pith. sign in

arxiv: 2604.10931 · v1 · submitted 2026-04-13 · 📡 eess.SP

Reliable Online Resource Allocation for Multi-User Semantic Communications: A Constraint Bayesian Optimization Approach

Huawei Hou , Suzhi Bi , Xian Li , Haixia Zhang , Zhi Quan This is my paper

Pith reviewed 2026-05-10 16:12 UTC · model grok-4.3

classification 📡 eess.SP
keywords semantic communicationresource allocationBayesian optimizationGaussian processmulti-user edge computingcompression ratioreconstruction qualitytransmission latency
0
0 comments X

The pith

Bayesian optimization using Gaussian process surrogates lets edge servers dynamically set compression ratios and rates for multiple semantic users to meet quality constraints while cutting latency.

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

Semantic communication systems compress source data into features via neural codecs and offload reconstruction to an edge server, but the black-box nature of the codecs plus channel noise makes it hard to guarantee quality while minimizing latency for many users at once. The paper formulates a joint optimization that balances overall reconstruction performance against total transmission time subject to per-user minimum quality floors. It solves the problem online with a Bayesian optimization routine that builds Gaussian process models relating reconstruction quality to compression ratio and SNR, then uses an acquisition function to pick feasible compression ratios. Simulations on high-resolution video frames show the method reaches 98 percent constraint satisfaction and cuts latency more than 45 percent versus fixed-ratio baselines.

Core claim

A constraint Bayesian optimization algorithm maintains Gaussian process surrogates that map each user's compression ratio and observed SNR to expected reconstruction quality; at each step the acquisition function selects the compression ratio and transmission rate pair that satisfies all quality constraints while minimizing the combined latency objective.

What carries the argument

Gaussian process surrogate models that predict reconstruction quality from compression ratio and SNR, paired with an acquisition function that enforces per-user quality constraints inside the Bayesian optimization loop.

If this is right

  • Edge servers can jointly tune compression ratios and transmission rates for many users without exhaustive search over the discrete CR space.
  • Transmission latency drops more than 45 percent relative to any fixed compression-ratio policy while still meeting quality floors.
  • The same GP-plus-acquisition structure can be re-used when channel conditions or user sets change, because the surrogate is updated from new observations.
  • Semantic codecs become practical for multi-user edge workloads once the mapping from CR and SNR to quality is learned on-line rather than assumed known.

Where Pith is reading between the lines

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

  • The same surrogate-modeling approach could be applied to other black-box performance surfaces in wireless systems such as power allocation or modulation selection.
  • If the GP begins to drift, a simple periodic re-sampling schedule would restore prediction accuracy without changing the overall algorithm.
  • Integrating the method with user scheduling or power control would let the edge server optimize an even larger joint space.
  • Hardware-in-the-loop experiments with actual neural codecs and real wireless channels would be the next direct test of the 98 percent satisfaction figure.

Load-bearing premise

The Gaussian process models built from initial observations will keep predicting reconstruction quality accurately for new compression ratios and SNRs without drift or excessive computation during real-time operation.

What would settle it

A live test in which measured reconstruction quality for a chosen compression ratio and instantaneous SNR falls well below the GP prediction for several users, driving the fraction of satisfied quality constraints below 90 percent.

Figures

Figures reproduced from arXiv: 2604.10931 by Haixia Zhang, Huawei Hou, Suzhi Bi, Xian Li, Zhi Quan.

Figure 1
Figure 1. Figure 1: Block diagram of the proposed multi-user edge co-inference system with online resource allocation. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: An illustration of TDMA-based transmission scheme. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Proposed BO-based multi-user online CR selection method. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Results of JSCC and oracle network on 4 datasets. [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: PSNR Performance and constraint violation rate of the proposed and benchmark methods across four users, evaluated on the first test dataset. TABLE III: Comprehensive Performance Comparison Across Methods (Averaged Over Three Testing Datasets). Best results are highlighted in bold and second-best results are underlined. Method Constraint Satisfaction (%) ↑ PSNR (dB) ↑ Latency (ms) ↓ Obj. User1 User2 User3 U… view at source ↗
Figure 6
Figure 6. Figure 6: Performance as time evolves. Results are from the first test dataset and smoothed over 5-time-slot windows. [34, 34, 26, 26] [34, 34, 27, 27] [35, 35, 27, 27] [35, 35, 28, 28] Constraints 70 75 80 85 90 95 100 Portion (%) 98.00 98.33 96.00 98.33 96.00 97.00 94.67 92.67 99.00 98.33 93.66 94.33 98.33 97.33 95.00 89.33 Portion of Meeting PSNR Requirements under Different Constraints User 1 User 2 User 3 User … view at source ↗
Figure 7
Figure 7. Figure 7: Portion of constraint satisfaction per user under differ [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Influence of α. 1 4 8 12 16 20 Users (Number of Models) 36 38 40 42 44 46 48 50 52 Update Runtime (ms) CPU Update (Left Y) GPU Update (Left Y) CPU Inference (Right Y) GPU Inference (Right Y) 0 2 4 6 8 10 Inference Runtime (ms) [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Model update and inference runtime under different users. VI. CONCLUSIONS AND FUTURE WORK This paper investigated multi-user semantic communications focused on joint CR and transmission rate optimization. To address the challenge of unobservable reconstruction quality at the receiver, we developed a transmitter-side oracle network for quality prediction and a BO-based CR selection method. By employing GPs … view at source ↗
read the original abstract

Semantic communication has been increasingly integrated into edge computing systems for reconstruction tasks, owing to its advantages in source compression, robustness to channel noise, and task execution efficiency. However, the black-box nature of neural-network (NN)-based semantic codecs, together with the noisy transmission of semantic features, makes it difficult to allocate transmission resources and guarantee reconstruction quality for multiple users. In this paper, we propose a reliable online resource allocation framework for a semantic-driven multi-user edge computing system, where multiple users encode source information into semantic features and offload reconstruction to an edge server. We formulate a multi-user resource optimization problem whose objective jointly accounts for system-wide reconstruction performance and transmission latency, under constraints that guarantee each user's minimum reconstruction quality. To solve this problem, we develop a Bayesian optimization (BO)-based online algorithm that enables flexible control of the user-side semantic compression ratio (CR) and allocation of transmission rates. The edge server jointly determines each user's CR and transmission rate by exploiting Gaussian-process (GP) models that capture the relationship between reconstruction performance, signal-to-noise ratio (SNR), and CR, and by employing an acquisition function to select CRs that satisfy the performance quality constraints while maximizing the objective. Simulation results on high-resolution video-frame reconstruction datasets demonstrate that the proposed method selects near-optimal CRs via the GP surrogate and acquisition function, achieving a 98.03% constraint-satisfaction rate and reducing transmission latency by more than 45% compared with fixed-CR schemes.

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

2 major / 2 minor

Summary. The paper proposes a constraint Bayesian optimization (BO) framework for online multi-user resource allocation in semantic communication systems. Gaussian process (GP) surrogates model the mapping from compression ratio (CR), SNR, and reconstruction quality; an acquisition function then selects per-user CRs and transmission rates to minimize system latency subject to per-user quality constraints. Simulations on high-resolution video-frame datasets report that the method achieves a 98.03% constraint-satisfaction rate and more than 45% latency reduction relative to fixed-CR baselines.

Significance. If the GP surrogates remain accurate for the CR/SNR points selected online, the approach provides a practical, data-driven method for handling black-box neural semantic codecs in edge systems while delivering explicit quality guarantees and substantial latency gains. The concrete simulation metrics on video reconstruction tasks constitute a strength, demonstrating measurable improvement over simple baselines.

major comments (2)
  1. [Abstract] Abstract and simulation results: the central claim that the BO loop reliably satisfies quality constraints (98.03% rate) depends on the GP surrogate correctly predicting reconstruction quality at the CR/SNR pairs it selects. No hold-out validation or online-update experiment is described that measures ground-truth NN reconstruction error on exactly those operating points; if the reported rate is computed from GP predictions rather than actual NN outputs, the result is circular and does not establish feasibility under surrogate mismatch.
  2. [Simulation results] The manuscript provides no information on GP training-data volume, baseline implementation details, statistical variance across runs, or sensitivity of the latency-reduction result to GP kernel hyperparameters and acquisition-function parameters. These omissions make it impossible to judge whether the >45% latency improvement is robust or an artifact of particular simulation settings.
minor comments (2)
  1. Notation for the acquisition function and the precise form of the quality constraint (e.g., how the minimum reconstruction quality threshold is encoded) could be stated more explicitly to aid reproducibility.
  2. The paper would benefit from a short discussion of computational overhead of the GP updates in real-time operation, even if only order-of-magnitude estimates are given.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and insightful comments on our manuscript. We address each major comment below and will incorporate revisions to improve clarity, reproducibility, and validation of the results.

read point-by-point responses

  1. Referee: [Abstract] Abstract and simulation results: the central claim that the BO loop reliably satisfies quality constraints (98.03% rate) depends on the GP surrogate correctly predicting reconstruction quality at the CR/SNR pairs it selects. No hold-out validation or online-update experiment is described that measures ground-truth NN reconstruction error on exactly those operating points; if the reported rate is computed from GP predictions rather than actual NN outputs, the result is circular and does not establish feasibility under surrogate mismatch.

    Authors: We agree that explicit validation against ground-truth NN outputs is essential to avoid any appearance of circularity. In the simulations, the 98.03% constraint-satisfaction rate is computed from the actual neural-network reconstruction quality (PSNR/SSIM) evaluated on the CR/SNR pairs selected by the BO procedure, not from GP predictions alone. The GP surrogate is used solely for guiding the online selection, while final metrics reflect true decoder performance. However, the manuscript does not describe the hold-out validation or online-update experiments in sufficient detail. We will add a dedicated subsection in the simulation results that (i) specifies the hold-out test points, (ii) reports ground-truth NN errors on the exact operating points chosen online, and (iii) includes an online-update experiment showing how the GP is retrained with new observations and how constraint satisfaction evolves. This revision will directly address the concern and strengthen the feasibility claim. revision: yes

  2. Referee: [Simulation results] The manuscript provides no information on GP training-data volume, baseline implementation details, statistical variance across runs, or sensitivity of the latency-reduction result to GP kernel hyperparameters and acquisition-function parameters. These omissions make it impossible to judge whether the >45% latency improvement is robust or an artifact of particular simulation settings.

    Authors: We acknowledge that the current manuscript lacks these critical implementation and robustness details, which limits reproducibility and assessment of result stability. We will expand the simulation section to include: (1) the exact volume and composition of the dataset used to train the GP surrogates (including how many video frames and SNR/CR combinations were sampled), (2) precise descriptions and parameter settings for all baseline schemes, (3) mean and standard deviation of latency and constraint-satisfaction metrics over at least 20 independent runs with different random seeds, and (4) sensitivity analysis (tables or figures) varying the GP kernel length-scale, variance, and acquisition-function hyperparameters (e.g., exploration-exploitation trade-off). These additions will allow readers to evaluate whether the >45% latency reduction is robust. revision: yes

Circularity Check

0 steps flagged

No significant circularity; standard BO surrogate applied to black-box function with independent simulation validation

full rationale

The paper formulates a resource allocation problem and solves it via Bayesian optimization using Gaussian process surrogates to approximate the NN-based reconstruction quality as a function of CR and SNR. This is a standard surrogate modeling technique for black-box objectives, not a self-definition or renaming of the target result. The GP is fitted to prior observations and used to guide selection via the acquisition function, but the reported metrics (98.03% constraint-satisfaction rate and >45% latency reduction) are obtained from simulations on high-resolution video-frame datasets that evaluate the actual system performance at the selected operating points. No equation reduces the claimed outcomes to fitted parameters by construction, and no load-bearing step relies on self-citation chains or imported uniqueness theorems. The derivation chain remains self-contained against external simulation benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The approach rests on standard assumptions of Gaussian process regression for surrogate modeling and Bayesian optimization for constrained acquisition; free parameters are the GP kernel hyperparameters and acquisition function trade-off weights, which are fitted during operation.

free parameters (2)
  • GP kernel hyperparameters
    Fitted to observed CR-SNR-reconstruction data to build the surrogate model.
  • Acquisition function parameters
    Control exploration-exploitation balance when selecting next CR values under constraints.
axioms (1)
  • domain assumption Gaussian processes provide a suitable probabilistic model for the unknown mapping from CR and SNR to reconstruction quality
    Invoked when the edge server uses GP models to predict performance and guide optimization.

pith-pipeline@v0.9.0 · 5573 in / 1168 out tokens · 34426 ms · 2026-05-10T16:12:47.153857+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

38 extracted references · 38 canonical work pages

  1. [1]

    Semantic communication: a survey on research landscape, challenges, and future directions,

    T. M. Getu, G. Kaddoum, and M. Bennis, “Semantic communication: a survey on research landscape, challenges, and future directions,”Proc. IEEE, vol. 112, no. 11, pp. 1649–1685, Nov. 2024

    work page 2024

  2. [2]

    A generalized semantic communication system: from sources to channels,

    Z. Qin, F. Gao, B. Lin, X. Tao, G. Liu, and C. Pan, “A generalized semantic communication system: from sources to channels,”IEEE Wireless Commun., vol. 30, no. 3, pp. 18–26, June 2023

    work page 2023

  3. [3]

    Joint source–channel coding: fundamentals and recent progress in practical designs,

    D. G ¨und¨uz, M. A. Wigger, T.-Y . Tung, P. Zhang, and Y . Xiao, “Joint source–channel coding: fundamentals and recent progress in practical designs,”Proc. IEEE, vol. 113, no. 9, pp. 888-919, Sept. 2025

    work page 2025

  4. [4]

    Deep joint source- channel coding for wireless image transmission,

    E. Bourtsoulatze, D. Burth Kurka, and D. G ¨und¨uz, “Deep joint source- channel coding for wireless image transmission,”IEEE Trans. Cogn. Commun. Netw., vol. 5, no. 3, pp. 567–579, Sept. 2019

    work page 2019

  5. [5]

    Task- oriented communications for 6G: vision, principles, and technologies,

    Y . Shi, Y . Zhou, D. Wen, Y . Wu, C. Jiang, and K. B. Letaief, “Task- oriented communications for 6G: vision, principles, and technologies,” IEEE Wireless Commun., vol. 30, no. 3, pp. 78–85, June 2023

    work page 2023

  6. [6]

    Learning task-oriented communication for edge inference: an information bottleneck approach,

    J. Shao, Y . Mao, and J. Zhang, “Learning task-oriented communication for edge inference: an information bottleneck approach,”IEEE J. Sel. Areas Commun., vol. 40, no. 1, pp. 197-211, Jan. 2022

    work page 2022

  7. [7]

    DeepJSCC-f: deep joint source-channel coding of images with feedback,

    D. B. Kurka and D. G ¨und¨uz, “DeepJSCC-f: deep joint source-channel coding of images with feedback,”IEEE J. Sel. Areas Inf. Theory, vol. 1, no. 1, pp. 178–193, May 2020

    work page 2020

  8. [8]

    Deep joint source-channel coding for wireless image transmission with adaptive rate control,

    M. Yang and H.-S. Kim, “Deep joint source-channel coding for wireless image transmission with adaptive rate control,” inProc. IEEE Int. Conf. Acoust., Speech Signal Process. (ICASSP), Singapore, May 2022, pp. 5193–5197

    work page 2022

  9. [9]

    Scalable multi-task edge sensing via task-oriented joint information gathering and broadcast,

    H. Hou, S. Bi, X. Li, S. Wang, L. Qian, and Z. Quan, “Scalable multi-task edge sensing via task-oriented joint information gathering and broadcast,”IEEE Trans. Wireless Commun., vol. 24, no. 11, pp. 9613–9628, Nov. 2025

    work page 2025

  10. [10]

    Compression before fusion: broadcast semantic communication system for heterogeneous tasks,

    M. Gong, S. Wang, F. Ye, and S. Bi, “Compression before fusion: broadcast semantic communication system for heterogeneous tasks,” IEEE Trans. Wireless Commun., vol. 23, no. 12, pp. 19428-19443, Dec. 2024

    work page 2024

  11. [11]

    DD-JSCC: dynamic deep joint source-channel coding for semantic communications,

    A. D. Raha, A. Adhikary, M. Gain, Y . Park, W. Saad, and C. S. Hong, “DD-JSCC: dynamic deep joint source-channel coding for semantic communications,” inProc. IEEE Int. Conf. Commun. (ICC), Montreal, QC, Canada, June 2025, pp. 3754–3759

    work page 2025

  12. [12]

    Deep joint source-channel coding for wireless image transmission with entropy-aware adaptive rate control,

    W. Chen, Y . Chen, Q. Yang, C. Huang, Q. Wang, and Z. Zhang, “Deep joint source-channel coding for wireless image transmission with entropy-aware adaptive rate control,” inProc. IEEE Global Com- mun. Conf. (GLOBECOM), Kuala Lumpur, Malaysia, Dec. 2023, pp. 2239–2244

    work page 2023

  13. [13]

    Predictive and adaptive deep coding for wireless image transmission in semantic communication,

    W. Zhang, H. Zhang, H. Ma, H. Shao, N. Wang, and V . C. M. Leung, “Predictive and adaptive deep coding for wireless image transmission in semantic communication,”IEEE Trans. Wireless Commun., vol. 22, no. 8, pp. 5486–5501, Aug. 2023

    work page 2023

  14. [14]

    Performance optimization for semantic communications: an attention- based reinforcement learning approach,

    Y . Wang, M. Chen, T. Luo, W. Saad, D. Niyato, and H. V . Poor, “Performance optimization for semantic communications: an attention- based reinforcement learning approach,”IEEE J. Sel. Areas Commun., vol. 40, no. 9, pp. 2598–2613, Sept. 2022

    work page 2022

  15. [15]

    Semantic communications for image recovery and classification via deep joint source and channel coding,

    Z. Lyu, G. Zhu, J. Xu, B. Ai, and S. Cui, “Semantic communications for image recovery and classification via deep joint source and channel coding,”IEEE Trans. Wireless Commun., vol. 23, no. 8, pp. 8388–8404, Aug. 2024

    work page 2024

  16. [16]

    SCAN: semantic communication with adaptive channel feedback,

    G. Zhang, Q. Hu, Y . Cai, and G. Yu, “SCAN: semantic communication with adaptive channel feedback,”IEEE Trans. Cogn. Commun. Netw., vol. 10, no. 5, pp. 1759–1773, Oct. 2024

    work page 2024

  17. [17]

    Modeling and performance analysis for semantic communications based on empirical results,

    S. Ma, B. Shen, C. Zhang, Y . Wu, H. Li, and S. Li, “Modeling and performance analysis for semantic communications based on empirical results,”IEEE Trans. Commun., vol. 73, no. 11, pp. 11078-11092, Nov. 2025

    work page 2025

  18. [18]

    Toward in- telligent resource allocation on task-oriented semantic communication,

    H. Zhang, H. Wang, Y . Li, K. Long, and V . C. M. Leung, “Toward in- telligent resource allocation on task-oriented semantic communication,” IEEE Wireless Commun., vol. 30, no. 3, pp. 70–77, June 2023

    work page 2023

  19. [19]

    Transferable de- ployment of semantic edge inference systems via unsupervised domain adaption,

    W. Jiao, S. Bi, X. Li, C. Guo, H. Chen, and Z. Quan, “Transferable de- ployment of semantic edge inference systems via unsupervised domain adaption,”IEEE Internet Things J., vol. 12, no. 14, pp. 27573-27587, July 2025

    work page 2025

  20. [20]

    Recent advances in Bayesian optimization,

    X. Wang, Y . Jin, S. Schmitt, and M. Olhofer, “Recent advances in Bayesian optimization,”ACM Comput. Surv., vol. 55, no. 287, pp. 1–36, July 2023

    work page 2023

  21. [21]

    Bayesian optimization with inequality constraints,

    J. Gardner, M. Kusner, Z. Xu, K. Weinberger, and J. Cunningham, “Bayesian optimization with inequality constraints,” inProc. Int. Conf. Mach. Learn. (ICML), June 2014, vol. 32, pp. 937–945

    work page 2014

  22. [22]

    A Bayesian optimization approach for online system configuration of edge video analytics,

    H. Tang, S. Bi, S. Wang, X. Lin, Y . Gu, and Z. Quan, “A Bayesian optimization approach for online system configuration of edge video analytics,” inProc. IEEE Int. Conf. Wireless Commun. Signal Process. (WCSP), Hefei, China, Oct. 2024, pp. 807–812

    work page 2024

  23. [23]

    Task-oriented computation offload- ing for edge inference: an integrated Bayesian optimization and deep reinforcement learning framework,

    X. Li, S. Bi, and Y .-J. A. Zhang, “Task-oriented computation offload- ing for edge inference: an integrated Bayesian optimization and deep reinforcement learning framework,”IEEE Trans. Mobile Comput., doi: 10.1109/TMC.2026.3672623

    work page doi:10.1109/tmc.2026.3672623 2026

  24. [24]

    Capacity optimizing resource allocation in joint source-channel coding systems with QoS constraints,

    K. Chi, Q. Yang, Z. Yang, Y . Duan, and Z. Zhang, “Capacity optimizing resource allocation in joint source-channel coding systems with QoS constraints,”IEEE Trans. Commun., vol. 73, no. 6, pp. 4198–4212, June 2025

    work page 2025

  25. [25]

    Adaptable semantic compression and resource allocation for task-oriented communications,

    C. Liu, C. Guo, Y . Yang, and N. Jiang, “Adaptable semantic compression and resource allocation for task-oriented communications,”IEEE Trans. Cogn. Commun. Netw., vol. 10, no. 3, pp. 769–782, June 2024

    work page 2024

  26. [26]

    Feature importance-aware task-oriented semantic transmission and optimiza- tion,

    Y . Wang, S. Han, X. Xu, H. Liang, R. Meng, and C. Dong, “Feature importance-aware task-oriented semantic transmission and optimiza- tion,”IEEE Trans. Cogn. Commun. Netw., vol. 10, no. 4, pp. 1175–1189, Aug. 2024

    work page 2024

  27. [27]

    Energy efficient semantic communication over wireless networks with rate splitting,

    Z. Yang, M. Chen, Z. Zhang, and C. Huang, “Energy efficient semantic communication over wireless networks with rate splitting,”IEEE J. Sel. Areas Commun., vol. 41, no. 5, pp. 1484–1495, May 2023

    work page 2023

  28. [28]

    Deep reinforced feature compression and channel equalization for semantic communications,

    J. Seon, S. Lee, S. H. Kim, Y . G. Sun, H. Seo, and D. I. Kim, “Deep reinforced feature compression and channel equalization for semantic communications,”IEEE Trans. Cogn. Commun. Netw., vol. 11, no. 6, pp. 3655-3668, Dec. 2025

    work page 2025

  29. [29]

    Compression ratio learning and semantic communications for video imaging,

    B. Zhang, Z. Qin, and G. Y . Li, “Compression ratio learning and semantic communications for video imaging,”IEEE J. Sel. Topics Signal Process., vol. 18, no. 3, pp. 312–324, Apr. 2024

    work page 2024

  30. [30]

    Task-oriented communication for multidevice cooperative edge inference,

    J. Shao, Y . Mao, and J. Zhang, “Task-oriented communication for multidevice cooperative edge inference,”IEEE Trans. Wirel. Commun., vol. 22, no. 1, pp. 73-87, Jan. 2023

    work page 2023

  31. [31]

    Enhancing information freshness via knowledge graph-aided semantic communication,

    J. Chen, S. Yang, T. -T. Chan, and H. Pan, “Enhancing information freshness via knowledge graph-aided semantic communication,” inProc. IEEE Int. Conf. Inf., Commun. Netw. (ICICN), Xi’an, China, Aug. 2023, pp. 155-160

    work page 2023

  32. [32]

    DeepJSCC- Q: constellation constrained deep joint source-channel coding,

    T. -Y . Tung, D. B. Kurka, M. Jankowski, and D. G ¨und¨uz, “DeepJSCC- Q: constellation constrained deep joint source-channel coding,”IEEE J. Sel. Areas Inf. Theory, vol. 3, no. 4, pp. 720-731, Dec. 2022

    work page 2022

  33. [33]

    Joint source-channel coding for channel-adaptive digital semantic communications,

    J. Park, Y . Oh, S. Kim, and Y . -S. Jeon, “Joint source-channel coding for channel-adaptive digital semantic communications,”IEEE Trans. Cogn. Commun. Netw., vol. 11, no. 1, pp. 75-89, Feb. 2025

    work page 2025

  34. [34]

    Digital semantic device-edge co-inference with task-oriented ARQ,

    X. Li, S. Bi, S. Wang, X. Li, and Y . -J. A. Zhang, “Digital semantic device-edge co-inference with task-oriented ARQ,”IEEE Trans. V eh. Technol., vol. 73, no. 9, pp. 13986-13990, Sept. 2024

    work page 2024

  35. [35]

    BDD100K: a diverse driving dataset for heterogeneous multitask learning,

    F. Yuet al., “BDD100K: a diverse driving dataset for heterogeneous multitask learning,” inProc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), Seattle, W A, USA, June 2020, pp. 2633–2642

    work page 2020

  36. [36]

    Fast and robust UA V to UA V detection and tracking from video,

    J. Li, D. H. Ye, M. Kolsch, J. P. Wachs, and C. A. Bouman, “Fast and robust UA V to UA V detection and tracking from video,”IEEE Trans. Emerg. Topics Comput., vol. 10, no. 3, pp. 1519–1531, Jul.–Sept. 2022

    work page 2022

  37. [37]

    The unmanned aerial vehicle benchmark: object detection and tracking,

    D. Duet al., “The unmanned aerial vehicle benchmark: object detection and tracking,” inProc. Eur . Conf. Comput. Vis. (ECCV), Munich, Germany, Sept. 2018, pp. 370–386

    work page 2018

  38. [38]

    Edge video analytics with adaptive information gathering: a deep reinforcement learning approach,

    S. Wang, S. Bi, and Y . J. Zhang, “Edge video analytics with adaptive information gathering: a deep reinforcement learning approach,”IEEE Trans. Wirel. Commun., vol. 22, no. 9, pp. 5800-5813, Sept. 2023

    work page 2023