Pith. sign in

REVIEW 4 major objections 4 minor 43 references

HEART: Achieving Timely Multi-Model Training for Vehicle-Edge-Cloud-Integrated Hierarchical Federated Learning

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper claims that jointly optimizing task-to-vehicle assignment and per-vehicle training order under a hybrid synchronous-asynchronous aggregation rule makes multi-model federated training over moving vehicles finish faster than four…

desk verdict Stress-test's contradiction dissolves on a close read; the paper is a competent heuristic contribution with a few presentation flaws that deserve peer review. read the letter →

arxiv 2501.09934 v3 pith:WRNIDUMJ submitted 2025-01-17 cs.LG cs.AI

Xiaohong Yang , Minghui Liwang , Xianbin Wang , Zhipeng Cheng , Seyyedali Hosseinalipour , Huaiyu Dai , Zhenzhen Jiao This is my paper · ORCID
classification cs.LGcs.AI
keywords hierarchicalfederatedlearningvehicle-edge-cloudmulti-modeltrainingtaskschedulinghybridsynchronous-asynchronousaggregationparticleswarmoptimizationgeneticalgorithmgreedyranking
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 addresses a setting in which vehicles simultaneously train several machine learning models in a hierarchical vehicle-edge-cloud federated learning system. It claims that the combination of a hybrid aggregation rule and a two-stage scheduler, HEART, minimizes the wall-clock time until every task reaches its target accuracy while keeping training effort balanced across tasks. The first stage assigns tasks to vehicles with a hybrid particle swarm and genetic algorithm; the second stage greedily orders each vehicle's assigned tasks by an aggregate score that rewards overlap with other vehicles and short upload times. If the claimed simulation results hold, balanced multi-task scheduling can cut total training time by roughly 20 to 40 percent compared with the tested baselines, without sacrificing individual-task convergence.

What carries the argument

The load-bearing mechanism is the hybrid synchronous-asynchronous aggregation rule combined with the two-stage scheduler HEART. At each edge server, aggregation for a task waits for all assigned vehicles' local models; at the cloud, global aggregation proceeds once a subset of edge models arrives, which the paper argues limits model obsolescence while keeping data utilization high. Stage 1 uses an improved PSO with dynamically decreasing inertia weight and a GA crossover-mutation step, guided by a fitness function that penalizes imbalanced task assignment and rewards assigning longer-training tasks. Stage 2 defines an aggregate-score, a weighted sum of an overlap score counting how many vehicles train the same task in the same rank position and an upload score inversely related to vehicle-to-edge upload time, and greedily builds each vehicle's training sequence to maximize it.

What would settle it

Run the same four-task, five-edge-server simulation with a nonzero downlink delay model, for example a finite broadcast data rate derived from a standard vehicular communication protocol, and check whether HEART's 110.2-second completion time still beats the 191.9-second TSPSO result; if a baseline improves more than HEART or the optimal task order shifts, the paper's central timing claim is undermined.

Watch

Extended reading notes

Core claim

The paper's central claim is that the min-max global training latency problem in VEC-HFL with multiple simultaneous ML tasks is NP-hard, and that its proposed HEART method reaches a near-optimal solution with low complexity. HEART combines an improved PSO-GA for balanced task-to-vehicle assignment with a greedy algorithm that ranks each vehicle's tasks to maximize an aggregate-score of inter-vehicle training overlap and upload efficiency. Under a hybrid rule that aggregates synchronously at edge servers and asynchronously at the cloud, the paper reports that all four tasks finish in 110.2 seconds, compared with 191.9 seconds for TSPSO, 154.1 seconds for TSGA, 215.8 seconds for TSSO, and 170.7 seconds for TSGD in the tested configuration.

Load-bearing premise

The timing model assumes edge servers and the cloud can broadcast updated models to vehicles with zero delay; if those downlink transmissions take real time, every completion time is underestimated and the optimized training order may no longer be optimal.

Editorial extensions

If this is right

  • Balancing task assignments across vehicles can shorten the overall wall-clock time even when no single task converges fastest, because all tasks finish at similar times instead of one task being prioritized at the expense of others.
  • The low-complexity greedy ranking step can be executed in near real time, which matters for vehicles that move between edge servers and have limited dwell time.
  • The hybrid synchronous-asynchronous aggregation rule reduces non-task-training time compared with a fully synchronous rule, with reported reductions of about 15.6 percent for 25 vehicles and 10 percent for 50 vehicles.
  • HEART's reported reductions in average and longest non-task-training time relative to TSGA, TSPSO, and TSSO suggest that task-order optimization mainly saves time by shrinking idle and upload gaps at the edge.
  • In the tested settings, increasing the number of vehicles from 25 to 50 preserves or improves HEART's time savings over the baselines, indicating the method scales with more participating vehicles.

Reading between the lines

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

  • Editorial inference: if downlink broadcast latency is added to the model, the optimal task order may change because vehicles would spend time receiving the next edge model; the greedy aggregate-score could be extended to include a downlink-aware term.
  • Editorial inference: the aggregate-score idea is not specific to vehicles and could be transferred to other hierarchical federated systems with mobile or heterogeneous clients, such as drones or smartphones, where upload windows vary over time.
  • Editorial inference: a testable extension is to combine HEART's assignment with stale-model control or adaptive client selection, since the paper's hybrid rule does not explicitly bound staleness at the cloud aggregation.
  • Editorial inference: the authors' own numbers imply that comparing HEART against a downlink-aware scheduler would isolate how much of the reported 110.2-second gain comes from task ordering versus from the zero-downlink-delay assumption.
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 considers the problem of training multiple ML models simultaneously in a vehicle-edge-cloud hierarchical federated learning (VEC-HFL) system. It proposes a hybrid synchronous-asynchronous aggregation rule, and a two-stage method, HEART, that first assigns tasks to vehicles using a combined PSO-GA heuristic and then orders the assigned tasks on each vehicle with a greedy algorithm that maximizes an 'aggregate score' of task-overlap and upload-time. The paper claims, based on simulations on real datasets, that HEART reduces global training latency while balancing task execution compared to several two-stage baselines. The central empirical claim is that HEART finishes all tasks faster than the baselines in wall-clock time.

Significance. The problem of multi-model training in VEC-HFL is timely and the paper addresses a genuine gap in the literature. The proposed hybrid aggregation rule is a plausible design choice, and the two-stage decomposition is practical. If the reported performance advantages were correct, the paper would provide a useful heuristic with reasonable complexity. However, the paper's headline claim of wall-clock superiority is contradicted by its own experimental data, and several load-bearing theoretical/modeling points are not adequately supported. The contribution is best seen as a design study with a heuristic method; the current empirical basis does not substantiate the claimed advantage.

major comments (4)
  1. [Section V, Fig. 4(d)] The claim that HEART achieves the fastest finish of all tasks is contradicted by the paper's own data. Fig. 4(d) reports the wall-clock completion time for the 20 Newsgroups task, which is the last task to finish for each method: TSGD completes at 103.4 s, while HEART completes at 110.2 s. The text compares HEART (110.2 s) against TSGD (170.7 s), but 170.7 s is the Driver-Yawning completion time from Fig. 4(c), not the overall completion time. Thus, the central superiority claim in the abstract and Section V is false as stated and cannot support the conclusion that HEART reduces training time.
  2. [Section IV-C] The NP-hardness of problem P is asserted without proof. The statement 'P belongs to a sub-category of integer programming (IP) problems which are known to be NP-hard' is insufficient, since many integer programs are polynomial-time solvable. Because the NP-hardness claim is used to justify the heuristic approach, the authors should provide a formal reduction from a known NP-hard problem (e.g., Partition or a scheduling problem) or qualify the claim as a conjecture.
  3. [Equation (19)] The fitness function in Eq. (19) appears to have a sign inconsistency with the text. The penalty for imbalance is written as -(ξ3 Σ|ψ_j−χ| + Σ ρ_j |ψ_j−χ|), i.e., with a negative sign, so larger ρ_j makes the fitness more negative for imbalance. However, the text states that ρ_j 'increases the reward for assigning tasks with longer training durations,' which suggests ρ_j should positively weight tasks with longer training times. Either the signs in Eq. (19) are wrong, or the description of ρ_j is misleading; this needs clarification.
  4. [Section III-A, footnote 1] The assumption that downlink communication delays (ES-to-vehicle and CS-to-ES) are negligible is load-bearing for the timing model. Equations (8)-(15) and the case (i)/(ii) classification depend on when a vehicle receives the next edge model; a nonzero downlink delay would add to the inactive time in case (ii) and could change the optimal task ordering. The authors should quantify typical downlink delays or provide a sensitivity analysis to show that the conclusions are robust to this assumption.
minor comments (4)
  1. [Section V] In Section V, the phrase 'notable lower than' should be 'notably lower than'.
  2. [Figure 3 caption] The caption of Fig. 3 labels the right subplot tasks as 'C1-C9' but the left subplot uses 'C1-C4'; this inconsistency should be resolved.
  3. [Algorithm 1, line 8] The notation 'for j ∈{ x⟨j⟩m,n;[g] = 1} in turn do' is unclear; it should specify the set of assigned tasks for vehicle n under ES m.
  4. [Equation (9)] The variable j' is used as the task preceding j in the vehicle's sequence, but the notation does not clearly define the relationship; a short explanation or a change of notation would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; HEART is a heuristic proposal evaluated against external baselines, though Section V contains a numerical inconsistency that is a correctness issue rather than a circularity.

full rationale

The derivation chain in HEART is self-contained. The timing equations (6)-(15) are definitions built from primitive quantities such as CPU cycles, clock frequency, data rates, and model sizes; they are identities used to construct the optimization objective, not fitted parameters later relabeled as predictions. Stage 1 (hybrid PSO-GA) and Stage 2 (greedy aggregate-score ranking) are heuristic optimization methods whose performance is measured empirically on real datasets against four external baselines plus a synchronous-aggregation baseline. No parameter is fitted to a subset of the reported completion times and then presented as a prediction of those same completion times. The one author-overlap citation, reference [15] (Chang, Hosseinalipour, Chiang), appears in the related-work survey as evidence that multi-model FL over wireless networks is a limited field; it is contextual and not load-bearing for HEART's algorithm, aggregation rule, or any claimed theorem. No uniqueness result is imported from it, and no ansatz is smuggled in via citation. The zero downlink delay assumption in Section III-A, footnote 1, is a stated modeling idealization; it affects the accuracy of the timing model but does not reduce any prediction to an input by construction. The fact that Stage 1's fitness function explicitly penalizes imbalanced assignment is an objective design choice, not a circular prediction of balance. The paper does contain an internal inconsistency in Section V: the text uses 170.7 s as TSGD's overall completion time, but Fig. 4(d) shows TSGD completing all tasks at 103.4 s, which is faster than HEART's 110.2 s. This affects the validity of the headline latency claim and is a correctness/accuracy concern, but it is not circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central optimization rests on a large set of hand-chosen coefficients and untested domain assumptions. No listed value is fitted against the target results, which keeps the circularity burden low, but the unreported xi_3 through xi_7, rho_j, chi, and pi parameters mean the search behavior and the success of the aggregate-score are effectively assumed. Dwell-time knowledge and zero downlink delay are the two most fragile domain assumptions.

free parameters (6)
  • task weight coefficient rho_j = unreported
    Introduced in Eq. (19) to reward assigning longer tasks; affects balance and scheduling; no value or tuning method given.
  • balance factors xi_3, xi_4, xi_5 = unreported
    xi_3 in Eq. (19) balances the fairness penalty; xi_4 and xi_5 in Eq. (21) are PSO social and cognitive coefficients; no values are reported.
  • aggregation score weights xi_6, xi_7 = unreported
    xi_6 in Eq. (28) scales upload time to overlap level; xi_7 in Eq. (29) trades overlap against upload time; both are central to the Stage 2 ranking and are unreported.
  • target assignment count chi = unreported
    Appears in Eq. (19) and is described as the target number of assignments per task; no concrete value or rule such as N/J is specified.
  • PSO inertia bounds pi_max, pi_min = unreported
    Control exploration in Eq. (20); values are not listed in Table II.
  • desired accuracy thresholds = 0.75, 0.9, 0.9, 0.75 for the four tasks
    Set in Section V; they define task completion and total time overhead, so they directly drive the headline wall-clock numbers.
assumptions (5)
  • domain assumption Downlink broadcast delays from ES to vehicles and CS to ES are negligible.
    Invoked in the footnote in Section III-A and used in all timing equations; the case (i)/(ii) classification and optimized rankings assume a model is received immediately after broadcast. If false, completion times are underestimated and task order may be suboptimal.
  • domain assumption Each vehicle remains inside its edge server's coverage long enough to finish all assigned tasks, and dwell time t_stay is known at scheduling time.
    Constraint (16) requires the sum of training times to be less than t_stay; the whole scheduler uses t_stay as a hard bound although real dwell times are stochastic and handovers can interrupt training.
  • ad hoc to paper Maximizing the hand-designed aggregate-score S_all yields training ranks that are near-optimal for problem P.
    Stage 2 solves a separate greedy maximization of Eq. (29). No theorem links S_all to the min-max objective in Eq. (18); if the score is not a good surrogate, the ranking stage has no formal grounding.
  • domain assumption The hybrid synchronous-asynchronous aggregation converges to target accuracy within the simulated global iterations.
    The convergence check (5) and target accuracies are used to stop training, but no convergence analysis of this hybrid aggregation rule is provided; the simulation results assume the rule behaves like a convergent FL method.
  • domain assumption Vehicle local training has deterministic per-round compute time with no contention or interruption.
    Eqs. (6)-(7) treat cycles per data point, CPU frequency, and model migration time as known constants; in actual vehicles, contention with other on-board functions could change completion times and invalidate the ranking.

how reviews work

0 comments
Cite this review

Pith. "Pith review of HEART: Achieving Timely Multi-Model Training for Vehicle-Edge-Cloud-Integrated Hierarchical Federated Learning." pith.science (2026). https://pith.science/paper/WRNIDUMJ

@misc{pith2026250109934,
  author       = {Pith},
  title        = {Pith review of: HEART: Achieving Timely Multi-Model Training for Vehicle-Edge-Cloud-Integrated Hierarchical Federated Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WRNIDUMJ}},
  note         = {Machine review of arXiv:2501.09934}
}
read the original abstract

The rapid growth of AI-enabled Internet of Vehicles (IoV) calls for efficient Machine Learning (ML) solutions that can handle high vehicular mobility and decentralized data. This has motivated the emergence of Hierarchical Federated Learning over vehicle-edge-cloud architectures (VEC-HFL). Nevertheless, one aspect which is underexplored in the literature on VEC-HFL is that vehicles often need to execute multiple ML tasks simultaneously, where this multi-model training environment introduces crucial challenges. First, improper aggregation rules can lead to model obsolescence and prolonged training times. Second, vehicular mobility may result in inefficient data utilization by preventing the vehicles from returning their models to the network edge. Third, achieving a balanced resource allocation across diverse tasks becomes of paramount importance as it majorly affects the effectiveness of collaborative training. We take one of the first steps towards addressing these challenges via proposing a framework for multi-model training in dynamic VEC-HFL with the goal of minimizing global training latency while ensuring balanced training across various tasks, a problem that turns out to be NP-hard. To facilitate timely model training, we introduce a hybrid synchronous-asynchronous aggregation rule. Building on this, we present a novel method called Hybrid Evolutionary And gReedy allocaTion (HEART). The framework operates in two stages: first, it achieves balanced task scheduling through a hybrid heuristic approach that combines improved Particle Swarm Optimization (PSO) and Genetic Algorithms (GA); second, it employs a low-complexity greedy algorithm to determine the training priority of assigned tasks on vehicles. Experiments on real-world datasets demonstrate the superiority of HEART over existing methods.

Figures

Figures reproduced from arXiv: 2501.09934 by the authors.

Figure 1
Figure 1. A schematic of multi-model training over VEC-HFL. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. A schematic of multi-task scheduling and training VEC-HFL [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The average number of times that a task has been executed on [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Test accuracy of the global models for different tasks: (a) [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]
Figure 6
Figure 6. Figure 6: Evaluations on non-task-training time overhead:(a) The aver [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 5
Figure 5. Figure 5: The time that it takes for the global model of all tasks [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 39 canonical work pages

  1. [1]

    Split Learning Over Wireless Networks: Parallel Design and Resource Management,

    W. Wu, M. Li, K. Qu, C. Zhou, X. Shen, W. Zhuang, X. Li, and W. Shi, “Split Learning Over Wireless Networks: Parallel Design and Resource Management,” IEEE J. Sel. Areas Commun., vol. 41, no. 4, pp. 1051–1066, 2023

    work page 2023

  2. [2]

    Joint Client Selection and Model Compression for Efficient FL in UA V-Assisted Wireless Networks,

    J. Cheng, G. Yuan, M. Zhou, S. Gao, C. Liu, H. Duan, and Q. Zeng, “Joint Client Selection and Model Compression for Efficient FL in UA V-Assisted Wireless Networks,” IEEE Trans. Veh. Technol., vol. 73, no. 10, pp. 15172–15184, 2024

    work page 2024

  3. [3]

    Adaptive Training and Aggregation for Federated Learning in Multi-Tier Computing Networks,

    W. Hou, H. Wen, N. Zhang, W. Lei, H. Lin, Z. Han, and Q. Liu, “Adaptive Training and Aggregation for Federated Learning in Multi-Tier Computing Networks,” IEEE Trans. Mobile. Com- put., vol. 23, no. 5, pp. 4376–4388, 2024

    work page 2024

  4. [4]

    GH- PFL: Advancing Personalized Edge-Based Learning Through Optimized Bandwidth Utilization,

    K. Mo, W. Lin, J. Lu, C. J. Xue, Y. Shao, and H. Xu, “GH- PFL: Advancing Personalized Edge-Based Learning Through Optimized Bandwidth Utilization,” IEEE Trans. Cloud Comput, vol. 13, no. 2, pp. 473–484, 2025

    work page 2025

  5. [5]

    HierFedML: Aggregator Placement and UE Assignment for Hierarchical Federated Learning in Mobile Edge Computing,

    Z. Xu, D. Zhao, W. Liang, O. F. Rana, P. Zhou, M. Li, W. Xu, H. Li, and Q. Xia, “HierFedML: Aggregator Placement and UE Assignment for Hierarchical Federated Learning in Mobile Edge Computing,” IEEE Trans. Parallel Distrib. Syst., vol. 34, no. 1, pp. 328–345, 2023

    work page 2023

  6. [6]

    HFL-TranWGAN: Knowledge-Driven Cross-Domain Collaborative Anomaly Detec- tion for End-to-End Network Slicing,

    Y. Wu, L. Liang, Y. Jia, and W. Wen, “HFL-TranWGAN: Knowledge-Driven Cross-Domain Collaborative Anomaly Detec- tion for End-to-End Network Slicing,” IEEE Trans. Netw. Serv. Manage, vol. 22, no. 1, pp. 760–776, 2025

    work page 2025

  7. [7]

    Device Scheduling and Assignment in Hierarchical Federated Learning for Internet of Things,

    T. Zhang, K.-Y. Lam, and J. Zhao, “Device Scheduling and Assignment in Hierarchical Federated Learning for Internet of Things,” IEEE Internet Things J., vol. 11, no. 10, pp. 18449– 18462, 2024

    work page 2024

  8. [8]

    Model-Oriented Training With Two-Stage Hierarchical Knowledge Distillation Under Non-IID Conditions in Federated Edge–Cloud Collabora- tion,

    Y. Chen, W. Dai, J. Han, Z. Qin, and S. Deng, “Model-Oriented Training With Two-Stage Hierarchical Knowledge Distillation Under Non-IID Conditions in Federated Edge–Cloud Collabora- tion,” IEEE Trans. Cloud Comput., vol. 13, no. 4, pp. 1205–1221, 2025

    work page 2025

Show all 43 references
  1. [9]

    Mobility Accelerates Learning: Convergence Analysis on Hierar- chical Federated Learning in Vehicular Networks,

    T. Chen, J. Yan, Y. Sun, S. Zhou, D. Gündüz, and Z. Niu, “Mobility Accelerates Learning: Convergence Analysis on Hierar- chical Federated Learning in Vehicular Networks,” IEEE Trans. Veh. Technol., vol. 74, no. 1, pp. 1657–1673, 2025

    2025

  2. [10]

    Adaptive Model Pruning for Hierarchical Wireless Federated Learning,

    X. Liu, S. Wang, Y. Deng, and A. Nallanathan, “Adaptive Model Pruning for Hierarchical Wireless Federated Learning,” IEEE Wirel. Commun. and Net. Conf. (WCNC)., Dubai, United Arab Emirates, 2024, pp. 1-6

    2024

  3. [11]

    MOB-FL: Mobility-Aware Federated Learning for Intelligent Connected Vehicles,

    B. Xie, Y. Sun, S. Zhou, Z. Niu, Y. Xu, J. Chen, et al., “MOB-FL: Mobility-Aware Federated Learning for Intelligent Connected Vehicles,” IEEE Int. Conf. Commun. (ICC), Rome, Italy, 2023, pp. 3951-3957

    2023

  4. [12]

    Distributed Deep Reinforcement Learning-Based Gradient Quantization for Federated Learning Enabled Vehicle Edge Computing,

    C. Zhang and W. Zhang, “Distributed Deep Reinforcement Learning-Based Gradient Quantization for Federated Learning Enabled Vehicle Edge Computing,” IEEE Internet Things J., vol. 12, no. 5, pp. 4899–4913, 2025

    2025

  5. [13]

    Efficient Vehicle Selec- tion and Resource Allocation for Knowledge Distillation-Based Federated Learning in UA V-Assisted VEC,

    C. Li, Y. Zhang, L. Yu, and M. Yang, “Efficient Vehicle Selec- tion and Resource Allocation for Knowledge Distillation-Based Federated Learning in UA V-Assisted VEC,” IEEE Trans. Intell. Transp. Syst., vol. 26, no. 5, pp. 6321–6331, 2025

    2025

  6. [14]

    FedICT: Federated Multi-Task Distillation for Multi-Access Edge Computing,

    Z. Wu, S. Sun, Y. Wang, M. Liu, Q. Pan, X. Jiang, et al., “FedICT: Federated Multi-Task Distillation for Multi-Access Edge Computing,” IEEE Trans. Parallel Distrib. Syst., vol. 35, no. 6, pp. 1107–1121, 2024

    2024

  7. [15]

    Asynchronous Multi-Model Dynamic Federated Learning Over Wireless Net- works: Theory, Modeling, and Optimization,

    Z. Chang, S. Hosseinalipour, and M. Chiang, “Asynchronous Multi-Model Dynamic Federated Learning Over Wireless Net- works: Theory, Modeling, and Optimization,” IEEE Trans. Cogn. Commun. Netw., vol. 10, no. 5, pp. 1989–2004, 2024

    1989

  8. [16]

    Communication-Efficient Federated Multi- task Learning Over Wireless Networks,

    H. Ma and H. Guo, “Communication-Efficient Federated Multi- task Learning Over Wireless Networks,” IEEE Internet Things J., vol. 10, no. 1, pp. 609–624, 2023

    2023

  9. [17]

    Matching Game for Multi-Task Federated Learning in Internet of Vehicles,

    Z. Li, H. Wu, Y. Lu, B. Ai, Z. Zhong, and Y. Zhang, “Matching Game for Multi-Task Federated Learning in Internet of Vehicles,” IEEE Trans. Veh. Technol., vol. 73, no. 2, pp. 1623–1636, 2024

    2024

  10. [18]

    Adaptive and Parallel Split Federated Learning in Vehicular Edge Computing,

    X. Qiang, Z. Chang, Y. Hu, L. Liu, and T. Hämäläinen, “Adaptive and Parallel Split Federated Learning in Vehicular Edge Computing,” IEEE Internet Things J., vol. 12, no. 5, pp. 4591–4604, 2025

    2025

  11. [19]

    Joint Participant Selection and Learning Scheduling for Multi-Model Federated Edge Learning,

    X. Wei, J. Liu, and Y. Wang, “Joint Participant Selection and Learning Scheduling for Multi-Model Federated Edge Learning,” IEEE Int. Conf. Mobile Ad Hoc and Smart Syst. (MASS), Denver, CO, USA, 2022, pp. 537-545

    2022

  12. [20]

    Mobility-Aware Multi-Task Decentralized Federated Learning for Vehicular Networks: Modeling, Analysis, and Op- timization,

    C. Dongyu, D. Tao, H. He, J. Juncheng, D. Mianxiong, Y. Di, et al., “Mobility-Aware Multi-Task Decentralized Federated Learning for Vehicular Networks: Modeling, Analysis, and Op- timization,” arXiv preprint arXiv:2503.06468, 2025

    2025 arXiv

  13. [21]

    Many-Task Feder- ated Fine-Tuning via Unified Task Vectors,

    T. Vasileios, O. Tanir, and M. Nirvana, “Many-Task Feder- ated Fine-Tuning via Unified Task Vectors,” arXiv preprint arXiv:2502.06376, 2025

    2025 arXiv

  14. [22]

    Fedml Parrot: A Scalable Federated Learning System via Heterogeneity-aware Scheduling on Sequential and Hierarchical Training,

    Z. Tang, X. Chu, R. Y. Ran, S. Lee, S. Shi, Y. Zhang, et al., “Fedml Parrot: A Scalable Federated Learning System via Heterogeneity-aware Scheduling on Sequential and Hierarchical Training,” arXiv preprint arXiv:2303.01778., 2023

    2023 arXiv

  15. [23]

    Task Selection and Resource Optimization in Multi-Task Federated Learning With Model Decomposition,

    H. Sun, M. Chen, Z. Yang, Y. Pan, Y. Cang, and Z. Zhang, “Task Selection and Resource Optimization in Multi-Task Federated Learning With Model Decomposition,” IEEE Commun. Lett., vol. 29, no. 1, pp. 225–229, 2025

    2025

  16. [24]

    Entropy and Mobility-Based Model Assignment for Multi-Model Vehicular Federated Learning,

    W. Lobato and D. Costa, “Entropy and Mobility-Based Model Assignment for Multi-Model Vehicular Federated Learning,” Inter. Conf. on Fed. Learning Tech. and App. (FLTA), Valencia, Spain, 2024, pp. 8-15

    2024

  17. [25]

    HiFlash: Communication-Efficient Hierarchical Federated Learning With Adaptive Staleness Control and Heterogeneity- Aware Client-Edge Association,

    Q. Wu, X. Chen, T. Ouyang, Z. Zhou, X. Zhang, S. Yang, et al., “HiFlash: Communication-Efficient Hierarchical Federated Learning With Adaptive Staleness Control and Heterogeneity- Aware Client-Edge Association,” IEEE Trans. Parallel Distrib. Syst., vol. 34, no. 5, pp. 1560–1579, 2023

    2023

  18. [26]

    Fed- Fetch: Faster Federated Learning with Adaptive Downstream Prefetching,

    Q. Yan, A. Liu, S. He, M. Lécuyer, and I. Beschastnikh, “Fed- Fetch: Faster Federated Learning with Adaptive Downstream Prefetching,” IEEE INFOCOM, London, United Kingdom, 2025, pp. 1-10

    2025

  19. [27]

    Optimizing Asynchronous Federated Learning: A Delicate Trade-Off Be- tween Model-Parameter Staleness and Update Frequency,

    A. Abdelkrim, C. Céline, J. Matthieu, and M. Éric, “Optimizing Asynchronous Federated Learning: A Delicate Trade-Off Be- tween Model-Parameter Staleness and Update Frequency,” arXiv preprint arXiv:2502.08206., 2025

    2025

  20. [28]

    HFEL: Joint edge association and resource allocation for cost-efficient hierar- chical federated edge learning,

    S. Luo, X. Chen, Q. Wu, Z. Zhou, and S. Yu, “HFEL: Joint edge association and resource allocation for cost-efficient hierar- chical federated edge learning,” IEEE Trans. Wireless Commun., vol. 19, no. 10, pp. 6535–6548, 2020

    2020

  21. [29]

    Joint Optimization of Platoon Control and Resource Scheduling in Cooperative Vehicle-Infrastructure System,

    P. Zhang, D. Tian, J. Zhou, X. Duan, Z. Sheng, D. Zhao, and D. Cao, “Joint Optimization of Platoon Control and Resource Scheduling in Cooperative Vehicle-Infrastructure System,” IEEE Trans. Intell. Veh., vol. 8, no. 6, pp. 3629–3646, 2023

    2023

  22. [30]

    Joint Optimiza- tion of Completion Ratio and Latency of Offloaded Tasks With Multiple Priority Levels in 5G Edge,

    P. Fard Moshiri, M. Simsek, and B. Kantarci, “Joint Optimiza- tion of Completion Ratio and Latency of Offloaded Tasks With Multiple Priority Levels in 5G Edge,” IEEE Trans. Netw. Serv. Manage, vol. 22, no. 2, pp. 1357–1371, 2025

    2025

  23. [31]

    Dynamic Stochastic Reorientation Particle Swarm Optimization for Adaptive Latent Factor Analysis in High-Dimensional Sparse Matrices,

    C. Lyu, Z. Ma, X. Luo, and Y. Shi, “Dynamic Stochastic Reorientation Particle Swarm Optimization for Adaptive Latent Factor Analysis in High-Dimensional Sparse Matrices,” IEEE Trans. Knowl. Data Eng., vol. 38, no. 1, pp. 222–234, 2026

    2026

  24. [32]

    UA V Swarm-Assisted Two-Tier Hierarchical Federated Learn- ing,

    T. Wang, X. Huang, Y. Wu, L. Qian, B. Lin, and Z. Su, “UA V Swarm-Assisted Two-Tier Hierarchical Federated Learn- ing,” IEEE Trans. Netw. Sci. Eng, vol. 11, no. 1, pp. 943–956, 2024

    2024

  25. [33]

    Optimization of STATCOM PI Controller Parameters Using the Hybrid GA-PSO Algorithm,

    K.-I. Mahider, M. Ferfra, and R. Rabeh, “Optimization of STATCOM PI Controller Parameters Using the Hybrid GA-PSO Algorithm,” IEEE Int. Conf. on Sys. and Cont. (ICSC)., Sousse, Tunisia, 2023, pp. 270-275

    2023

  26. [34]

    PSO-Algorithm- Assisted Attack-Compensated Control for 2-D Fuzzy Systems Under Cyber Attacks,

    Q. Zhang, H. Zhang, J. Zhang, and Z. Liang, “PSO-Algorithm- Assisted Attack-Compensated Control for 2-D Fuzzy Systems Under Cyber Attacks,” IEEE Trans. Cybern., vol. 56, no. 1, pp. 357–366, 2026

    2026

  27. [35]

    GA-MADDPG: A Demand-Aware UA V Network Adaptation Method for Joint Communication and Positioning in Emergency Scenarios,

    K. Zhuang, L. Xu, L. Li, L. Wang, and A. Fei, “GA-MADDPG: A Demand-Aware UA V Network Adaptation Method for Joint Communication and Positioning in Emergency Scenarios,” IEEE Wirel. Commun. and Net. Conf. (WCNC)., Glasgow, United Kingdom, 2023, pp. 1-6

    2023

  28. [36]

    Accuracy Assessment of Industrial Heritage Mapping Based on GA+BP Neural Networks Forecast,

    Q. Wang and P. Du, “Accuracy Assessment of Industrial Heritage Mapping Based on GA+BP Neural Networks Forecast,” IEEE Int. Conf. Appl. Intell. Sustain. Comput. (ICAISC), Dharwad, India, 2023, pp. 1-5

    2023

  29. [37]

    Very Deep Convolutional Net- works for Large-Scale Image Recognition,

    S. Karen and Z. Andrew, “Very Deep Convolutional Net- works for Large-Scale Image Recognition,” arXiv preprint arXiv:1409.1556., 2015

    2015 arXiv

  30. [38]

    Deep Residual Learning for Image Recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep Residual Learning for Image Recognition,” arXiv preprint arXiv:1512.03385, 2015

    2015 arXiv

  31. [39]

    Long Short-term Memory RNN,

    B. V. Christian, K. Adrian, and S. B. Erling, “Long Short-term Memory RNN,” arXiv preprint arXiv:2105.06756, 2021

    2021 arXiv

  32. [40]

    Privacy-Preserving Federated Learning for UA V-Enabled Net- works: Learning-Based Joint Scheduling and Resource Manage- ment,

    H. Yang, J. Zhao, Z. Xiong, K.-Y. Lam, S. Sun, and L. Xiao, “Privacy-Preserving Federated Learning for UA V-Enabled Net- works: Learning-Based Joint Scheduling and Resource Manage- ment,” IEEE J. Sel. Areas Commun., vol. 39, no. 10, pp. 3144– 3159, 2021

    2021

  33. [41]

    JHPF A-Net: Joint Head Pose and Facial Action Network for Driver Yawning Detection Across Arbitrary Poses in Videos,

    Y. Lu, C. Liu, F. Chang, H. Liu, and H. Huan, “JHPF A-Net: Joint Head Pose and Facial Action Network for Driver Yawning Detection Across Arbitrary Poses in Videos,” IEEE Trans. Intell. Transp. Syst., vol. 24, no. 11, pp. 11850–11863, 2023

    2023

  34. [42]

    On the optimization of UA V- assisted wireless networks for hierarchical federated learning,

    R. Khelf, E. Driouch, and W. Ajib, “On the optimization of UA V- assisted wireless networks for hierarchical federated learning,” IEEE Annual Int. Symp. Pers., Ind. Mobile Radio Commun. (PIMRC), Toronto, Canada, 2023, pp. 1-6

    2023

  35. [43]

    Toward Robust Hierarchical Federated Learning in Internet of Vehicles,

    H. Zhou, Y. Zheng, H. Huang, J. Shu, and X. Jia, “Toward Robust Hierarchical Federated Learning in Internet of Vehicles,” IEEE Trans. Intell. Transp. Syst., vol. 24, no. 5, pp. 5600–5614, 2023

    2023

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.