Personalized Federated Distillation Assisted Vehicle Edge Caching Strategy

Cui Zhang; Kezhi Wang; Pingyi Fan; Qiong Wu; Wen Chen; Xun Li

arxiv: 2512.09378 · v2 · submitted 2025-12-10 · 💻 cs.LG

Personalized Federated Distillation Assisted Vehicle Edge Caching Strategy

Xun Li , Qiong Wu , Pingyi Fan , Kezhi Wang , Wen Chen , Cui Zhang This is my paper

Pith reviewed 2026-05-16 23:35 UTC · model grok-4.3

classification 💻 cs.LG

keywords vehicle edge cachingfederated distillationpersonalized learningcontent predictioncommunication overhead reductionvehicular networksprivacy preservation

0 comments

The pith

Personalized federated distillation enables vehicle edge caching with reduced communication overhead while remaining robust to speed changes.

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

The paper proposes replacing full-model exchanges in federated learning with a distillation step so that vehicles can share condensed knowledge rather than entire parameter sets when predicting content for edge caching. This change addresses two practical failures of standard federated learning: excessive communication volume and the risk that a vehicle leaves the roadside-unit coverage before training finishes. Simulations show the resulting strategy keeps prediction quality high across a range of vehicle speeds and delivers measurable drops in total bits exchanged. The approach still relies on local data staying on the vehicle, preserving privacy without requiring a central server to see raw requests.

Core claim

The central claim is that a personalized federated distillation procedure can drive vehicle-edge content caching by letting each vehicle distill its local interest model into a compact representation that is periodically aggregated, thereby lowering communication cost and avoiding training interruptions caused by high mobility while still producing accurate cache decisions.

What carries the argument

Personalized federated distillation, which condenses each vehicle's local model into a smaller set of soft targets or logits that are exchanged instead of full parameter vectors.

Load-bearing premise

The simulated vehicle trajectories, content request patterns, and network conditions sufficiently represent real-world scenarios so that the reported robustness and overhead reductions will translate outside the simulation.

What would settle it

A field experiment that logs actual uplink and downlink bytes per vehicle and cache-hit rates while vehicles travel at recorded speeds through live roadside units, then compares those measurements against the paper's simulation curves.

Figures

Figures reproduced from arXiv: 2512.09378 by Cui Zhang, Kezhi Wang, Pingyi Fan, Qiong Wu, Wen Chen, Xun Li.

**Figure 1.** Figure 1: , where the vehicle edge computing network consists of three layers: a macro base station (MBS), RSUs, and vehicles. Each RSU r = 1, 2, . . . , R is connected to the MBS via a reliable wired link, and vehicles i = 1, 2, . . . , I are located within the coverage area of the RSUs. Each vehicle has its own local data di , and is equipped with a communication module, a distillation module, and locally pre-trai… view at source ↗

**Figure 2.** Figure 2: Execution between Vehicles and RSUs. 2) User Similarity Construction: Unlike [21], we calculate the similarity between users rather than individual samples. The main reason for this is that users only rate a limited number of items, which results in their data containing many zero values. Directly calculating the similarity between each sample would treat these unrated items as content that VUs are not int… view at source ↗

**Figure 3.** Figure 3: Loss versus episodes. • CAFR [13]: This scheme proposed the combination of asynchronous FL and autoencoder for edge caching, combined with a mobility aware mechanism, provides a feasible solution for vehicle edge caching. • N-τ -greedy: This scheme selects the contents with the N largest request counts based on probability 1 − τ and selects N contents randomly based on probability τ to cache. In the experi… view at source ↗

**Figure 4.** Figure 4: Request content delay versus cache capacity. [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗

**Figure 5.** Figure 5: Cache hit percentage versus vehicle speed. [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗

read the original abstract

Vehicle edge caching is a promising technology that can significantly reduce the latency for vehicle users (VUs) to access content by pre-caching user-interested content at edge nodes. It is crucial to accurately predict the content that VUs are interested in without exposing their privacy. Traditional federated learning (FL) can protect user privacy by sharing models rather than raw data. However, the training of FL requires frequent model transmission, which can result in significant communication overhead. Additionally, vehicles may leave the road side unit (RSU) coverage area before training is completed, leading to training failures. To address these issues, in this paper, we propose a personalized federated distillation assisted vehicle edge caching strategy. The simulation results demonstrate that the proposed vehicle edge caching strategy has good robustness to variations in vehicle speed, significantly reducing communication overhead.

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.

Desk Editor's Note private letter to a colleague

The paper applies personalized federated distillation to vehicle edge caching to cut communication overhead and manage mobility dropouts, with simulation results that still need tighter validation on the models used.

read the letter

The main takeaway is that this work adapts personalized federated distillation to vehicle edge caching so that content predictions can happen with less model transmission and without as many training failures when vehicles leave an RSU area. It directly targets the overhead and mobility problems that standard federated learning runs into in this setting. The simulations then report that the strategy holds up across speed changes and delivers clear overhead savings. That combination is the concrete step forward here. It takes an existing distillation technique and tunes it to the vehicular constraints rather than inventing a new framework from scratch. The focus on robustness to vehicle speed is a sensible check for the mobility claim. The soft spot is the evidence base. All performance numbers come from simulations, yet the abstract gives no baselines, no error bars, no exact definitions of the metrics, and no description of how trajectories, request patterns, or channel conditions were generated. If those inputs are stylized, the reported gains could be tied to the setup rather than general. Without sensitivity checks or external benchmarks, it is hard to judge how far the robustness extends. This paper is for researchers working on federated methods for edge networks in transportation or mobile IoT. Someone already looking at privacy-preserving caching would see a targeted application and could use the simulation results as a starting point for their own tests. I would send it for peer review. The problem is practical, the proposed fix is reasonable, and referees can push on the experimental details to strengthen the claims.

Referee Report

2 major / 1 minor

Summary. The paper proposes a personalized federated distillation assisted vehicle edge caching strategy to predict content interests of vehicle users (VUs) while protecting privacy. It addresses limitations of traditional federated learning (FL), including high communication overhead from frequent model transmissions and training failures when vehicles leave RSU coverage before completion. The method combines personalization and distillation to reduce transmissions. Simulation results are claimed to show good robustness to vehicle speed variations and significant communication overhead reduction versus standard FL.

Significance. If the simulation results hold under more rigorous validation, the work could advance privacy-preserving edge caching in vehicular networks by lowering communication costs and improving reliability in mobile settings. The integration of personalized distillation offers a plausible direction for adapting FL to dynamic vehicle environments, though the absence of detailed benchmarks limits immediate assessment of novelty relative to existing FL variants in caching.

major comments (2)

[Abstract] Abstract: The central claims of robustness to vehicle speed variations and significant communication overhead reduction rest entirely on simulation results, yet no details are provided on baselines, exact metrics, error bars, data exclusion rules, or sensitivity analysis, preventing verification that the math and data support the stated gains.
[Simulation Results] Simulation setup (assumed results section): The reported performance deltas depend on modeled vehicle trajectories, content request patterns, and network conditions; without explicit external benchmarks, parameter-free derivations, or validation that these models are independent of the fitted strategy, the robustness and overhead claims risk being tied to the specific unvalidated simulation parameters.

minor comments (1)

[Abstract] Abstract: Consider adding one sentence summarizing the core mechanism of the personalized federated distillation component to clarify how it differs from standard FL.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which identify key areas where additional rigor is needed in presenting our simulation results. We address each point below and will revise the manuscript to improve clarity and verifiability.

read point-by-point responses

Referee: [Abstract] Abstract: The central claims of robustness to vehicle speed variations and significant communication overhead reduction rest entirely on simulation results, yet no details are provided on baselines, exact metrics, error bars, data exclusion rules, or sensitivity analysis, preventing verification that the math and data support the stated gains.

Authors: We agree that the current presentation lacks sufficient detail to allow independent verification. In the revised manuscript we will expand both the abstract and the dedicated simulation section to specify the baselines (standard FL without personalization or distillation, and a non-federated caching baseline), the exact metrics (communication overhead in total transmitted parameters per training round and cache-hit ratio), error bars computed over 10 independent runs with distinct random seeds, any data-exclusion rules (vehicles that exit RSU coverage mid-training), and a sensitivity analysis sweeping vehicle speeds from 20 km/h to 120 km/h while holding other parameters fixed. These additions will directly support the robustness and overhead-reduction claims. revision: yes
Referee: [Simulation Results] Simulation setup (assumed results section): The reported performance deltas depend on modeled vehicle trajectories, content request patterns, and network conditions; without explicit external benchmarks, parameter-free derivations, or validation that these models are independent of the fitted strategy, the robustness and overhead claims risk being tied to the specific unvalidated simulation parameters.

Authors: We acknowledge the risk of simulation-specific bias. The revised version will (i) cite and compare against external benchmarks drawn from recent vehicular-caching literature, (ii) supply a parameter-free derivation of the communication-overhead reduction that follows directly from the reduced number of model transmissions enabled by distillation, and (iii) add cross-validation experiments that replace the original trajectory generator with SUMO-derived traces and alter the Zipf content-request exponent, thereby demonstrating that the reported gains persist across modeling choices. Any remaining assumptions will be stated explicitly. revision: partial

Circularity Check

0 steps flagged

No circularity: claims rest on simulation results without self-referential derivations

full rationale

The paper proposes a personalized federated distillation strategy for vehicle edge caching and validates performance via simulations showing robustness to speed variations and reduced overhead. No derivation chain, equations, or load-bearing steps are present in the abstract or described text that reduce by construction to inputs (e.g., no self-definitional parameters, fitted values renamed as predictions, or uniqueness theorems imported via self-citation). The central claims are empirical outcomes from modeled scenarios; absent any quoted reduction of a result to its own fitted components or ansatz, the analysis finds the work self-contained against the specified circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard domain assumptions from federated learning and edge caching literature without introducing new invented entities or many explicitly fitted free parameters in the abstract description.

axioms (2)

domain assumption Sharing model updates instead of raw data protects user privacy in federated learning
Invoked when contrasting traditional FL with the proposed distillation method.
domain assumption Distillation can transfer sufficient knowledge with lower communication cost than full model exchange
Core premise of the proposed strategy for reducing overhead.

pith-pipeline@v0.9.0 · 5444 in / 1254 out tokens · 28027 ms · 2026-05-16T23:35:10.208449+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

24 extracted references · 24 canonical work pages · 1 internal anchor

[1]

Resource allocation for twin maintenance and task processing in vehicular edge computing network,

Y . Xie, Q. Wu, P. Fan, N. Cheng, W. Chen, J. Wang, and K. B. Letaief, “Resource allocation for twin maintenance and task processing in vehicular edge computing network,”IEEE Internet of Things Journal, vol. 12, no. 15, pp. 32 008–32 021, 2025

work page 2025
[2]

Energy consumption minimization in secure multi-antenna uav-assisted mec networks with channel uncertainty,

W. Mao, K. Xiong, Y . Lu, P. Fan, and Z. Ding, “Energy consumption minimization in secure multi-antenna uav-assisted mec networks with channel uncertainty,”IEEE Transactions on Wireless Communications, vol. 22, no. 11, pp. 7185–7200, 2023

work page 2023
[3]

Co- operative edge caching based on elastic federated and multi-agent deep reinforcement learning in next-generation networks,

Q. Wu, W. Wang, P. Fan, Q. Fan, H. Zhu, and K. B. Letaief, “Co- operative edge caching based on elastic federated and multi-agent deep reinforcement learning in next-generation networks,”IEEE Transactions on Network and Service Management, 2024

work page 2024
[4]

Age-upon-decisions minimizing scheduling in internet of things: To be random or to be deterministic?

Y . Dong, Z. Chen, S. Liu, P. Fan, and K. B. Letaief, “Age-upon-decisions minimizing scheduling in internet of things: To be random or to be deterministic?”IEEE Internet of Things Journal, vol. 7, no. 2, pp. 1081– 1097, 2020

work page 2020
[5]

Optimal resource allocation in wireless powered communication networks with user cooperation,

X. Di, K. Xiong, P. Fan, H.-C. Yang, and K. B. Letaief, “Optimal resource allocation in wireless powered communication networks with user cooperation,”IEEE Transactions on Wireless Communications, vol. 16, no. 12, pp. 7936–7949, 2017

work page 2017
[6]

Reconfigurable-intelligent-surface-aided vehicular edge computing: Joint phase-shift optimization and multiuser power allocation,

K. Qi, Q. Wu, P. Fan, N. Cheng, W. Chen, and K. B. Letaief, “Reconfigurable-intelligent-surface-aided vehicular edge computing: Joint phase-shift optimization and multiuser power allocation,”IEEE Internet of Things Journal, vol. 12, no. 1, pp. 764–777, 2025

work page 2025
[7]

Optimum transmission policies for energy harvesting sensor networks powered by a mobile control center,

T. Li, P. Fan, Z. Chen, and K. B. Letaief, “Optimum transmission policies for energy harvesting sensor networks powered by a mobile control center,”IEEE Transactions on Wireless Communications, vol. 15, no. 9, pp. 6132–6145, 2016

work page 2016
[8]

Doppler frequency offsets estimation and diversity reception scheme of high speed railway with multiple antennas on separated carriages,

Y . Yang, P. Fan, and Y . Huang, “Doppler frequency offsets estimation and diversity reception scheme of high speed railway with multiple antennas on separated carriages,” in2012 International Conference on Wireless Communications and Signal Processing (WCSP), 2012, pp. 1– 6

work page 2012
[9]

Adaptive federated learning in resource constrained edge computing systems,

S. Wang, T. Tuor, T. Salonidis, K. K. Leung, C. Makaya, T. He, and K. Chan, “Adaptive federated learning in resource constrained edge computing systems,”IEEE journal on selected areas in communications, vol. 37, no. 6, pp. 1205–1221, 2019

work page 2019
[10]

Global proportional fair scheduling for networks with multiple base stations,

H. Zhou, P. Fan, and J. Li, “Global proportional fair scheduling for networks with multiple base stations,”IEEE Transactions on V ehicular Technology, vol. 60, no. 4, pp. 1867–1879, 2011

work page 2011
[11]

A neighbor-table-based multipath routing in ad hoc networks,

Z. Yao, J. Jiang, P. Fan, Z. Cao, and V . Li, “A neighbor-table-based multipath routing in ad hoc networks,” inThe 57th IEEE Semiannual V ehicular Technology Conference, 2003. VTC 2003-Spring., vol. 3, 2003, pp. 1739–1743 vol.3

work page 2003
[12]

Investigation of the time-offset- based qos support with optical burst switching in wdm networks,

P. Fan, C. Feng, Y . Wang, and N. Ge, “Investigation of the time-offset- based qos support with optical burst switching in wdm networks,” in 2002 IEEE International Conference on Communications. Conference Proceedings. ICC 2002 (Cat. No.02CH37333), vol. 5, 2002, pp. 2682– 2686 vol.5

work page 2002
[13]

Mobility-aware cooperative caching in vehicular edge computing based on asynchronous federated and deep reinforcement learning,

Q. Wu, Y . Zhao, Q. Fan, P. Fan, J. Wang, and C. Zhang, “Mobility-aware cooperative caching in vehicular edge computing based on asynchronous federated and deep reinforcement learning,”IEEE Journal of Selected Topics in Signal Processing, vol. 17, no. 1, pp. 66–81, 2023. 6

work page 2023
[14]

Performance modeling of ieee 802.11 dcf based fair channel access for vehicular-to-roadside communication in a non- saturated state,

Q. Wu and J. Zheng, “Performance modeling of ieee 802.11 dcf based fair channel access for vehicular-to-roadside communication in a non- saturated state,” in2014 IEEE International Conference on Communi- cations (ICC). IEEE, 2014, pp. 2575–2580

work page 2014
[15]

Mobility- aware proactive edge caching for connected vehicles using federated learning,

Z. Yu, J. Hu, G. Min, Z. Zhao, W. Miao, and M. S. Hossain, “Mobility- aware proactive edge caching for connected vehicles using federated learning,”IEEE Transactions on Intelligent Transportation Systems, vol. 22, no. 8, pp. 5341–5351, 2020

work page 2020
[16]

U-net: Convolutional networks for biomedical image segmentation,

O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” inMedical image computing and computer-assisted intervention–MICCAI 2015: 18th international con- ference, Munich, Germany, October 5-9, 2015, proceedings, part III 18. Springer, 2015, pp. 234–241

work page 2015
[17]

DiffWave: A Versatile Diffusion Model for Audio Synthesis

Z. Kong, W. Ping, J. Huang, K. Zhao, and B. Catanzaro, “Dif- fwave: A versatile diffusion model for audio synthesis,”arXiv preprint arXiv:2009.09761, 2020

work page internal anchor Pith review Pith/arXiv arXiv 2009
[18]

Grad- tts: A diffusion probabilistic model for text-to-speech,

V . Popov, I. V ovk, V . Gogoryan, T. Sadekova, and M. Kudinov, “Grad- tts: A diffusion probabilistic model for text-to-speech,” inInternational conference on machine learning. PMLR, 2021, pp. 8599–8608

work page 2021
[19]

Denoising diffusion probabilistic models,

J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in neural information processing systems, vol. 33, 2020

work page 2020
[20]

High- resolution image synthesis with latent diffusion models,

R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High- resolution image synthesis with latent diffusion models,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 10 684–10 695

work page 2022
[21]

Fedcache: A knowledge cache-driven federated learning architecture for personalized edge intelligence,

Z. Wu, S. Sun, Y . Wang, M. Liu, K. Xu, W. Wang, X. Jiang, B. Gao, and J. Lu, “Fedcache: A knowledge cache-driven federated learning architecture for personalized edge intelligence,”IEEE Transactions on Mobile Computing, vol. 23, no. 10, pp. 9368–9382, 2024

work page 2024
[22]

Efficient and robust approxi- mate nearest neighbor search using hierarchical navigable small world graphs,

Y . A. Malkov and D. A. Yashunin, “Efficient and robust approxi- mate nearest neighbor search using hierarchical navigable small world graphs,”IEEE transactions on pattern analysis and machine intelligence, vol. 42, no. 4, pp. 824–836, 2018

work page 2018
[23]

Semantic- aware resource allocation based on deep reinforcement learning for 5g- v2x hetnets,

Z. Shao, Q. Wu, P. Fan, N. Cheng, Q. Fan, and J. Wang, “Semantic- aware resource allocation based on deep reinforcement learning for 5g- v2x hetnets,”IEEE Communications Letters, vol. 28, no. 10, pp. 2452– 2456, 2024

work page 2024
[24]

Context- aware proactive content caching with service differentiation in wireless networks,

S. M ¨uller, O. Atan, M. Van Der Schaar, and A. Klein, “Context- aware proactive content caching with service differentiation in wireless networks,”IEEE Transactions on Wireless Communications, vol. 16, no. 2, pp. 1024–1036, 2016

work page 2016

[1] [1]

Resource allocation for twin maintenance and task processing in vehicular edge computing network,

Y . Xie, Q. Wu, P. Fan, N. Cheng, W. Chen, J. Wang, and K. B. Letaief, “Resource allocation for twin maintenance and task processing in vehicular edge computing network,”IEEE Internet of Things Journal, vol. 12, no. 15, pp. 32 008–32 021, 2025

work page 2025

[2] [2]

Energy consumption minimization in secure multi-antenna uav-assisted mec networks with channel uncertainty,

W. Mao, K. Xiong, Y . Lu, P. Fan, and Z. Ding, “Energy consumption minimization in secure multi-antenna uav-assisted mec networks with channel uncertainty,”IEEE Transactions on Wireless Communications, vol. 22, no. 11, pp. 7185–7200, 2023

work page 2023

[3] [3]

Co- operative edge caching based on elastic federated and multi-agent deep reinforcement learning in next-generation networks,

Q. Wu, W. Wang, P. Fan, Q. Fan, H. Zhu, and K. B. Letaief, “Co- operative edge caching based on elastic federated and multi-agent deep reinforcement learning in next-generation networks,”IEEE Transactions on Network and Service Management, 2024

work page 2024

[4] [4]

Age-upon-decisions minimizing scheduling in internet of things: To be random or to be deterministic?

Y . Dong, Z. Chen, S. Liu, P. Fan, and K. B. Letaief, “Age-upon-decisions minimizing scheduling in internet of things: To be random or to be deterministic?”IEEE Internet of Things Journal, vol. 7, no. 2, pp. 1081– 1097, 2020

work page 2020

[5] [5]

Optimal resource allocation in wireless powered communication networks with user cooperation,

X. Di, K. Xiong, P. Fan, H.-C. Yang, and K. B. Letaief, “Optimal resource allocation in wireless powered communication networks with user cooperation,”IEEE Transactions on Wireless Communications, vol. 16, no. 12, pp. 7936–7949, 2017

work page 2017

[6] [6]

Reconfigurable-intelligent-surface-aided vehicular edge computing: Joint phase-shift optimization and multiuser power allocation,

K. Qi, Q. Wu, P. Fan, N. Cheng, W. Chen, and K. B. Letaief, “Reconfigurable-intelligent-surface-aided vehicular edge computing: Joint phase-shift optimization and multiuser power allocation,”IEEE Internet of Things Journal, vol. 12, no. 1, pp. 764–777, 2025

work page 2025

[7] [7]

Optimum transmission policies for energy harvesting sensor networks powered by a mobile control center,

T. Li, P. Fan, Z. Chen, and K. B. Letaief, “Optimum transmission policies for energy harvesting sensor networks powered by a mobile control center,”IEEE Transactions on Wireless Communications, vol. 15, no. 9, pp. 6132–6145, 2016

work page 2016

[8] [8]

Doppler frequency offsets estimation and diversity reception scheme of high speed railway with multiple antennas on separated carriages,

Y . Yang, P. Fan, and Y . Huang, “Doppler frequency offsets estimation and diversity reception scheme of high speed railway with multiple antennas on separated carriages,” in2012 International Conference on Wireless Communications and Signal Processing (WCSP), 2012, pp. 1– 6

work page 2012

[9] [9]

Adaptive federated learning in resource constrained edge computing systems,

S. Wang, T. Tuor, T. Salonidis, K. K. Leung, C. Makaya, T. He, and K. Chan, “Adaptive federated learning in resource constrained edge computing systems,”IEEE journal on selected areas in communications, vol. 37, no. 6, pp. 1205–1221, 2019

work page 2019

[10] [10]

Global proportional fair scheduling for networks with multiple base stations,

H. Zhou, P. Fan, and J. Li, “Global proportional fair scheduling for networks with multiple base stations,”IEEE Transactions on V ehicular Technology, vol. 60, no. 4, pp. 1867–1879, 2011

work page 2011

[11] [11]

A neighbor-table-based multipath routing in ad hoc networks,

Z. Yao, J. Jiang, P. Fan, Z. Cao, and V . Li, “A neighbor-table-based multipath routing in ad hoc networks,” inThe 57th IEEE Semiannual V ehicular Technology Conference, 2003. VTC 2003-Spring., vol. 3, 2003, pp. 1739–1743 vol.3

work page 2003

[12] [12]

Investigation of the time-offset- based qos support with optical burst switching in wdm networks,

P. Fan, C. Feng, Y . Wang, and N. Ge, “Investigation of the time-offset- based qos support with optical burst switching in wdm networks,” in 2002 IEEE International Conference on Communications. Conference Proceedings. ICC 2002 (Cat. No.02CH37333), vol. 5, 2002, pp. 2682– 2686 vol.5

work page 2002

[13] [13]

Mobility-aware cooperative caching in vehicular edge computing based on asynchronous federated and deep reinforcement learning,

Q. Wu, Y . Zhao, Q. Fan, P. Fan, J. Wang, and C. Zhang, “Mobility-aware cooperative caching in vehicular edge computing based on asynchronous federated and deep reinforcement learning,”IEEE Journal of Selected Topics in Signal Processing, vol. 17, no. 1, pp. 66–81, 2023. 6

work page 2023

[14] [14]

Performance modeling of ieee 802.11 dcf based fair channel access for vehicular-to-roadside communication in a non- saturated state,

Q. Wu and J. Zheng, “Performance modeling of ieee 802.11 dcf based fair channel access for vehicular-to-roadside communication in a non- saturated state,” in2014 IEEE International Conference on Communi- cations (ICC). IEEE, 2014, pp. 2575–2580

work page 2014

[15] [15]

Mobility- aware proactive edge caching for connected vehicles using federated learning,

Z. Yu, J. Hu, G. Min, Z. Zhao, W. Miao, and M. S. Hossain, “Mobility- aware proactive edge caching for connected vehicles using federated learning,”IEEE Transactions on Intelligent Transportation Systems, vol. 22, no. 8, pp. 5341–5351, 2020

work page 2020

[16] [16]

U-net: Convolutional networks for biomedical image segmentation,

O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” inMedical image computing and computer-assisted intervention–MICCAI 2015: 18th international con- ference, Munich, Germany, October 5-9, 2015, proceedings, part III 18. Springer, 2015, pp. 234–241

work page 2015

[17] [17]

DiffWave: A Versatile Diffusion Model for Audio Synthesis

Z. Kong, W. Ping, J. Huang, K. Zhao, and B. Catanzaro, “Dif- fwave: A versatile diffusion model for audio synthesis,”arXiv preprint arXiv:2009.09761, 2020

work page internal anchor Pith review Pith/arXiv arXiv 2009

[18] [18]

Grad- tts: A diffusion probabilistic model for text-to-speech,

V . Popov, I. V ovk, V . Gogoryan, T. Sadekova, and M. Kudinov, “Grad- tts: A diffusion probabilistic model for text-to-speech,” inInternational conference on machine learning. PMLR, 2021, pp. 8599–8608

work page 2021

[19] [19]

Denoising diffusion probabilistic models,

J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in neural information processing systems, vol. 33, 2020

work page 2020

[20] [20]

High- resolution image synthesis with latent diffusion models,

R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High- resolution image synthesis with latent diffusion models,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 10 684–10 695

work page 2022

[21] [21]

Fedcache: A knowledge cache-driven federated learning architecture for personalized edge intelligence,

Z. Wu, S. Sun, Y . Wang, M. Liu, K. Xu, W. Wang, X. Jiang, B. Gao, and J. Lu, “Fedcache: A knowledge cache-driven federated learning architecture for personalized edge intelligence,”IEEE Transactions on Mobile Computing, vol. 23, no. 10, pp. 9368–9382, 2024

work page 2024

[22] [22]

Efficient and robust approxi- mate nearest neighbor search using hierarchical navigable small world graphs,

Y . A. Malkov and D. A. Yashunin, “Efficient and robust approxi- mate nearest neighbor search using hierarchical navigable small world graphs,”IEEE transactions on pattern analysis and machine intelligence, vol. 42, no. 4, pp. 824–836, 2018

work page 2018

[23] [23]

Semantic- aware resource allocation based on deep reinforcement learning for 5g- v2x hetnets,

Z. Shao, Q. Wu, P. Fan, N. Cheng, Q. Fan, and J. Wang, “Semantic- aware resource allocation based on deep reinforcement learning for 5g- v2x hetnets,”IEEE Communications Letters, vol. 28, no. 10, pp. 2452– 2456, 2024

work page 2024

[24] [24]

Context- aware proactive content caching with service differentiation in wireless networks,

S. M ¨uller, O. Atan, M. Van Der Schaar, and A. Klein, “Context- aware proactive content caching with service differentiation in wireless networks,”IEEE Transactions on Wireless Communications, vol. 16, no. 2, pp. 1024–1036, 2016

work page 2016