Pith. sign in

REVIEW 4 major objections 4 minor 25 references

Computation- and Communication-Efficient Online FL for Resource-Constrained Aerial Vehicles

T0 review · 4 major / 4 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 Resource-limited drones can prune models and quantize gradients in online federated learning with little accuracy loss.

desk verdict A systems recipe that works in simulation, but Theorem 1's proof is invalid as written and the empirical support is too thin to compensate. read the letter →

arxiv 2506.02972 v2 pith:UCBH4R4Y submitted 2025-06-03 cs.LG cs.NIcs.SYeess.SY

Ferdous Pervej , Richeng Jin , Md Moin Uddin Chowdhury , Simran Singh , İsmail Güvenç , Huaiyu Dai This is my paper
classification cs.LGcs.NIcs.SYeess.SY
keywords aerialfederatedlearningonlinemodelpruninggradientquantizationtrajectoryoptimizationtime-varyingdatadistributionsconvergenceanalysisresource-constrainedUAVs
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

This paper tries to show that online federated learning can be made practical for battery- and bandwidth-limited aerial vehicles without sacrificing model quality. The proposed 2CEOAFL algorithm has each aerial client prune the received dense model into a shallow winning ticket, train only that pruned model, and then probabilistically quantize the accumulated gradient before uploading it to the central server. Trajectories are planned so each vehicle collects fresh data from clusters whose time-varying importance it estimates from a Gaussian mixture model. The main theoretical result, Theorem 1, bounds the expected squared norm of the masked global gradient by a sum of error terms from stochastic gradients, quantization, local data distribution shift, gradient dissimilarity, and a pruning term that only disappears when nothing is pruned. Simulations on CIFAR-10 with ResNet-18 show the pruned-and-quantized scheme reaching test accuracy within a few points of the full unpruned, unquantized federated baseline while using less computation and communication.

What carries the argument

The load-bearing object is the masked global gradient $\overline{\nabla} f(w^t|D^t) := \sum_{u=0}^{U-1} \alpha_u \nabla f_u(w^t|D^t_u) \odot m^t_u$, whose expected squared norm Theorem 1 bounds. The binary mask $m^t_u$ records which weights survive pruning at vehicle $u$ in round $t$; it is what makes the trained model shallow, and it is also what keeps the pruning error alive in the bound. The proof combines this masked gradient with the probabilistic upload rule $\Pi^t_u = d^t_u$ with probability $q^t_u$ and $\Pi^t_u = Q(d^t_u)$ otherwise, where $Q$ is the low-precision stochastic quantizer of Definition 1 and $d^t_u$ is the accumulated pruned-model gradient difference. The machinery is a standard smoothness-and-variance decomposition: $\beta$-Lipschitz smoothness turns the loss change into gradient inner products, the bias-free assumptions strip the quantization and mini-batch randomness, and the distribution-shift assumption $\|\nabla f_u(w|D^{t-1}_u)-\nabla f_u(w|D^t_u)\|^2\le\Phi^t_u$ converts the online data drift into one of the error terms. The trajectory model is a separate supporting mechanism: a mixed-integer convex program chooses cluster visits according to a time-varying class distribution $\psi^t_{u,c}$ normalized from $M_u z_u(t)$.

What would settle it

Run 2CEOAFL on a synthetic non-stationary task where the data distribution shifts sharply at a known round, and check whether Theorem 1's inequality (19) holds at that round using measured values; if the expected squared masked-gradient norm exceeds the right-hand side while all stated assumptions hold, the loss-drift step in the proof fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that a fleet of independently owned aerial connected vehicles can run online federated learning under tight onboard constraints by combining three ideas: selfish trajectory optimization that follows each vehicle's time-varying data distribution, model pruning via the lottery-ticket hypothesis so each vehicle trains only a small subnetwork, and probabilistic low-precision quantization of the uploaded gradient differences. Under the stated assumptions, Theorem 1 upper-bounds the expected squared norm of the masked global gradient $\overline{\nabla} f(w^t|D^t)$ by a telescoping loss-change term plus explicit error terms for stochastic gradient noise, quantization variance, local data distribution shift, gradient dissimilarity, and model pruning. The pruning error term is proportional to $\sum_u \alpha_u \delta^t_u \mathbb{E}[\|w^t_u\|^2]$ and does not vanish as learning rates decay unless no weights are removed, so convergence is to a neighborhood of a stationary point rather than to the point itself. The experiments compare 2CEOAFL against the unpruned/unquantized AFL baseline, pruning-only and quantization-only baselines, and centralized SGD; after 100 rounds on CIFAR-10 the proposed algorithm reaches roughly 47.5% test accuracy against about 48.7% for the full AFL baseline, with substantially lower communication overhead.

Load-bearing premise

The argument depends on the training objective not changing too much from one round to the next as new data arrives, but the paper only bounds how much the gradients change, never how much the loss itself changes.

Editorial extensions

If this is right

  • Aerial vehicles can cut per-round computation by training only the pruned subnetwork and cut uplink payload by quantizing gradients, while the global model stays within a few accuracy points of the unpruned, unquantized baseline on the tested CIFAR-10 task.
  • The convergence bound gives designers a checklist of where accuracy loss comes from: stochastic gradient noise, quantization variance, drifting local data, heterogeneity, and pruning; each term can be read off and, except for pruning, driven down with decaying learning rates.
  • Because the pruning error term does not vanish unless $\bar w^{t,0}_u = w^{t,0}_u$, even infinite rounds will not reach a stationary point exactly when any weights are removed; the practical correction is to keep pruning ratios modest.
  • The trajectory-optimization component connects data freshness to model quality: vehicles fly toward clusters whose time-varying importance is high, so the online dataset and the training objective evolve together rather than in opposition.
  • The probabilistic quantization rule with probability $q^t_u$ tied to the pruning ratio gives a tuneable knob: when resources are scarce, quantize more; when they are plentiful, send raw gradients.

Reading between the lines

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

  • The proof of Theorem 1 applies $\beta$-smoothness to $f(w^{t+1}|D^{t+1})$ versus $f(w^t|D^t)$, two different loss functions because the dataset changes every round, without bounding the drift of the loss itself; a reader should treat the convergence guarantee as conditional on an unstated loss-drift condition.
  • A natural stress test is to hold the quantization probability fixed and compare lottery-ticket pruning with random pruning at equal sparsity; the paper's framework predicts that the lottery-ticket mask, not sparsity alone, is what keeps accuracy comparable.
  • The bound suggests a per-round protocol where the central server adaptively sets $q^t_u$ and $\delta^t_u$ from radio-resource feedback, because the error terms depend explicitly on both; the authors list joint optimization as future work, so this is an extrapolation.
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

4 major / 4 minor

Summary. The paper proposes 2CEOAFL, an online aerial federated learning algorithm that combines trajectory-aware data sensing with two resource-saving mechanisms: pruning of the received dense model before local training, and probabilistic stochastic quantization of uploaded gradient differences. The authors model the time-varying spatial data distribution as a Gaussian mixture with time-dependent class weights, solve a per-ACV trajectory optimization problem, and then state Theorem 1, which upper-bounds the expected squared norm of the masked global gradient by a sum of terms attributed to stochastic gradient noise, quantization variance, local data distribution shift, gradient dissimilarity, and model pruning. The empirical section reports CIFAR10/ResNet18 experiments showing that 2CEOAFL achieves test accuracy comparable to the unpruned/unquantized AFL baseline while reducing communication overhead. The proof of the theorem is relegated to an online supplementary file.

Significance. If the theoretical bound were valid, the paper would give a useful decomposition of error sources for online FL with model pruning and gradient quantization, and the trajectory/data-distribution model is an interesting attempt to couple ACV motion planning with continual learning. The empirical comparison against AFL, AFL-Prune, AFL-Quant, and centralized SGD is informative and supports the qualitative claim of communication savings at modest accuracy cost. On the other hand, the paper does not ship code or machine-checked proofs, the theoretical derivation is the central advertised contribution, and the experimental evaluation is a single run without multiple seeds or confidence intervals.

major comments (4)
  1. [Supplementary Materials, Eq. (23)] The first inequality of the proof, f(w^{t+1}|D^{t+1}) ≤ f(w^t|D^t) + ⟨∇f(w^t|D^t), w^{t+1} − w^t⟩ + (β/2)‖w^{t+1} − w^t‖², is not justified by Assumption 1. Lipschitz smoothness is a property of a single loss function; for the loss at round t+1 the valid expansion would use f(w^t|D^{t+1}) and ∇f(w^t|D^{t+1}). The written inequality requires an additional bound on the drift of the loss itself, such as f(w|D^{t+1}) − f(w|D^t), as well as a bound on the gradient difference ⟨∇f(w^t|D^{t+1}) − ∇f(w^t|D^t), w^{t+1} − w^t⟩. No such assumption is stated. Consequently the derivation of Theorem 1 is not established from the stated assumptions.
  2. [Theorem 1 and Eq. (19)-(20)] The term E[f(w^t|D^t)] − E[f(w^{t+1}|D^{t+1})] is treated as though it telescopes when the bound is averaged over t. Since the dataset D^t changes every round, this is not a telescoping difference; an extra term of the form E[f(w^{t+1}|D^t) − f(w^{t+1}|D^{t+1})] appears. Without a bound on this loss drift, the time-averaged bound in Eq. (20) is not a convergence bound and the stated θ-suboptimal convergence claim is unsupported. The paper's own Section III-C states that the impacts of newly arrived samples are 'not explicitly modeled,' which is consistent with this gap. The proof must either introduce an explicit loss-drift assumption or reformulate the result as a regret bound with the drift appearing as a non-vanishing error term.
  3. [Assumption 5, Eq. (17)] Assumption 5 imposes δ^t_u ≥ ‖w^t_u − w̄^{t,0}_u‖² / ‖w^t_u‖², which ties the pruning ratio to the relative squared difference between the dense and pruned models. This is not a property of the lottery-ticket pruning procedure described in Section III-A, and in the experiments δ^t_u is simply chosen uniformly at random from [0.05, 0.7] with no verification that (17) holds. Since the final pruning-error term in Theorem 1 is derived directly through this assumption, the bound applies only to systems satisfying a condition that the proposed algorithm does not guarantee. The authors should either prove that their pruning rule satisfies (17), measure the condition in the simulations, or replace the assumption with a verifiable bound.
  4. [Section III-B, Theorem 1 statement] Even before the proof details, the theorem's bound is not a statement of convergence: the leading term 2(E[f(w^t|D^t)] − E[f(w^{t+1}|D^{t+1})])/(η_t κ) may be negative when the data distribution shifts the loss upward, and the pruning term does not vanish for any positive δ. The authors' Remark 1 acknowledges that the global gradient 'may only converge to a neighborhood of a stationary point.' The paper should state precisely what is being proven: a per-round inequality is not the same as a convergence guarantee, and the notation θ (defined as an average over T) is used in Eq. (20) as if it were a bound that tends to zero, which is not shown.
minor comments (4)
  1. [Figure 3 caption] The caption contains a typo: 'CIF AR10' should be 'CIFAR10'.
  2. [Section II-C, constraint C5] The notation '∀t1 := t ∋ [(t+1) mod C = 0]' is confusing; please rewrite the constraint in a clearer form indexing the blocks of C rounds.
  3. [Section IV-A] The experimental section reports a single run; adding multiple seeds with mean and standard deviation would substantially strengthen the claim of 'comparable performance.'
  4. [Eq. (14)] The definition D^t := ⋃_u D^t_u is correct for the union of local datasets, but the notation f(w^t|D^t) is later used both for the global loss and for the loss evaluated on the time-varying global dataset; please clarify that the global function is the weighted sum defined in Eq. (14).

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Theorem 1 is a conditional bound with explicit assumptions; the Eq. (23) proof gap is a correctness issue, not a circular reduction.

full rationale

The paper's central result, Theorem 1, is a conditional convergence bound proved in the supplementary material; it is not a fitted prediction and it does not reduce to its inputs by construction. The pruning term in Eq. (19) follows from Assumption 5, which is a stated hypothesis relating the pruning ratio to the relative squared distance between dense and pruned models; using that hypothesis to rewrite the distance term is a legitimate conditional step, not a definitional identity, because delta is an algorithm parameter and the inequality is an explicit assumption. The self-citations in Definition 2, Eq. (7), the quantization scheme, and the overhead formulas are not load-bearing: Definition 2 is stated in full as an assumption, the dataset update is a simple union, and the time/energy comparison is elementary algebra; the supplementary proof is cited only as the location of the proof. The empirical comparison uses independently configured baselines (SGD, AFL, AFL-Prune, AFL-Quant) with fixed hyperparameters, so no fitted parameter is relabeled as a prediction. There is, however, a genuine mathematical gap in the supplement's Eq. (23), where beta-smoothness is applied across two different loss functions f(center dot | D^{t+1}) and f(center dot | D^t), and the first difference E[f(w^t|D^t)] - E[f(w^{t+1}|D^{t+1})] does not telescope under changing loss functions; this undermines Theorem 1 as stated, but it is a correctness/assumption gap rather than circularity. The paper's own limitations paragraph (Section III-C) also concedes that the impacts of newly arrived samples are not explicitly modeled. Accordingly, there is no significant circularity.

Assumptions & free parameters 6 free parameters · 8 assumptions · 0 invented entities

The theory rests on standard smoothness and bounded-variance assumptions plus several assumptions tailored to the online setting, most importantly the pruning ratio bound and the data distribution shift bound. The one critical unstated premise is the cross-dataset smoothness step in the proof. The simulation relies on many hand-chosen hyperparameters that shape the generated data and trajectories, but these are experiment settings rather than fitted values that generate the claimed result.

free parameters (6)
  • Nmax for training data = 420
    Controls the maximum number of new training samples acquired at each trajectory point in Equation (6). Chosen by hand for the simulation; the central learning results depend on it.
  • Nmax for test data = 80
    Controls the number of new test samples per round, making the test distribution also time-varying. Chosen by hand.
  • Cluster association threshold zeta = not stated numerically
    Appears in Equation (3) to decide whether a drone is inside a cluster. Its value affects trajectory constraints and is not specified in the paper.
  • Time-varying basis functions z_u(t) = sin(2pi t/T), cos(5pi t/T), sin(5pi t/T), cos(2pi t/T)
    The frequencies and number of basis functions are hand-chosen for the experiments and directly define how the class distributions evolve over time.
  • Random basis-to-class mapping matrix M_u = random, seed not given
    Generated randomly to map the low-dimensional basis to class distributions. Exact realization is unreported, preventing exact replication.
  • Pruning ratio delta_t_u and upload probability q_t_u = delta uniformly in [0.05,0.7], q_t_u = delta_t_u
    Chosen randomly for the simulation rather than optimized. The paper's own limitations section notes that pruning ratios are not optimized.
assumptions (8)
  • standard math Local loss functions are beta-Lipschitz smooth (Assumption 1).
    Stated in Section III-B and used throughout the proof of Theorem 1 to bound the decrease of the loss after a gradient step.
  • standard math Stochastic mini-batch gradients are unbiased with bounded variance (Assumption 2).
    Standard in SGD analysis; used in the proof to obtain the sigma^2 error terms.
  • domain assumption Bounded gradient dissimilarity between local and global gradients (Assumption 3).
    Assumes local gradient norms are controlled by the global gradient norm plus an epsilon^t_u term that depends on statistical data heterogeneity.
  • domain assumption The stochastic quantizer is unbiased and has variance bounded by q times the squared norm (Assumption 4).
    This is an idealization of QSGD-style quantization; it does not hold exactly for all finite-precision implementations but is a common analytic convenience.
  • ad hoc to paper Pruning ratio bound: delta_t_u >= ||w_t_u - bar w_t,0_u||^2 / ||w_t_u||^2 (Assumption 5).
    This assumption is tailored so that the pruning error in the final bound can be expressed directly in terms of delta_t_u. It is not implied by the lottery ticket pruning procedure described in the paper.
  • domain assumption Local data distribution shift is bounded by Phi_t_u (Definition 2).
    Assumes the squared difference of local gradients between consecutive rounds is bounded, which is needed to quantify the effect of continually sensed data.
  • ad hoc to paper Smoothness and Taylor expansion apply across f(w^t|D^t) and f(w^{t+1}|D^{t+1}) as if they were the same function (Supplementary Equation (23)).
    This is an unstated and unjustified assumption. The loss functions at D^t and D^{t+1} differ, and no bound on loss drift between datasets is provided. This is the load-bearing gap in the proof.
  • domain assumption Data is spatially distributed according to a Gaussian mixture model and temporal distributions follow a linear basis expansion with softmax normalization (Equations (1)-(2)).
    This is a modeling choice for the simulation and trajectory optimization, not a proven property of real sensor data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Computation- and Communication-Efficient Online FL for Resource-Constrained Aerial Vehicles." pith.science (2026). https://pith.science/paper/UCBH4R4Y

@misc{pith2026250602972,
  author       = {Pith},
  title        = {Pith review of: Computation- and Communication-Efficient Online FL for Resource-Constrained Aerial Vehicles},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UCBH4R4Y}},
  note         = {Machine review of arXiv:2506.02972}
}
read the original abstract

Privacy-preserving distributed machine learning (ML) and aerial connected vehicle (ACV)-assisted edge computing have drawn significant attention lately. Since the onboard sensors of ACVs can capture new data as they move along their trajectories, the continual arrival of such 'newly' sensed data leads to online learning and demands carefully crafting the trajectories. Besides, as typical ACVs are inherently resource-constrained, computation- and communication-efficient ML solutions are needed. Therefore, we propose a computation- and communication-efficient online aerial federated learning (2CEOAFL) algorithm to take the benefits of continual sensed data and limited onboard resources of the ACVs. In particular, considering independently owned ACVs act as selfish data collectors, we first model their trajectories according to their respective time-varying data distributions. We then propose a 2CEOAFL algorithm that allows the flying ACVs to (a) prune the received dense ML model to make it shallow, (b) train the pruned model, and (c) probabilistically quantize and offload their trained accumulated gradients to the central server (CS). Our extensive simulation results show that the proposed 2CEOAFL algorithm delivers comparable performances to its non-pruned and nonquantized, hence, computation- and communication-inefficient counterparts.

Figures

Figures reproduced from arXiv: 2506.02972 by the authors.

Figure 2
Figure 2. ACV’s cluster association 0 10 20 30 40 50 60 70 80 90 100 Training Round 0.30 0.32 0.34 0.36 0.38 0.40 0.42 0.44 0.46 0.48 0.50 0.52 0.54 Test Accuracy FL Round vs Test Acc on CIFAR10 with ResNet18 2CEOAFL AFL AFL-Prune AFL-Quant SGD [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Test accuracies with ResNet18: s = 3,δth = 0.7 0.2 41.1 82.0 122.9 163.7 204.6 245.5 286.4 327.3 368.2 Communication Overhead [bits] ×106 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 CDF CDF of Communication Overheads 2CEOAFL AFL AFL-Prune AFL-Quant [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

25 extracted references · 25 canonical work pages

  1. [1]

    UA V-enabled intelligent transportation systems for the smart city: Applications and challenges,

    H. Menouar, I. Guvenc, K. Akkaya, A. S. Uluagac, A. Kadri, and A. Tuncer, “UA V-enabled intelligent transportation systems for the smart city: Applications and challenges,” IEEE Commun. Magaz. , vol. 55, no. 3, pp. 22–28, 2017

    work page 2017

  2. [2]

    UA V hyperspectral remote sensing image classification: A systematic review,

    Z. Zhang, L. Huang, Q. Wang, L. Jiang, Y . Qi, S. Wang, T. Shen, B.-H. Tang, and Y . Gu, “UA V hyperspectral remote sensing image classification: A systematic review,” IEEE J. Sel. Top. Appl. Earth Observ. Remote Sens. , vol. 18, pp. 3099–3124, 2025

    work page 2025

  3. [3]

    UA V-enabled integrated sensing and communication: Opportunities and challenges,

    K. Meng, Q. Wu, J. Xu, W. Chen, Z. Feng, R. Schober, and A. L. Swindlehurst, “UA V-enabled integrated sensing and communication: Opportunities and challenges,” IEEE Wireless Commun., vol. 31, no. 2, pp. 97–104, 2024. 6Note that in our considered system model, while the training data distribution continually changes, the test dataset, which is accumulated ...

    work page 2024

  4. [4]

    Mobile edge computing and machine learning in the internet of unmanned aerial vehicles: a survey,

    Z. Ning, H. Hu, X. Wang, L. Guo, S. Guo, G. Wang, and X. Gao, “Mobile edge computing and machine learning in the internet of unmanned aerial vehicles: a survey,” ACM Computing Surveys, vol. 56, no. 1, pp. 1–31, 2023

    work page 2023

  5. [5]

    Resource constrained vehicular edge federated learning with highly mobile connected vehicles,

    M. F. Pervej, R. Jin, and H. Dai, “Resource constrained vehicular edge federated learning with highly mobile connected vehicles,” IEEE J. Sel. Areas in Commun. , vol. 41, no. 6, pp. 1825–1844, 2023

    work page 2023

  6. [6]

    Federated learning via unmanned aerial vehicle,

    M. Fu, Y . Shi, and Y . Zhou, “Federated learning via unmanned aerial vehicle,” IEEE Trans. Wireless Commun., vol. 23, no. 4, pp. 2884–2900, 2024

    work page 2024

  7. [7]

    Latency minimization for UA V-enabled federated learning: Trajectory design and resource allocation,

    X. Zhang, W. Liu, J. Ren, H. Xing, G. Gui, Y . Shen, and S. Cui, “Latency minimization for UA V-enabled federated learning: Trajectory design and resource allocation,” IEEE Internet Things J. , 2025

    work page 2025

  8. [8]

    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 Proc. AIStat. PMLR, 2017, pp. 1273–1282

    work page 2017

Show all 25 references
  1. [9]

    Online-score-aided federated learning: Taming the resource constraints in wireless networks,

    M. F. Pervej, M. Choi, and A. F. Molisch, “Online-score-aided federated learning: Taming the resource constraints in wireless networks,” arXiv preprint arXiv:2408.05886, 2024

    2024 arXiv

  2. [10]

    Federated learning in the sky: Joint power allocation and scheduling with UA V swarms,

    T. Zeng, O. Semiari, M. Mozaffari, M. Chen, W. Saad, and M. Bennis, “Federated learning in the sky: Joint power allocation and scheduling with UA V swarms,” in Proc. IEEE ICC , 2020

    2020

  3. [11]

    Federated learning-based collaborative wideband spectrum sensing and scheduling for uavs in utm systems,

    S. R. Chintareddy, K. Roach, K. Cheung, and M. Hashemi, “Federated learning-based collaborative wideband spectrum sensing and scheduling for uavs in utm systems,” IEEE Trans. Machine Learn. Commun. Network., 2025

    2025

  4. [12]

    UA V-assisted online machine learning over multi-tiered networks: A hierarchical nested personalized federated learning approach,

    S. Wang, S. Hosseinalipour, M. Gorlatova, C. G. Brinton, and M. Chi- ang, “UA V-assisted online machine learning over multi-tiered networks: A hierarchical nested personalized federated learning approach,” IEEE Trans. Network Serv. Manag. , vol. 20, no. 2, pp. 1847–1865, 2022

    2022

  5. [13]

    Air-ground integrated online federated learning under unreliable communication,

    Y . Jing, Y . Qu, T. Wu, C. Dong, S. Guo, and Q. Wu, “Air-ground integrated online federated learning under unreliable communication,” IEEE Trans. Cognitive Commun. Network. , 2025

    2025

  6. [14]

    Participant and sample selection for efficient online federated learning in UA V swarms,

    F. Wu, Y . Qu, T. Wu, C. Dong, K. Guo, Q. Wu, and S. Guo, “Participant and sample selection for efficient online federated learning in UA V swarms,” IEEE Internet Things J. , vol. 11, no. 12, pp. 21 202–21 214, 2024

    2024

  7. [15]

    Split federated learning for UA V-enabled integrated sensing, computation, and communication,

    X. Hou, J. Wang, Z. Zhang, J. Wang, L. Liu, and Y . Ren, “Split federated learning for UA V-enabled integrated sensing, computation, and communication,” arXiv preprint arXiv:2504.01443 , 2025

    2025

  8. [16]

    CVXPY: A python-embedded modeling language for convex optimization,

    S. Diamond and S. Boyd, “CVXPY: A python-embedded modeling language for convex optimization,” J. Machine Learn. Research, vol. 17, no. 83, pp. 1–5, 2016

    2016

  9. [17]

    Resource-aware hierarchical federated learning in wireless video caching networks,

    M. F. Pervej and A. F. Molisch, “Resource-aware hierarchical federated learning in wireless video caching networks,” IEEE Trans. Wireless Commun., vol. 24, no. 1, pp. 165–180, 2025

    2025

  10. [18]

    The lottery ticket hypothesis: Finding sparse, trainable neural networks,

    J. Frankle and M. Carbin, “The lottery ticket hypothesis: Finding sparse, trainable neural networks,” in Proc. ICLR, 2019

    2019

  11. [19]

    QSGD: Communication-efficient SGD via gradient quantization and encoding,

    D. Alistarh, D. Grubic, J. Li, R. Tomioka, and M. V ojnovic, “QSGD: Communication-efficient SGD via gradient quantization and encoding,” in Proc. NeurIPS, vol. 30, 2017

    2017

  12. [20]

    Model pruning enables efficient federated learning on edge devices,

    Y . Jiang, S. Wang, V . Valls, B. J. Ko, W.-H. Lee, K. K. Leung, and L. Tassiulas, “Model pruning enables efficient federated learning on edge devices,” IEEE Trans. Neur. Netw. Learn. Syst. , pp. 1–13, Apr. 2022

    2022

  13. [21]

    Joint model pruning and device selection for communication-efficient federated edge learning,

    S. Liu, G. Yu, R. Yin, J. Yuan, L. Shen, and C. Liu, “Joint model pruning and device selection for communication-efficient federated edge learning,” IEEE Trans. Commun., vol. 70, no. 1, pp. 231–244, Jan. 2022

    2022

  14. [22]

    Sparsified SGD with memory,

    S. U. Stich, J.-B. Cordonnier, and M. Jaggi, “Sparsified SGD with memory,” Advan. in NeurIPS , vol. 31, 2018

    2018

  15. [23]

    FedPAQ: A communication-efficient federated learning method with periodic averaging and quantization,

    A. Reisizadeh, A. Mokhtari, H. Hassani, A. Jadbabaie, and R. Pedarsani, “FedPAQ: A communication-efficient federated learning method with periodic averaging and quantization,” in Proc. AIStats, vol. 108. PMLR, 26–28 Aug 2020, pp. 2021–2031

    2020

  16. [24]

    Hierarchical federated learning in wireless networks: Pruning tackles bandwidth scarcity and system heterogeneity,

    M. F. Pervej, R. Jin, and H. Dai, “Hierarchical federated learning in wireless networks: Pruning tackles bandwidth scarcity and system heterogeneity,” IEEE Trans. Wireless Commun. , vol. 23, no. 9, pp. 11 417–11 432, 2024

    2024

  17. [25]

    Computation-and communication-efficient online FL for resource-constrained aerial vehicles,

    M.-F. Pervej, R. Jin, M. M. U. Chowdhury, S. Singh, ˙I. G ¨uvenc ¸, and H. Dai, “Computation-and communication-efficient online FL for resource-constrained aerial vehicles,” arXiv preprint arXiv:2506.02972 , 2025. SUPPLEMENTARY MATERIALS Additional notations: ¯gu ¯wt,τ u |Dt u...

    2025 arXiv

Pith tools

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