pith. sign in

arxiv: 2508.09532 · v3 · submitted 2025-08-13 · 💻 cs.LG · cs.AI· cs.NI

Decentralized Rank Scheduling for Energy-Constrained Multi-Task Federated Fine-Tuning in Edge-Assisted IoV Networks

Bokeng Zheng , Jianqiang Zhong , Jiayi Liu , Lei Xue , Xu Chen , Xiaoxi Zhang This is my paper

Pith reviewed 2026-05-18 23:08 UTC · model grok-4.3

classification 💻 cs.LG cs.AIcs.NI
keywords federated fine-tuningInternet of Vehiclesenergy constraintsbandit algorithmsrank selectionmulti-task learningedge-assisted networksonline learning
0
0 comments X

The pith

A hierarchical framework with the UCB-DUAL bandit algorithm lets vehicles adapt foundation models to multiple tasks while respecting global energy budgets in volatile IoV networks.

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

The paper proposes a method to handle the conflict between global energy limits, diverse task needs, and changing vehicle connectivity during federated fine-tuning of foundation models for IoV perception tasks. It splits the work into an infrastructure layer that uses feedback to reallocate energy budgets across tasks according to how fast each one is converging, and a vehicle layer where each car picks its model rank through an online learning problem. The vehicle decisions are handled by a new primal-dual bandit algorithm called UCB-DUAL that comes with sublinear regret bounds and folds the global energy rules into local choices. Simulations driven by real vehicle trajectories show the approach beats standard federated fine-tuning methods. This setup matters because it lets resource-limited vehicles keep adapting models without breaking energy rules or needing constant central coordination.

Core claim

The authors show that multi-task federated fine-tuning in edge-assisted IoV networks can be solved by decoupling the problem into two linked phases: an infrastructure-level feedback loop that redistributes global energy budgets across concurrent tasks according to real-time convergence dynamics and resource utilization, plus a vehicle-level formulation of intra-task rank selection as an energy-constrained online learning problem that is solved by the UCB-DUAL primal-dual bandit algorithm, which supplies theoretical sublinear regret guarantees and internalizes the global energy constraints so vehicles can independently trade off accuracy, latency, and power consumption.

What carries the argument

The UCB-DUAL primal-dual bandit algorithm for energy-constrained online rank selection at each vehicle, paired with the infrastructure feedback loop for dynamic energy budget redistribution.

If this is right

  • Global energy budgets are automatically shifted toward tasks that are converging more slowly or using resources less efficiently.
  • Each vehicle can choose model ranks locally without waiting for central commands while still respecting overall energy limits.
  • The bandit algorithm supplies regret bounds that grow only sublinearly with the number of decisions.
  • The framework scales to large numbers of vehicles and tasks by keeping most decisions decentralized.
  • Empirical results in real-trajectory simulations show clear gains over existing federated fine-tuning baselines.

Where Pith is reading between the lines

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

  • The same two-phase structure might apply to energy-aware adaptation of models in other mobile edge settings such as drones or smartphones with intermittent links.
  • Bandit methods for rank selection could be combined with other federated constraints like privacy budgets or latency deadlines.
  • Replacing the simulator with hardware-in-the-loop experiments on real vehicles would reveal whether the regret bounds hold under unmodeled radio effects.
  • Adding new perception tasks could be done by extending only the infrastructure redistribution rule without changing the vehicle algorithm.

Load-bearing premise

The large-scale simulator driven by real-world trajectory data is assumed to faithfully represent the high volatility of vehicular network connectivity and the heterogeneous task demands that shape energy redistribution and rank decisions.

What would settle it

A test in which UCB-DUAL fails to deliver sublinear regret or the full method loses its performance edge over baselines when run against actual fluctuating IoV links and task mixes would disprove the central claims.

Figures

Figures reproduced from arXiv: 2508.09532 by Bokeng Zheng, Jianqiang Zhong, Jiayi Liu, Lei Xue, Xiaoxi Zhang, Xu Chen.

Figure 1
Figure 1. Figure 1: Multi-task Federated Fine-Tuning with IoV [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: Impact of LoRA Rank on Federated Fine-Tuning. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Accuracy and latency comparison across methods. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Reward over commu￾nication rounds [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Scalability under in￾creasing number of clients [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
read the original abstract

Large-scale Internet of Vehicles (IoV) deployments increasingly demand the on-device adaptation of foundation models to support diverse, mission-critical perception tasks. While federated fine-tuning offers a promising solution for efficient model specialization, existing approaches often struggle to reconcile the inherent conflict between stringent global energy budgets, heterogeneous task demands, and the high volatility of vehicular network connectivity. In this work, we introduce a hierarchical, adaptive framework that decouples multi-task fine-tuning into two interdependent optimization phases. First, we implement a feedback-loop mechanism at the infrastructure level that dynamically redistributes global energy budgets across concurrent tasks based on real-time convergence dynamics and resource utilization. Second, at the vehicle level, we formulate intra-task rank selection as an energy-constrained online learning problem, solved via a novel primal-dual bandit algorithm, UCB-DUAL, which provides theoretical guarantees on sublinear regret. Our approach effectively internalizes global energy constraints into local decision-making, allowing vehicles to autonomously navigate the complex trade-off between model accuracy, latency, and power consumption. Empirical evaluations using a large-scale IoV simulator, driven by real-world trajectory data, confirm that our proposed method significantly outperforms current federated fine-tuning baselines, offering a robust and scalable solution for resource-constrained vehicular intelligence.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript proposes a hierarchical framework for energy-constrained multi-task federated fine-tuning in edge-assisted IoV networks. It decouples optimization into an infrastructure-level feedback mechanism that redistributes global energy budgets across tasks according to real-time convergence and resource utilization, and a vehicle-level formulation of intra-task rank selection as an energy-constrained online learning problem solved by the novel primal-dual bandit algorithm UCB-DUAL, which is claimed to deliver sublinear regret. Empirical results from a large-scale IoV simulator driven by real-world trajectory data are presented as confirming significant outperformance over existing federated fine-tuning baselines.

Significance. If the regret analysis holds and the simulator faithfully reproduces vehicular volatility and task heterogeneity, the work would offer a practical way to internalize global energy constraints into local rank decisions for foundation-model adaptation in dynamic IoV settings. The explicit coupling of a feedback energy allocator with a primal-dual bandit constitutes a clear technical contribution over prior federated fine-tuning methods that treat energy budgets as static.

major comments (3)
  1. [Theoretical Analysis / UCB-DUAL] The abstract asserts sublinear regret guarantees for UCB-DUAL, yet the manuscript provides no derivation outline, key lemmas, or explicit regret bound (e.g., dependence on time horizon T and number of ranks). Without these details the claim that the primal-dual formulation correctly enforces energy constraints while preserving sublinear regret cannot be verified and is load-bearing for the theoretical contribution.
  2. [Empirical Evaluation] The empirical evaluation relies on a single simulator driven by real-world trajectories but reports no error bars, ablation on hyperparameter choices for UCB-DUAL, or sensitivity analysis of simulator parameters governing connectivity volatility and task heterogeneity. These omissions directly weaken support for the central claim of significant outperformance over baselines.
  3. [Simulator Description] The simulator fidelity assumption—that real-world trajectory data faithfully reproduces the high volatility of vehicular connectivity and heterogeneous task demands that drive energy redistribution and rank selection—is stated without validation against held-out traces or comparison to alternative mobility models. This is load-bearing for the outperformance results.
minor comments (2)
  1. [Algorithm Description] Notation for the dual variable and energy budget constraint in the UCB-DUAL formulation should be introduced with a single consistent symbol set to avoid reader confusion between global and local budgets.
  2. [Abstract] The abstract would benefit from a brief statement of the concrete regret order (e.g., O(√T)) rather than the generic term “sublinear.”

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments on our manuscript. We address each major comment point by point below, indicating planned revisions where appropriate to strengthen the presentation of our contributions.

read point-by-point responses

  1. Referee: [Theoretical Analysis / UCB-DUAL] The abstract asserts sublinear regret guarantees for UCB-DUAL, yet the manuscript provides no derivation outline, key lemmas, or explicit regret bound (e.g., dependence on time horizon T and number of ranks). Without these details the claim that the primal-dual formulation correctly enforces energy constraints while preserving sublinear regret cannot be verified and is load-bearing for the theoretical contribution.

    Authors: We agree that the absence of a derivation outline, key lemmas, and explicit regret bound in the current manuscript makes independent verification difficult. The UCB-DUAL algorithm is constructed via a primal-dual approach in which dual variables dynamically enforce the energy constraints while the primal updates follow an upper-confidence-bound selection rule; the analysis aims to show that the cumulative regret remains sublinear in the time horizon. In the revised manuscript we will add a dedicated subsection (or appendix) that outlines the key lemmas, states the explicit regret bound and its dependence on T and the number of ranks, and sketches the main proof steps so that the theoretical claim can be verified. revision: yes

  2. Referee: [Empirical Evaluation] The empirical evaluation relies on a single simulator driven by real-world trajectories but reports no error bars, ablation on hyperparameter choices for UCB-DUAL, or sensitivity analysis of simulator parameters governing connectivity volatility and task heterogeneity. These omissions directly weaken support for the central claim of significant outperformance over baselines.

    Authors: We concur that reporting error bars, hyperparameter ablations, and sensitivity results would materially strengthen the empirical support. In the revision we will rerun the experiments over multiple independent seeds to obtain error bars, add an ablation study varying the exploration coefficient and dual-step-size parameters of UCB-DUAL, and include sensitivity plots for the simulator parameters that control connectivity volatility and task heterogeneity. revision: yes

  3. Referee: [Simulator Description] The simulator fidelity assumption—that real-world trajectory data faithfully reproduces the high volatility of vehicular connectivity and heterogeneous task demands that drive energy redistribution and rank selection—is stated without validation against held-out traces or comparison to alternative mobility models. This is load-bearing for the outperformance results.

    Authors: The simulator is driven by publicly available real-world trajectory datasets that have been employed in prior IoV studies precisely because they exhibit the volatility and heterogeneity characteristic of vehicular environments. While the current manuscript does not contain an explicit fidelity validation section, we will add a short discussion that reports basic statistical comparisons (e.g., inter-contact time distributions) between the used traces and held-out segments, together with a brief comparison to a standard random-waypoint mobility model to address the referee’s concern. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation and evaluation remain independent

full rationale

The paper presents a hierarchical framework decoupling energy redistribution (infrastructure-level feedback loop) from intra-task rank selection (vehicle-level UCB-DUAL primal-dual bandit with claimed sublinear regret bounds), then validates via empirical runs on a simulator driven by real-world trajectories. No quoted step equates a claimed prediction or first-principles result to its own fitted inputs, self-citations, or ansatz by construction. The central outperformance claim rests on simulator results rather than tautological re-labeling of hyperparameters or data subsets, and the theoretical guarantees are presented as derived rather than imported from overlapping prior work. This is the normal non-circular outcome for an algorithmic proposal with external empirical grounding.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 1 invented entities

The framework depends on standard assumptions about convergence dynamics and network volatility plus the new algorithm; no explicit free parameters or invented physical entities are named in the abstract.

invented entities (1)
  • UCB-DUAL algorithm no independent evidence
    purpose: Solves the energy-constrained intra-task rank selection as an online learning problem
    Novel primal-dual bandit introduced to provide sublinear regret guarantees under energy constraints

pith-pipeline@v0.9.0 · 5776 in / 1203 out tokens · 30217 ms · 2026-05-18T23:08:21.262304+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

44 extracted references · 44 canonical work pages · 1 internal anchor

  1. [1]

    Green internet of vehicles (iov) in the 6g era: Toward sustainable vehicular communications and networking,

    J. Wang, K. Zhu, and E. Hossain, “Green internet of vehicles (iov) in the 6g era: Toward sustainable vehicular communications and networking,” IEEE Transactions on Green Communications and Networking , vol. 6, no. 1, pp. 391–423, 2021

    work page 2021

  2. [2]

    Smart traffic man- agement system using internet of things,

    S. Javaid, A. Sufian, S. Pervaiz, and M. Tanveer, “Smart traffic man- agement system using internet of things,” in 2018 20th international conference on advanced communication technology (ICACT) . IEEE, 2018, pp. 393–398

    work page 2018

  3. [3]

    Smart traffic management system for traffic control using au- tomated mechanical and electronic devices,

    M. Rath, “Smart traffic management system for traffic control using au- tomated mechanical and electronic devices,” in IOP Conference Series: Materials Science and Engineering , vol. 377, no. 1. IOP Publishing, 2018, p. 012201

    work page 2018

  4. [4]

    Overview of intelligent vehicle infrastructure cooperative simulation technology for iov and automatic driving,

    Z. Ding and J. Xiang, “Overview of intelligent vehicle infrastructure cooperative simulation technology for iov and automatic driving,” World Electric Vehicle Journal, vol. 12, no. 4, p. 222, 2021

    work page 2021

  5. [5]

    Tomtit: Hierarchical federated fine- tuning of giant models based on autonomous synchronization,

    T. Qi, Y . Zhan, P. Li, and Y . Xia, “Tomtit: Hierarchical federated fine- tuning of giant models based on autonomous synchronization,” in IEEE INFOCOM 2024 - IEEE Conference on Computer Communications , 2024, pp. 1910–1919

    work page 2024

  6. [6]

    Language mod- els are few-shot learners,

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Kaplan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell et al., “Language mod- els are few-shot learners,” Advances in neural information processing systems, vol. 33, pp. 1877–1901, 2020

    work page 1901

  7. [7]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly et al., “An image is worth 16x16 words: Transformers for image recognition at scale,” in International Conference on Learning Representations , 2020

    work page 2020

  8. [8]

    Lora: Low-rank adaptation of large language models

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, W. Chen et al., “Lora: Low-rank adaptation of large language models.” ICLR, vol. 1, no. 2, p. 3, 2022

    work page 2022

  9. [9]

    Parameter-efficient fine-tuning for large models: A comprehensive survey,

    Z. Han, C. Gao, J. Liu, J. Zhang, and S. Q. Zhang, “Parameter-efficient fine-tuning for large models: A comprehensive survey,” Transactions on Machine Learning Research , 2024

    work page 2024

  10. [10]

    Federated Learning: Strategies for Improving Communication Efficiency

    J. Kone ˇcn´y, H. B. McMahan, F. X. Yu, P. Richt ´arik, A. T. Suresh, and D. Bacon, “Federated learning: Strategies for improving communication efficiency,”CoRR, vol. abs/1610.05492, 2016. [Online]. Available: http://arxiv.org/abs/1610.05492

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  11. [11]

    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 Artificial intelligence and statistics . PMLR, 2017, pp. 1273– 1282

    work page 2017

  12. [12]

    Fedlps: Heterogeneous federated learning for multiple tasks with local parameter sharing,

    Y . Jia, X. Zhang, A. Beheshti, and W. Dou, “Fedlps: Heterogeneous federated learning for multiple tasks with local parameter sharing,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, no. 11, 2024, pp. 12 848–12 856

    work page 2024

  13. [13]

    A blockchain-based roadside unit-assisted authentication and key agreement protocol for internet of vehicles,

    Z. Xu, W. Liang, K.-C. Li, J. Xu, and H. Jin, “A blockchain-based roadside unit-assisted authentication and key agreement protocol for internet of vehicles,” Journal of Parallel and Distributed Computing , vol. 149, pp. 29–39, 2021

    work page 2021

  14. [14]

    Federated learning: Challenges, methods, and future directions

    T. Li, A. K. Sahu, A. Talwalkar, and V . Smith, “Federated learning: Challenges, methods, and future directions,” IEEE Signal Processing Magazine , p. 50–60, May 2020. [Online]. Available: http://dx.doi.org/10.1109/msp.2020.2975749

    work page doi:10.1109/msp.2020.2975749 2020

  15. [15]

    Federated optimization in heterogeneous networks,

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

    work page 2020

  16. [16]

    Asynchronous fed- erated learning for edge-assisted vehicular networks,

    S. Wang, Q. Wu, Q. Fan, P. Fan, and J. Wang, “Asynchronous fed- erated learning for edge-assisted vehicular networks,” arXiv preprint arXiv:2208.01901, 2022

    work page arXiv 2022

  17. [17]

    T-drive trajectory data sample,

    Y . Zheng, “T-drive trajectory data sample,” August 2011, t-Drive sample dataset. [Online]. Available: https://www.microsoft.com/en- us/research/publication/t-drive-trajectory-data-sample/

    work page 2011

  18. [18]

    Parameter-efficient transfer learning for nlp,

    N. Houlsby, A. Giurgiu, S. Jastrzebski, B. Morrone, Q. De Laroussilhe, A. Gesmundo, M. Attariyan, and S. Gelly, “Parameter-efficient transfer learning for nlp,” in International conference on machine learning . PMLR, 2019, pp. 2790–2799

    work page 2019

  19. [19]

    Bitfit: Simple parameter- efficient fine-tuning for transformer-based masked language-models,

    E. B. Zaken, Y . Goldberg, and S. Ravfogel, “Bitfit: Simple parameter- efficient fine-tuning for transformer-based masked language-models,” in Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers) , 2022, pp. 1–9

    work page 2022

  20. [20]

    Prefix-tuning: Optimizing continuous prompts for generation,

    X. L. Li and P. Liang, “Prefix-tuning: Optimizing continuous prompts for generation,” in Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). Association for Computational Linguistics, 2021

    work page 2021

  21. [21]

    Combining lora to gpt-neo to reduce large language model hallucination,

    S.-h. Huang and C.-y. Chen, “Combining lora to gpt-neo to reduce large language model hallucination,” Research Square, 2024. [Online]. Available: https://doi.org/10.21203/rs.3.rs-4515250/v1

    work page doi:10.21203/rs.3.rs-4515250/v1 2024

  22. [22]

    Feddd: Toward communication-efficient federated learning with differential pa- rameter dropout,

    Z. Feng, X. Chen, Q. Wu, W. Wu, X. Zhang, and Q. Huang, “Feddd: Toward communication-efficient federated learning with differential pa- rameter dropout,” IEEE Transactions on Mobile Computing , vol. 23, no. 5, pp. 5366–5384, 2023

    work page 2023

  23. [23]

    Vfl- cafe: Communication-efficient vertical federated learning via dynamic caching and feature selection,

    J. Zhou, H. Liang, T. Wu, X. Zhang, Y . Jiang, and C. W. Tan, “Vfl- cafe: Communication-efficient vertical federated learning via dynamic caching and feature selection,” Entropy, vol. 27, no. 1, p. 66, 2025

    work page 2025

  24. [24]

    Giov: Achieving generative ai services in internet of vehicles via collaborative edge intelligence,

    G. Xie, R. Xie, X. Zhang, J. Nie, Q. Tang, W. Y . B. Lim, and D. Niyato, “Giov: Achieving generative ai services in internet of vehicles via collaborative edge intelligence,” in2024 IEEE Wireless Communications and Networking Conference (WCNC) . IEEE, 2024, pp. 1–6

    work page 2024

  25. [25]

    Gai-iov: Bridging generative ai and vehicular networks for ubiquitous edge intelligence,

    G. Xie, Z. Xiong, X. Zhang, R. Xie, S. Guo, M. Guizani, and H. V . Poor, “Gai-iov: Bridging generative ai and vehicular networks for ubiquitous edge intelligence,” IEEE Transactions on Wireless Communications , vol. 23, no. 4, pp. 1234–1248, 2024

    work page 2024

  26. [26]

    Transfer learning-driven intrusion detection for internet of vehicles (iov),

    Y . Otoum, Y . Wan, and A. Nayak, “Transfer learning-driven intrusion detection for internet of vehicles (iov),” in 2022 International Wireless Communications and Mobile Computing (IWCMC) . IEEE, 2022, pp. 342–347

    work page 2022

  27. [27]

    Online location planning for ai-defined vehicles: optimizing joint tasks of order serving and spatio-temporal heterogeneous model fine-tuning,

    B. Zheng, B. Rao, T. Zhu, C. W. Tan, J. Duan, Z. Zhou, X. Chen, and X. Zhang, “Online location planning for ai-defined vehicles: optimizing joint tasks of order serving and spatio-temporal heterogeneous model fine-tuning,” arXiv preprint arXiv:2502.04399 , 2025

    work page arXiv 2025

  28. [28]

    Federated learning on non-iid data silos: An experimental study,

    Q. Li, Y . Diao, Q. Chen, and B. He, “Federated learning on non-iid data silos: An experimental study,” in 2022 IEEE 38th international conference on data engineering (ICDE) . IEEE, 2022, pp. 965–978

    work page 2022

  29. [29]

    Heterogeneous lora for federated fine-tuning of on-device foundation models,

    Y . J. Cho, L. Liu, Z. Xu, A. Fahrezi, and G. Joshi, “Heterogeneous lora for federated fine-tuning of on-device foundation models,” in Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, 2024, pp. 12 903–12 913

    work page 2024

  30. [30]

    Computation-effective personalized federated learning: A meta learning approach,

    Z. Zhan and X. Zhang, “Computation-effective personalized federated learning: A meta learning approach,” in 2023 IEEE 43rd International Conference on Distributed Computing Systems (ICDCS) . IEEE, 2023, pp. 957–958

    work page 2023

  31. [31]

    Fedrema: Improving personalized federated learning via leveraging the most relevant clients,

    H. Liang, Z. Zhan, W. Liu, X. Zhang, C. W. Tan, and X. Chen, “Fedrema: Improving personalized federated learning via leveraging the most relevant clients,” in ECAI, 2024

    work page 2024

  32. [32]

    Adaptive per- sonalized federated learning for non-iid data with continual distribution shift,

    S. Chen, W. Liu, X. Zhang, H. Xu, W. Lin, and X. Chen, “Adaptive per- sonalized federated learning for non-iid data with continual distribution shift,” in 2024 IEEE/ACM 32nd International Symposium on Quality of Service (IWQoS). IEEE, 2024, pp. 1–6

    work page 2024

  33. [33]

    Dynamite: Dynamic interplay of mini-batch size and aggregation frequency for federated learning with static and streaming datasets,

    W. Liu, X. Zhang, J. Duan, C. Joe-Wong, Z. Zhou, and X. Chen, “Dynamite: Dynamic interplay of mini-batch size and aggregation frequency for federated learning with static and streaming datasets,” IEEE Transactions on Mobile Computing, vol. 23, no. 7, pp. 7664–7679, 2023

    work page 2023

  34. [34]

    Adacoopt: Leverage the interplay of batch size and aggregation frequency for federated learning,

    ——, “Adacoopt: Leverage the interplay of batch size and aggregation frequency for federated learning,” in 2023 IEEE/ACM 31st International Symposium on Quality of Service (IWQoS) . IEEE, 2023, pp. 1–10

    work page 2023

  35. [35]

    Edgec3: Online management for edge-cloud collaborative continuous learning,

    S. Lin, X. Zhang, Y . Li, C. Joe-Wong, J. Duan, and X. Chen, “Edgec3: Online management for edge-cloud collaborative continuous learning,” in 2023 20th Annual IEEE International Conference on Sensing, Com- munication, and Networking (SECON) . IEEE, 2023, pp. 411–419

    work page 2023

  36. [36]

    Online management for edge-cloud collaborative continuous learning: A two-timescale approach,

    S. Lin, X. Zhang, Y . Li, C. Joe-Wong, J. Duan, D. Yu, Y . Wu, and X. Chen, “Online management for edge-cloud collaborative continuous learning: A two-timescale approach,” IEEE Transactions on Mobile Computing, 2024

    work page 2024

  37. [37]

    A cloud-edge collaborative architecture for multimodal llms-based advanced driver assistance sys- tems in iot networks,

    Y . Hu, D. Ye, J. Kang, M. Wu, and R. Yu, “A cloud-edge collaborative architecture for multimodal llms-based advanced driver assistance sys- tems in iot networks,” IEEE Internet of Things Journal , vol. 11, no. 2, pp. 567–580, 2024

    work page 2024

  38. [38]

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

    H. Wang, Z. Kaplan, D. Niu, and B. Li, “Optimizing federated learning on non-iid data with reinforcement learning,” in IEEE INFOCOM 2020- IEEE Conference on computer communications. IEEE, 2020, pp. 1698– 1707

    work page 2020

  39. [39]

    Rosevin: Employing resource-and rate-adaptive edge super-resolution for video streaming,

    X. Zhang, H. Xu, L. Zou, J. Duan, C. Wu, Y . Xue, Z. Chen, and X. Chen, “Rosevin: Employing resource-and rate-adaptive edge super-resolution for video streaming,” in IEEE INFOCOM 2024-IEEE Conference on Computer Communications. IEEE, 2024, pp. 491–500

    work page 2024

  40. [40]

    Upper confidence primal- dual reinforcement learning for cmdp with adversarial loss,

    S. Qiu, X. Wei, Z. Yang, J. Ye, and Z. Wang, “Upper confidence primal- dual reinforcement learning for cmdp with adversarial loss,” Advances in Neural Information Processing Systems , vol. 33, pp. 15 277–15 287, 2020

    work page 2020

  41. [41]

    Recursive deep models for semantic compositionality over a sentiment treebank,

    R. Socher, A. Perelygin, J. Wu, J. Chuang, C. D. Manning, A. Y . Ng, and C. Potts, “Recursive deep models for semantic compositionality over a sentiment treebank,” in Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing . Association for Computational Linguistics, 2013, pp. 1631–1642

    work page 2013

  42. [42]

    A mathematical theory of communication,

    C. E. Shannon, “A mathematical theory of communication,” The Bell system technical journal , vol. 27, no. 3, pp. 379–423, 1948

    work page 1948

  43. [43]

    Introduction to the conll- 2003 shared task: Language-independent named entity recognition,

    E. F. Tjong Kim Sang and F. De Meulder, “Introduction to the conll- 2003 shared task: Language-independent named entity recognition,” in Proceedings of CoNLL-2003 . Association for Computational Linguis- tics, 2003, pp. 142–147

    work page 2003

  44. [44]

    Visual choice of plausible alternatives: An evaluation of image- based commonsense causal reasoning,

    J. Yeo, G. Lee, G. Wang, S. Choi, H. Cho, R. K. Amplayo, and S.-w. Hwang, “Visual choice of plausible alternatives: An evaluation of image- based commonsense causal reasoning,” in Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018). European Language Resources Association, 2018, pp. 1234– 1240

    work page 2018