in over-the-air federated learning via matching pursuit,

    A. Bereyhi, A. Vagollari, S. Asaad, R. R. M ¨uller, W. Gerstacker, and H. V . Poor, “Device scheduling in over-the-air federated learning via matching pursuit,”IEEE Trans. Signal Process., vol. 71, pp. 2188–2203, Jun. 2023

    2023

  4. [10]

    Beamforming vector design and device selection in over-the-air federated learning,

    M. Kim, A. L. Swindlehurst, and D. Park, “Beamforming vector design and device selection in over-the-air federated learning,”IEEE Trans. Wireless Commun., vol. 22, no. 11, pp. 7464–7477, Mar. 2023

    2023

  5. [12]

    Blind federated learning at the wireless edge with low-resolution ADC and DAC,

    B. Tegin and T. M. Duman, “Blind federated learning at the wireless edge with low-resolution ADC and DAC,”IEEE Trans. Wireless Com- mun., vol. 20, no. 12, pp. 7786–7798, Jun. 2021

    2021

  6. [13]

    Over-the- air federated edge learning with hierarchical clustering,

    O. Ayg ¨un, M. Kazemi, D. G ¨und¨uz, and T. M. Duman, “Over-the- air federated edge learning with hierarchical clustering,”IEEE Trans. Wireless Commun., vol. 23, no. 12, pp. 17 856–17 871, Dec. 2024

    2024

  7. [14]

    Energy harvesting wireless communications: A review of recent advances,

    S. Ulukus, A. Yener, E. Erkip, O. Simeone, M. Zorzi, P. Grover, and K. Huang, “Energy harvesting wireless communications: A review of recent advances,”IEEE J. Sel. Areas Commun., vol. 33, no. 3, pp. 360– 381, Mar. 2015

    2015

  8. [15]

    Federated learning with energy harvesting devices,

    L. Zeng, D. Wen, G. Zhu, C. You, Q. Chen, and Y . Shi, “Federated learning with energy harvesting devices,”IEEE Trans. Green Commun. Netw., vol. 8, no. 1, pp. 190–204, Aug. 2024

    2024

  9. [16]

    Federated learning over energy harvesting wireless networks,

    R. Hamdi, M. Chen, A. B. Said, M. Qaraqe, and H. V . Poor, “Federated learning over energy harvesting wireless networks,”IEEE Internet Things J., vol. 9, no. 1, pp. 92–103, Jun. 2022

    2022

  10. [17]

    Joint client selection and receive beamforming for over-the-air federated learning with energy harvesting,

    C. Chen, Y .-H. Chiang, H. Lin, J. C. Lui, and Y . Ji, “Joint client selection and receive beamforming for over-the-air federated learning with energy harvesting,”IEEE Open J. Commun. Soc., vol. 4, pp. 1127–1140, May. 2023

    2023

  11. [18]

    Over-the-air federated learning with energy harvesting devices,

    O. Ayg ¨un, M. Kazemi, D. G ¨und¨uz, and T. M. Duman, “Over-the-air federated learning with energy harvesting devices,” inProc. IEEE Global Commun. Conf. (GLOBECOM), Rio de Janeiro, Brazil, Dec 2022, pp. 1942–1947

    2022

  12. [19]

    Online optimization for over-the-air federated learning with energy harvesting,

    Q. An, Y . Zhou, Z. Wang, H. Shan, Y . Shi, and M. Bennis, “Online optimization for over-the-air federated learning with energy harvesting,” IEEE Trans. Wireless Commun., vol. 23, no. 7, pp. 7291–7306, July 2024

    2024

  13. [20]

    Federated learning with energy harvesting devices: An MDP framework,

    K. Zhang and X. Cao, “Federated learning with energy harvesting devices: An MDP framework,”arXiv preprint arXiv:2405.10513, 2024

    2024 arXiv

  14. [21]

    Federated learning with non-IID data,

    Y . Zhao, M. Li, L. Lai, N. Suda, D. Civin, and V . Chandra, “Federated learning with non-IID data,”arXiv preprint arXiv:1806.00582, 2018. 17

    2018 arXiv

  15. [22]

    On the convergence of FedAvg on non-IID data,

    X. Li, K. Huang, W. Yang, S. Wang, and Z. Zhang, “On the convergence of FedAvg on non-IID data,” inProc. Int. Conf. Learn. Represent. (ICLR), Apr. 2020, pp. 1–26

    2020

  16. [23]

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

    Y . Fraboni, R. Vidal, L. Kameni, and M. Lorenzi, “Clustered sampling: Low-variance and improved representativity for clients selection in federated learning,” inProc. Int. Conf. Mach. Learn. (ICML). PMLR, Jul. 2021, pp. 3407–3416

    2021

  17. [25]

    Clustered federated learning: Model-agnostic distributed multitask optimization under privacy con- straints,

    F. Sattler, K.-R. M ¨uller, and W. Samek, “Clustered federated learning: Model-agnostic distributed multitask optimization under privacy con- straints,”IEEE Trans. Neural Netw. Learn. Syst., vol. 32, no. 8, pp. 3710–3722, Aug. 2021

    2021

  18. [26]

    An efficient framework for clustered federated learning,

    A. Ghosh, J. Chung, D. Yin, and K. Ramchandran, “An efficient framework for clustered federated learning,”IEEE Trans. Inf. Theory, vol. 68, no. 12, pp. 8076–8091, Dec. 2022

    2022

  19. [27]

    Over-the-air clustered federated learning,

    H. U. Sami and B. G ¨uler, “Over-the-air clustered federated learning,” IEEE Trans. Wireless Commun., vol. 23, no. 7, pp. 7877–7893, July 2024

    2024

  20. [28]

    Rethinking clustered federated learning in NOMA enhanced wireless networks,

    Y . Lin, K. Wang, and Z. Ding, “Rethinking clustered federated learning in NOMA enhanced wireless networks,”IEEE Trans. Wireless Commun., vol. 23, no. 11, pp. 16 875–16 890, Nov. 2024

    2024

  21. [29]

    Personalized federated learning over the air,

    Z. Li, Z. Chen, T. Q. S. Quek, and H. H. Yang, “Personalized federated learning over the air,”IEEE Trans. Wireless Commun., vol. 24, no. 11, pp. 9509–9523, Nov. 2025

    2025

  22. [30]

    Over-the- air multitask federated learning via cost-effective hybrid beamforming,

    J. Yao, W. Xu, G. Zhu, Z. Yang, K. Huang, and D. Niyato, “Over-the- air multitask federated learning via cost-effective hybrid beamforming,” IEEE Trans. Signal Process., vol. 73, pp. 3977–3993, 2025

    2025

  23. [31]

    Energy harvesting sensor nodes: Survey and implications,

    S. Sudevalayam and P. Kulkarni, “Energy harvesting sensor nodes: Survey and implications,”IEEE Commun. Surveys Tuts., vol. 13, no. 3, pp. 443–461, Third Quarter 2011

    2011

  24. [32]

    Federated learning with over-the-air ag- gregation over time-varying channels,

    B. Tegin and T. M. Duman, “Federated learning with over-the-air ag- gregation over time-varying channels,”IEEE Trans. Wireless Commun., vol. 22, no. 8, pp. 5671–5684, Aug. 2023

    2023

  25. [33]

    Optimizing federated learning by entropy-based client selection,

    A. Lutz, G. Steidl, K. M ¨uller, and W. Samek, “Optimizing federated learning by entropy-based client selection,”arXiv preprint 2411.01240, 2024

    2024

  26. [34]

    Optimizing federated learning on non-IID data with reinforcement learning,

    H. Wang, Z. Kaplan, D. Niu, and B. Li, “Optimizing federated learning on non-IID data with reinforcement learning,” inProc. IEEE INFOCOM, Toronto, ON, Canada, July 2020, pp. 1698–1707

    2020

  27. [35]

    The MNIST database of handwritten digits,

    Y . LeCun, “The MNIST database of handwritten digits,” http://yann.lecun.com/exdb/mnist/, 1998

    1998

  28. [36]

    Fashion-MNIST: a novel image dataset for benchmarking machine learning algorithms,

    H. Xiao, K. Rasul, and R. V ollgraf, “Fashion-MNIST: a novel image dataset for benchmarking machine learning algorithms,”arXiv preprint arXiv:1708.07747, 2017

    2017 arXiv

  29. [37]

    Learning multiple layers of features from tiny images,

    A. Krizhevsky, “Learning multiple layers of features from tiny images,” M.S. thesis, University of Toronto, Department of Computer Science, Toronto, ON, Canada, 2009

    2009

  30. [38]

    Federated learning based on dynamic regularization,

    D. A. E. Acar, Y . Zhao, R. Matas, M. Mattina, P. Whatmough, and V . Saligrama, “Federated learning based on dynamic regularization,” in Proc. Int. Conf. Learn. Represent. (ICLR), May 2021, p. 1–36

    2021

Pith tools

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