Pith. sign in

REVIEW 3 major objections 6 minor 29 references

When a pinching antenna can move only at finite speed during local training, jointly choosing which devices train and where the antenna stops cuts age-of-information and speeds federated learning.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-30 17:58 UTC pith:UJWCGKYS

load-bearing objection Real systems increment on finite-speed pinching antennas in AoI-FL; math holds, learning-gain attribution in sims is messier than the abstract implies. the 3 major comments →

arxiv 2607.23595 v1 pith:UJWCGKYS submitted 2026-07-26 eess.SP

Age-of-Information Aware Federated Learning with Finite Speed Pinching Antenna

Kaidi Wang , Daniel K C So , Zhiguo Ding This is my paper
classification eess.SP
keywords federated learningpinching antennafinite antenna speedage-of-informationdevice selectionantenna placementcoalitional gamebranch-and-bound
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Federated learning over wireless links is slowed by stragglers and by repeatedly picking the same easy-to-reach devices, which leaves other devices’ updates stale. This paper treats a pinching antenna—a radiating point that can slide along a waveguide—as a practical resource whose speed is finite, so it can only travel during each round’s local-training window and only as far as the slowest selected device allows. That coupling means device choice simultaneously sets the training delay, the antenna’s reachable interval, the resulting upload rates, and the age-of-information of every device. The authors minimize total AoI under a hard round deadline by a coalitional-game device selector that re-evaluates antenna placement inside every candidate coalition, plus a branch-and-bound placement routine whose search interval is tightened to the intersection of the mobility window and the selected devices’ horizontal span. Simulations on non-IID MNIST and CIFAR-10 show faster convergence, lower sum AoI, and higher participation than fixed-antenna, midpoint, sequential, and random baselines, indicating that even modest antenna mobility can be turned into fresher, fairer federated updates.

Core claim

Under a finite pinching-antenna speed, the feasible movement range in each round is set by the slowest selected device’s training time; jointly optimizing the selected set and the antenna position inside that range to minimize overall AoI under a latency deadline accelerates learning convergence, reduces sum AoI, and raises device participation relative to fixed-antenna and non-AoI baselines.

What carries the argument

The mobility-constrained coalition utility: each candidate device set is scored by the negative of the resulting overall AoI, where the upload component is obtained by solving a one-dimensional non-convex placement problem over the interval formed by intersecting the antenna’s reachable segment with the horizontal span of the selected devices (closed-form for one device; branch-and-bound otherwise).

Load-bearing premise

All selected devices finish training before any upload begins, and every upload in the round uses the single antenna position reached inside that common training window.

What would settle it

Rerun the MNIST/CIFAR-10 experiments while allowing the antenna to keep moving during TDMA uploads or to serve devices from multiple simultaneous radiating points; if the AoI and accuracy gains over the fixed-antenna baseline disappear or reverse, the claimed coupling benefit does not hold under more flexible hardware.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • A device with longer local training time can be preferred because it enlarges the antenna’s reachable interval and may cut total upload time enough to lower overall AoI.
  • There exists a saturation moving speed equal to the horizontal distance to a device divided by its training time; above that speed further mobility yields no extra rate for that device.
  • The optimal antenna location always lies inside the intersection of the mobility window and the selected devices’ x-span (or on the nearer boundary if the two intervals are disjoint).
  • Moderate antenna speeds already capture most of the learning-speed gain of the infinite-speed ideal, so hardware need not be arbitrarily fast.

Where Pith is reading between the lines

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

  • The same training-window mobility budget could be exploited by other slowly reconfigurable surfaces (e.g., movable relays or liquid-metal antennas) whenever local compute time is non-negligible.
  • If multi-waveguide or multi-pinching-point hardware becomes available, the coalition utility would need an extra combinatorial layer over which points are activated, likely changing the participation–latency trade-off.
  • AoI-driven selection under mobility constraints may automatically compensate for mild non-IID skew by periodically “paying” extra training time to reach distant, under-sampled devices.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper studies AoI-aware device selection and antenna placement for federated learning assisted by a single pinching antenna with finite moving speed. The antenna is repositioned during the local training phase, so its feasible movement range in round t is set by the slowest selected device's training time (Eqs. (4)–(6)), creating a coupling between device selection, placement, uploading latency, and AoI. The authors (i) analyze the single-device rate gain over a fixed antenna and the gap to the infinite-speed benchmark with explicit bounds (Prop. 1); (ii) formulate an overall-AoI minimization under a round latency deadline (21) and decompose it into device selection (22) and placement (23); (iii) propose a join/leave coalitional game with coalition utility equal to negative overall AoI, showing finite termination and Nash stability of the terminal coalition; and (iv) derive a closed-form singleton placement (35), a reduced search region (Prop. 2, Cor. 1), and an ε-optimal branch-and-bound placement algorithm with valid interval bounds. Simulations on MNIST/CIFAR-10 and Monte Carlo system-level studies compare against fixed-antenna, midpoint-placement, sequential, and random baselines.

Significance. The finite-speed treatment is a genuinely practical gap in the pinching-antenna-FL literature, which has largely assumed instantaneous repositioning; modeling the training phase as the repositioning budget is a natural and well-motivated device. The analytical core is solid: I verified the algebra of Proposition 1 (including the ln(1+z) ≥ z/(1+z) step yielding (15)), the monotonicity claims, the search-region reduction in Proposition 2 and Corollary 1, and the validity of the BnB lower bound (44) and ε-optimality argument in §V-E. The paper ships parameter-free bounds, a closed-form singleton solution, and a globally convergent placement algorithm with explicit ε-optimality, which are real strengths. The simulation study covers two datasets and three system-level sweeps with stated parameters and averaging over 10 initial models. If the empirical attribution issues below are resolved, this is a useful, publishable contribution to an active area.

major comments (3)
  1. [§VI-A, Figs. 2–3] The antenna placement rule paired with the sequential-selection and random-selection baselines is never specified. The text states that the midpoint scheme uses midpoint placement and the proposed scheme uses BnB placement, but is silent on what placement accompanies the sequential and random selection baselines in Figs. 2–3. If those baselines are not given the same BnB-optimal placement, the reported convergence and sum-AoI gaps confound the selection policy with placement quality, and the headline claim (Abstract: the joint design outperforms baselines) cannot be cleanly attributed. This must be stated explicitly, and ideally an ablation (proposed selection + midpoint placement; sequential/random + BnB placement) should be added so the contributions of selection and placement are separable.
  2. [§VI-A, Figs. 2–3 (protocol)] In the learning experiments, 'device locations are randomly regenerated in each communication round.' This is difficult to reconcile with the paper's central mechanism: the cross-round mobility constraint (5)–(6) and the whole finite-speed narrative are built on spatial persistence, with the antenna repositioning sequentially from its previous location across rounds (§I-B). Re-scattering devices every round makes the protocol unlike any real FL deployment and turns the constraint into a memoryless per-round restriction. At minimum the authors should justify this choice and clarify how x_pin^(t-1) is carried over (the system-level section does carry the previous optimized position forward, so the two subsections appear to use different conventions). A supplementary experiment with a fixed deployment over rounds — as in Fig. 1 — would substantially strengthen the evidence for Figs. 2–3.
  3. [§VI-A, Fig. 1] Fig. 1 is the only experiment that isolates antenna moving speed as the causal variable for learning, and it uses a single hand-set geometry (uniform 2.5 m spacing below the waveguide, Cn=1 GHz, Dlm=50 Mbits, Tmax=8 s). With v_pin=0.5 m/s and T_tr of a few seconds, v·T_tr ≈ 3 m against a 47.5 m device span, so the mobility constraint binds very hard in this particular geometry; the claimed '~100 rounds' acceleration over the fixed antenna may not be representative. The claim in the Abstract that the scheme 'accelerates learning convergence' rests largely on this figure. At least one additional deployment geometry (e.g., clustered or two-sided device placement) or a sweep of v_pin·T_tr relative to the span should be shown to establish that the qualitative conclusion is not an artifact of this configuration.
minor comments (6)
  1. [§IV-C, item 3] The stated complexity of Algorithm 1 is O(CN), but each candidate-coalition evaluation requires solving the placement problem (via (35) or Algorithm 2). The true cost is O(CN·T_place); please state this, and ideally report measured per-round runtimes, since the placement is invoked inside the join/leave loops.
  2. [§V-B] The notation x(t) and x̄(t) (the reduced interval endpoints) collides visually with the round index and with x_min^(t)/x_max^(t); consider distinct symbols such as a^(t), b^(t). Also 'problem (31) is reduced to to the following problem' (duplicated word), and 'equivalent minimizing' (§III, after (23)) should be 'equivalent to minimizing'.
  3. [§II-E, Eq. (19)] Selected devices reset AoI to 0 rather than to the round latency T(t), so a device that uploads an update computed T_tr seconds earlier is treated as perfectly fresh. This follows the convention of [25] but should be acknowledged in one sentence, since it affects the interpretation of the sum-AoI curves.
  4. [Fig. 1(a)] The antenna-trajectory panel would benefit from marking device positions and the waveguide extent explicitly, and the caption should state that curves are averaged over 10 initial models with a 20-round moving-average smoothing (this is in the text but easy to miss).
  5. [References] Ref. [3] lists pages '1–1' (early access) — please update if final pagination is available; several 2025/2026 arXiv preprints (e.g., [11], [13], [14], [15], [17], [18]) should be updated if published versions now exist.
  6. [Table I] Table I omits several parameters that vary across figures (v_pin, Cn, Dlm, Pt, Tmax, A0). These are given in figure captions, but a consolidated table or a statement of which parameters are figure-specific would help reproducibility.

Circularity Check

0 steps flagged

No significant circularity: objectives, rate bounds, and algorithms are derived from stated models rather than fitted to or defined as their claimed outcomes.

full rationale

This is a self-contained wireless optimization paper. Overall AoI (20) is obtained directly from the standard AoI evolution (19) and round latency (18); the finite-speed mobility constraint (5)–(6) is an explicit physical budget, not a fit. Proposition 1’s rate gain/gap and bounds follow algebraically from the single-device rate (11) via ln(1+z) inequalities and monotonicity in residual distance rn(vpin); Proposition 2 and Corollary 1 reduce the multi-device placement region from monotonicity of fn in squared horizontal distance. Device selection maximizes coalition utility (24)–(25), which is defined as the negative of the same AoI objective, and BnB minimizes the stated uploading-time sum with valid interval bounds—so the algorithms optimize declared costs rather than rediscovering fitted targets. Self-citations ([7], [25], [8], etc.) supply background AoI-FL and pinching channel models that are fully restated with equations in §II; they are not load-bearing uniqueness theorems or ansatzes that force the numerical claims. Simulation comparisons are empirical evaluations against baselines, not predictions forced by construction. No self-definitional loop, fitted-input-as-prediction, or renaming of a known empirical law was found.

Axiom & Free-Parameter Ledger

4 free parameters · 6 axioms · 1 invented entities

Central claims rest on standard wireless and FL modeling choices plus one paper-specific operational protocol (move only during local training; single common upload position). No new physical entities. Free parameters are ordinary simulation knobs (speeds, T_max, powers), not fitted universal constants that manufacture the result.

free parameters (4)
  • v_pin (antenna moving speed) = scenario-dependent (e.g. 0.5 m/s in several figs)
    Swept in simulations (e.g. 0.5–1 m/s vs 0 and ∞); performance claims depend on chosen speeds relative to geometry and T_tr.
  • T_max (round latency deadline) = 7–20 s across figures
    Hard constraint in (21f); number of selected devices and AoI are highly sensitive to this hand-chosen threshold.
  • A_0 (initial AoI) = 20 or 50
    Sets freshness baseline; appears as 20 or 50 in experiments.
  • BnB optimality tolerance ε = 1e-4
    Stopping criterion for global antenna search; Table I sets 1e-4.
axioms (6)
  • domain assumption Pinching channel magnitude depends only on free-space distance to the radiating point; waveguide path contributes phase only (eq. 7).
    Standard simplified pinching-antenna model used throughout rate and placement analysis.
  • ad hoc to paper Local training time of a round equals the max over selected devices; antenna may move only within that window at constant speed v_pin (eqs. 4–6).
    Operational protocol that creates the paper’s coupling; not forced by physics alone.
  • domain assumption TDMA uplink; round latency is training max plus sum of upload times; AoI resets to 0 if selected else accumulates full round latency (18)–(19).
    Standard FL scheduling and AoI evolution assumptions from cited AoI-FL literature.
  • domain assumption Downlink global-model broadcast and selection notification complete before local training and are ignored in latency (footnote 1).
    Removes downlink from the optimization; common but load-bearing for round-time accounting.
  • domain assumption FedAvg aggregation with sample-size weights; non-IID data makes freshness/participation matter for accuracy.
    Motivation for AoI objective; learning gains are empirical under this premise.
  • ad hoc to paper Coalition utility equals negative overall AoI, with −∞ for empty or deadline-violating sets; join/leave only on strict utility increase.
    Defines Algorithm 1’s search dynamics and Nash-stability claim.
invented entities (1)
  • Training-time-constrained pinching-antenna movement budget coupled to AoI coalition utility no independent evidence
    purpose: Links device selection to feasible antenna interval and upload latency inside one round.
    Not a new particle or force; a protocol-level construct. Independent evidence would be hardware timing traces of pinching motion during on-device training—none provided.

pith-pipeline@v1.2.0-grok45-kimik3 · 23704 in / 3507 out tokens · 65666 ms · 2026-07-30T17:58:34.771204+00:00 · methodology

0 comments
read the original abstract

This paper investigates age-of-information (AoI) aware federated learning over wireless networks with finite speed pinching antennas. In contrast to existing studies that assume an infinitely high antenna moving speed, a practical round based training procedure is considered, where the pinching antenna is repositioned during the local training phase and its feasible movement range depends on the selected devices. This creates a new coupling among device selection, antenna placement, local training time, model uploading time, and AoI evolution. To characterize the impact of antenna moving speed, the rate gain over the fixed antenna and the gap to the infinite speed benchmark are analyzed. Subsequently, an overall AoI minimization problem is formulated under a round latency deadline by jointly optimizing the selected device set and the pinching antenna position. A coalitional game based device selection algorithm is proposed, where finite speed antenna placement is incorporated into the coalition utility evaluation. For antenna placement, the optimal search region is derived by exploiting the mobility constraint and device location span, based on which a branch-and-bound (BnB) algorithm is developed to obtain the global optimum. Simulation results show that the proposed scheme can accelerate learning convergence, reduce the sum AoI, and improve device participation compared with baseline schemes, demonstrating the potential of pinching antennas for enhancing federated learning through flexible spatial reconfiguration.

Figures

Figures reproduced from arXiv: 2607.23595 by Daniel K C So, Kaidi Wang, Zhiguo Ding.

Figure 2
Figure 2. Figure 2: Learning performance on MNIST under different [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 1
Figure 1. Figure 1: Learning performance on MNIST under different [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 4
Figure 4. Figure 4: Impact of antenna moving speed on system per [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: Learning performance on CIFAR-10 under different [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figure 6
Figure 6. Figure 6: Impact of latency deadline on system performance [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

29 extracted references · 1 linked inside Pith

  1. [1]

    Communication-efficient learning of deep networks from de centralized data,

    B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y Arc as, “Communication-efficient learning of deep networks from de centralized data,” in Artificial intelligence and statistics . PMLR, 2017, pp. 1273– 1282

    2017

  2. [2]

    Federated learnin g via over- the-air computation,

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

    2022

  3. [3]

    A j oint learning and communications framework for federated learn ing over wireless networks,

    M. Chen, Z. Y ang, W. Saad, C. Yin, H. V . Poor, and S. Cui, “A j oint learning and communications framework for federated learn ing over wireless networks,” IEEE Trans. Wireless Commun. , pp. 1–1, 2020

    2020

  4. [4]

    Federated learning: A signal processing perspective,

    T. Gafni, N. Shlezinger, K. Cohen, Y . C. Eldar, and H. V . Po or, “Federated learning: A signal processing perspective,” IEEE Signal Process. Mag., vol. 39, no. 3, pp. 14–41, 2022

    2022

  5. [5]

    Federated learn ing for wireless communications: Motivation, opportunities, and challenges,

    S. Niknam, H. S. Dhillon, and J. H. Reed, “Federated learn ing for wireless communications: Motivation, opportunities, and challenges,” IEEE Commun. Mag. , vol. 58, no. 6, pp. 46–51, 2020

    2020

  6. [6]

    Client selec tion in federated learning: Principles, challenges, and opport unities,

    L. Fu, H. Zhang, G. Gao, M. Zhang, and X. Liu, “Client selec tion in federated learning: Principles, challenges, and opport unities,” IEEE Internet Things J. , vol. 10, no. 24, pp. 21 811–21 819, 2023

    2023

  7. [7]

    Exploring age-o f- information weighting in federated learning under data het erogeneity,

    K. Wang, Z. Ding, D. K. C. So, and Z. Ding, “Exploring age-o f- information weighting in federated learning under data het erogeneity,” IEEE Trans. Wireless Commun. , vol. 24, no. 6, pp. 5023–5036, 2025

    2025

  8. [8]

    Flexible-ante nna systems: A pinching-antenna perspective,

    Z. Ding, R. Schober, and H. Vincent Poor, “Flexible-ante nna systems: A pinching-antenna perspective,” IEEE Trans. Commun. , pp. 1–1, 2025

    2025

  9. [9]

    Pinching antennas: Principles, applica- tions and challenges,

    Z. Y ang, N. Wang, Y . Sun, Z. Ding, R. Schober, G. K. Karagia nnidis, V . W. Wong, and O. A. Dobre, “Pinching antennas: Principles, applica- tions and challenges,” IEEE Wireless Commun. , pp. 1–10, Oct. 2025

    2025

  10. [10]

    Pinc hing- antenna systems: Architecture designs, opportunities, an d outlook,

    Y . Liu, Z. Wang, X. Mu, C. Ouyang, X. Xu, and Z. Ding, “Pinc hing- antenna systems: Architecture designs, opportunities, an d outlook,” IEEE Commun. Mag. , pp. 1–7, Sept. 2025

    2025

  11. [11]

    Pinching anten nas meet AI in next-generation wireless networks,

    F. Fang, Z. Ding, V . Leung, and L. Hanzo, “Pinching anten nas meet AI in next-generation wireless networks,” arXiv preprint arXiv:2511.07442, 2025

    arXiv 2025

  12. [12]

    Antenna activation fo r NOMA as- sisted pinching-antenna systems,

    K. Wang, Z. Ding, and R. Schober, “Antenna activation fo r NOMA as- sisted pinching-antenna systems,” IEEE Wireless Commun. Lett., vol. 14, no. 5, pp. 1526–1530, Mar. 2025

    2025

  13. [13]

    Generalized pinching-antenna systems: A tutorial on principles, desig n strategies, and future directions,

    Y . Xu, J. Cui, Y . Zhu, Z. Ding, T.-H. Chang, R. Schober, V . W. Wong, O. A. Dobre, G. K. Karagiannidis, H. V . Poor et al. , “Generalized pinching-antenna systems: A tutorial on principles, desig n strategies, and future directions,” arXiv preprint arXiv:2510.14166 , 2025

    arXiv 2025

  14. [14]

    Generalized pinching-a ntenna systems: A leaky-coaxial-cable perspective,

    K. Wang, Z. Ding, and L. Hanzo, “Generalized pinching-a ntenna systems: A leaky-coaxial-cable perspective,” arXiv preprint arXiv:2512.04979, 2025

    Pith/arXiv arXiv 2025

  15. [15]

    Pinching-antenna-enable d federated learning: Tail latency, participation, and convergence an alysis,

    Y . Lin, Z. Chen, and Z. Ding, “Pinching-antenna-enable d federated learning: Tail latency, participation, and convergence an alysis,” arXiv preprint arXiv:2510.23315, 2025

    arXiv 2025

  16. [16]

    Straggler-resilie nt federated learning over a hybrid conventional and pinching antenna ne twork,

    B. Wu, F. Fang, M. Zeng, and X. Wang, “Straggler-resilie nt federated learning over a hybrid conventional and pinching antenna ne twork,” IEEE Trans. V eh. Technol., pp. 1–6, 2026

    2026

  17. [17]

    Energy-efficient over-the-ai r federated learn- ing via pinching antenna systems,

    S. Asaad and A. Bereyhi, “Energy-efficient over-the-ai r federated learn- ing via pinching antenna systems,” arXiv preprint arXiv:2602.14250 , 2026

    arXiv 2026

  18. [18]

    Pinching antennas- assisted low- latency federated learning over multi-user wireless netwo rks,

    S. Asaad, H. Tabassum, and P . Wang, “Pinching antennas- assisted low- latency federated learning over multi-user wireless netwo rks,” arXiv preprint arXiv:2603.08595, 2026

    arXiv 2026

  19. [19]

    Tail-latency-aware federated lear ning with pinching antenna: Latency, participation, and placement,

    Y . Lin and Z. Ding, “Tail-latency-aware federated lear ning with pinching antenna: Latency, participation, and placement,” Entropy, vol. 28, no. 3, p. 341, 2026

    2026

  20. [20]

    Rate maximizat ion for downlink pinching-antenna systems,

    Y . Xu, Z. Ding, and G. K. Karagiannidis, “Rate maximizat ion for downlink pinching-antenna systems,” IEEE Wireless Commun. Lett. , pp. 1–1, 2025

    2025

  21. [21]

    Modeling a nd beamforming optimization for pinching-antenna systems,

    Z. Wang, C. Ouyang, X. Mu, Y . Liu, and Z. Ding, “Modeling a nd beamforming optimization for pinching-antenna systems,” IEEE Trans. Commun., vol. 73, no. 12, pp. 13 904–13 919, Oct. 2025

    2025

  22. [22]

    Minimum data rate maximization for uplink pinching-anten na systems,

    S. A. Tegos, P . D. Diamantoulakis, Z. Ding, and G. K. Kara giannidis, “Minimum data rate maximization for uplink pinching-anten na systems,” IEEE Wireless Commun. Lett., vol. 14, no. 5, pp. 1516–1520, Mar. 2025

    2025

  23. [23]

    Uplink and downlink communications in segmented waveguide-enabled p inching- antenna systems (SW ANs),

    C. Ouyang, H. Jiang, Z. Wang, Y . Liu, and Z. Ding, “Uplink and downlink communications in segmented waveguide-enabled p inching- antenna systems (SW ANs),” IEEE Trans. Commun. , pp. 1–1, 2026

    2026

  24. [24]

    Antenna acti vation and resource allocation in multi-waveguide pinching-antenna systems,

    K. Wang, Z. Ding, and G. K. Karagiannidis, “Antenna acti vation and resource allocation in multi-waveguide pinching-antenna systems,” IEEE Trans. Wireless Commun. , pp. 1–1, Sept. 2025

    2025

  25. [25]

    Age-of-inform ation min- imization in federated learning based networks with non-II D dataset,

    K. Wang, Z. Ding, D. K. C. So, and Z. Ding, “Age-of-inform ation min- imization in federated learning based networks with non-II D dataset,” IEEE Trans. Wireless Commun. , vol. 23, no. 8, pp. 8939–8953, 2024

    2024

  26. [26]

    Convergence acceleration in wireless federated learning : A stackelberg game approach,

    K. Wang, Y . Ma, M. B. Mashhadi, C. H. Foh, R. Tafazolli, an d Z. Ding, “Convergence acceleration in wireless federated learning : A stackelberg game approach,” IEEE Trans. V eh. Technol., vol. 74, no. 1, pp. 714–729, Jan. 2025

    2025

  27. [27]

    The age of information in networks: Moment s, distri- butions, and sampling,

    R. D. Y ates, “The age of information in networks: Moment s, distri- butions, and sampling,” IEEE Trans. Inf. Theory , vol. 66, no. 9, pp. 5712–5728, 2020

    2020

  28. [28]

    Han, Game theory in wireless and communication networks: theory , models, and applications

    Z. Han, Game theory in wireless and communication networks: theory , models, and applications . Cambridge University Press, 2012

    2012

  29. [29]

    Branch and bound methods,

    S. Boyd and J. Mattingley, “Branch and bound methods,” Notes for EE364b, Stanford University , vol. 2006, p. 07, 2007

    2006