arxiv: 2604.09255 · v1 · submitted 2026-04-10 · 📡 eess.SP

Recognition: unknown

Semantic Feature Multiple Access Empowered Integrated Learning and Communication Networks

Jiaxiang Wang , Zhouxiang Zhao , Yahao Ding , Zhijin Qin , Zhaohui Yang , Mingzhe Chen , Mohammad Shikh-Bahaei

Authors on Pith no claims yet

Pith reviewed 2026-05-10 17:01 UTC · model grok-4.3

classification 📡 eess.SP

keywords semantic feature multiple accessintegrated learning and communicationSwin Transformer transceiveruser pairingsum-rate maximizationimage transmissionjoint source-channel coding

0 comments

The pith

Semantic feature multiple access with similarity-conditioned transceivers improves image quality and network sum rates over standard baselines.

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

The paper develops an integrated learning and communication approach where users transmit learned semantic features of their data over shared wireless resources instead of raw signals. By conditioning how these features are fused on the semantic similarity between user pairs and modeling the resulting interference as a function of power and compression, the system converts a difficult mixed-integer optimization into a tractable form. A three-block alternating algorithm then jointly handles user pairing, power allocation, bandwidth, and compression ratios while respecting quality, delay, and energy limits. Simulations show higher reconstructed image quality measured by peak signal-to-noise ratio and structural similarity, plus higher overall data rates, compared with conventional joint source-channel coding and separation schemes.

Core claim

Semantic feature multiple access lets paired users superpose their learned feature representations on shared time-frequency resources, where interference depends jointly on the user pair, transmit power, and compression ratio. The proposed similarity-conditioned SFMA transceiver employs a Swin Transformer with a dual-conditioned similarity modulator that gates cross-user feature fusion according to inter-user semantic similarity. This interference is characterized by a bivariate logistic function of power and compression ratio, which allows formulation of a sum-rate maximization problem subject to per-user distortion, latency, energy, power, and bandwidth constraints. The problem is solvedby

What carries the argument

The dual-conditioned similarity modulator that gates cross-user feature fusion according to inter-user semantic similarity, together with the bivariate logistic function that models pair-dependent interference as a function of transmit power and compression ratio, enabling the three-block alternating optimization.

If this is right

Higher peak signal-to-noise ratio and multi-scale structural similarity for image reconstruction than deep joint source-channel coding and separation baselines.
Larger network sum rates than conventional multiple access schemes under the same constraints.
Joint optimization of binary pairing decisions with continuous power, bandwidth, and compression variables through alternating blocks for compression allocation, power-bandwidth approximation, and graph-based pairing.
Satisfaction of combined distortion, latency, energy, power, and bandwidth constraints while maximizing total rate.

Where Pith is reading between the lines

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

The similarity-conditioned fusion approach could extend to video or sensor data streams beyond still images.
Dynamic semantic-similarity pairing may improve scalability in denser networks with many users.
Similar logistic-style interference characterizations might apply to other learned transceiver designs in wireless systems.

Load-bearing premise

The pair-dependent interference in the learned feature space can be accurately characterized by a bivariate logistic function parameterized by transmit power and compression ratio.

What would settle it

Hardware measurements or real-channel experiments that show the actual interference levels in the feature space deviate substantially from the bivariate logistic predictions across different power and compression settings.

Figures

Figures reproduced from arXiv: 2604.09255 by Jiaxiang Wang, Mingzhe Chen, Mohammad Shikh-Bahaei, Yahao Ding, Zhaohui Yang, Zhijin Qin, Zhouxiang Zhao.

**Figure 2.** Figure 2: The semantic interference factor between users [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 4.** Figure 4: The reconstruction quality when SNR = 13 dB and CBR = 0.125 under different baselines. Each row corresponds to the reconstructed image of a single user across all baselines, while each column displays the concurrently transmitted images of two users. at low SNR and low CBR to over 31 dB at high SNR and high CBR, with diminishing returns in CBR at high values [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗

**Figure 3.** Figure 3: Performance evaluation under different channel and [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗

**Figure 5.** Figure 5: Average total transmission rate versus total transm [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗

**Figure 8.** Figure 8: Pairing strategy comparison under different number [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗

read the original abstract

Integrated learning and communication (ILAC) unifies learned transceivers with radio resource management, where semantic feature multiple access (SFMA) enables paired users to superpose their learned representations over shared time-frequency resources. Unlike conventional multiple access schemes, SFMA interference arises in the learned feature space and depends jointly on the user pair, the transmit power, and the compression ratio. This coupling ties binary pairing decisions to continuous resource variables, yielding a mixed-integer non-convex optimization problem. To address this problem, we first propose similarity-conditioned SFMA (SC-SFMA), a Swin Transformer-based transceiver whose dual-conditioned similarity modulator (DC-SimM) gates cross-user feature fusion according to the inter-user semantic similarity. We then characterize the resulting pair-dependent interference by a bivariate logistic function parameterized by transmit power and compression ratio, thereby bridging the learned transceiver with network-level optimization. On this basis, we formulate a sum-rate maximization problem subject to per-user distortion, latency, energy, power, and bandwidth constraints. To solve this problem, we develop a three-block alternating optimization algorithm that integrates dual-decomposition-assisted compression ratio allocation, trust-region successive convex approximation (SCA) for joint power-bandwidth optimization, and dynamic feasible graph-based user pairing. Simulation results show that SC-SFMA achieves considerable peak signal-to-noise ratio (PSNR) and multi-scale structural similarity index measure (MS-SSIM) gains over deep joint source-channel coding (JSCC) and separation-based baselines. The proposed optimization framework attains significant sum rate improvements over conventional multiple access baselines.

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 ties a Swin-Transformer semantic transceiver with similarity-conditioned feature fusion to a three-block optimizer by fitting pair-dependent interference to a bivariate logistic function of power and compression, but that fit is not shown to match the transceiver outputs.

read the letter

The main thing to know is that this work proposes SC-SFMA, a Swin Transformer transceiver with a dual-conditioned similarity modulator that gates cross-user feature fusion by semantic similarity, then approximates the resulting interference with a bivariate logistic function so the mixed-integer sum-rate problem can be solved by alternating compression allocation, trust-region SCA on power and bandwidth, and dynamic graph pairing. That specific combination of learned transceiver and tractable interference model for resource optimization looks new relative to the JSCC and separation baselines they compare against. The simulations report PSNR and MS-SSIM gains plus better sum rates than conventional multiple access under the listed constraints on distortion, latency, energy, power, and bandwidth. The formulation is coherent on its own terms and the three-block algorithm is a reasonable way to handle the coupling between pairing and continuous variables. The soft spot is exactly the one the stress-test flags: the bivariate logistic is the bridge that lets the optimizer run, yet the abstract gives no derivation, fitting procedure, or goodness-of-fit numbers against the actual DC-SimM outputs, especially at low similarity or extreme compression ratios. If the approximation deviates, the reported allocations and gains will not hold in the real system. This is a material concern for reproducibility rather than a minor detail. The paper is aimed at researchers working on semantic communications and integrated learning-communication systems who want to see learned transceivers plugged into network-level optimization. A reader in that area would get value from the architecture and the reported gains, provided the interference model is checked. It deserves a serious referee to examine the validation of the logistic fit, the simulation setup, and whether the gains survive when the model is replaced by the actual transceiver behavior.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes semantic feature multiple access (SFMA) for integrated learning and communication networks. It introduces similarity-conditioned SFMA (SC-SFMA) using a Swin Transformer-based transceiver with a dual-conditioned similarity modulator (DC-SimM) to control cross-user feature fusion based on semantic similarity. Pair-dependent interference in the learned feature space is characterized by a bivariate logistic function of transmit power and compression ratio. This model supports formulation of a sum-rate maximization problem subject to per-user distortion, latency, energy, power, and bandwidth constraints, which is solved by a three-block alternating optimization algorithm integrating dual-decomposition for compression ratios, trust-region successive convex approximation for power-bandwidth allocation, and dynamic feasible graph-based pairing. Simulations claim considerable PSNR and MS-SSIM gains over deep JSCC and separation-based baselines, plus significant sum-rate improvements over conventional multiple access schemes.

Significance. If the interference model proves accurate and reproducible, the work could advance semantic communications by tightly coupling learned transceivers with radio resource management, offering a pathway to improved multi-user efficiency under semantic constraints. The three-block alternating algorithm provides a practical way to handle the mixed-integer coupling between discrete pairing and continuous resources. However, the absence of validation for the core modeling step limits immediate impact and generalizability.

major comments (2)

[Abstract and interference characterization section] Abstract and the section on interference characterization: the pair-dependent interference is modeled by a bivariate logistic function parameterized by power and compression ratio, which is substituted directly into the three-block alternating algorithm (trust-region SCA and dynamic graph pairing steps). No derivation, fitting procedure, parameter values, or goodness-of-fit metrics (e.g., MSE or R² against DC-SimM outputs across similarity levels and compression ratios) are supplied, rendering the reported sum-rate allocations potentially mismatched to the actual Swin-Transformer transceiver.
[Simulation results section] Simulation results section: the claimed PSNR, MS-SSIM, and sum-rate gains are stated without error bars, confidence intervals, number of Monte Carlo runs, or statistical tests, so the magnitude and reliability of the improvements over JSCC and conventional baselines cannot be rigorously assessed.

minor comments (2)

The abstract refers to 'considerable' and 'significant' gains; providing at least one quantitative example (e.g., average dB improvement or percentage) would improve readability.
[Notation and optimization formulation] Ensure all parameters of the bivariate logistic function are explicitly defined with symbols and ranges when first introduced, and confirm they remain consistent in the optimization formulation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript proposing semantic feature multiple access for integrated learning and communication networks. We address each major comment point by point below and will revise the manuscript accordingly to improve documentation and statistical rigor.

read point-by-point responses

Referee: [Abstract and interference characterization section] Abstract and the section on interference characterization: the pair-dependent interference is modeled by a bivariate logistic function parameterized by power and compression ratio, which is substituted directly into the three-block alternating algorithm (trust-region SCA and dynamic graph pairing steps). No derivation, fitting procedure, parameter values, or goodness-of-fit metrics (e.g., MSE or R² against DC-SimM outputs across similarity levels and compression ratios) are supplied, rendering the reported sum-rate allocations potentially mismatched to the actual Swin-Transformer transceiver.

Authors: We agree that the current manuscript lacks sufficient documentation on the bivariate logistic interference model. This model was empirically derived by fitting to interference values generated from the DC-SimM within the Swin Transformer transceiver across a range of inter-user semantic similarities, transmit powers, and compression ratios. In the revised version, we will insert a new subsection detailing the data generation process, the nonlinear least-squares fitting procedure, the optimized parameter values, and goodness-of-fit metrics including R² and MSE. These additions will confirm the model's accuracy and justify its direct use in the trust-region SCA and dynamic graph pairing steps of the three-block algorithm. revision: yes
Referee: [Simulation results section] Simulation results section: the claimed PSNR, MS-SSIM, and sum-rate gains are stated without error bars, confidence intervals, number of Monte Carlo runs, or statistical tests, so the magnitude and reliability of the improvements over JSCC and conventional baselines cannot be rigorously assessed.

Authors: We concur that the simulation results section would be strengthened by explicit statistical reporting. The presented metrics are averages, but variability information is omitted. In the revised manuscript, we will add error bars denoting standard deviations computed over 100 independent Monte Carlo runs for all PSNR, MS-SSIM, and sum-rate curves. We will also state the run count explicitly and include a brief note on t-test results confirming the statistical significance of the gains relative to the deep JSCC and conventional multiple access baselines. revision: yes

Circularity Check

1 steps flagged

Bivariate logistic interference model bridges transceiver to optimizer but is presented without derivation or fit validation

specific steps

fitted input called prediction [Abstract (bridging paragraph after SC-SFMA proposal)]
"We then characterize the resulting pair-dependent interference by a bivariate logistic function parameterized by transmit power and compression ratio, thereby bridging the learned transceiver with network-level optimization."

The logistic form is introduced as the link that enables the three-block alternating optimization (dual-decomposition compression, trust-region SCA power-bandwidth, dynamic graph pairing). If this bivariate logistic is obtained by fitting to DC-SimM simulation outputs (as implied by the reader's note and absence of analytic derivation), then the optimization variables and claimed sum-rate improvements are computed inside a model whose parameters are themselves extracted from the same transceiver behavior; the 'prediction' of network performance therefore reduces to re-optimization of the fitted input rather than an independent result.

full rationale

The derivation chain proceeds from DC-SimM transceiver design to an empirical characterization of pair-dependent interference, then directly to a mixed-integer optimization solved by alternating algorithm. The characterization step supplies the functional form used in all subsequent power, bandwidth, compression, and pairing decisions. No first-principles derivation or external validation metric is quoted in the provided text; the logistic is introduced solely to 'bridge' the learned transceiver to network-level optimization. This makes the reported sum-rate gains dependent on the accuracy of that specific functional substitution rather than an independent analytical model. The step is therefore a fitted-input substitution rather than a closed derivation, warranting a moderate circularity score but not a full reduction to self-definition.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

The central claim rests on the accuracy of modeling feature-space interference via a fitted logistic function and on simulation-based validation of the optimization under stated constraints.

free parameters (1)

bivariate logistic function parameters
Parameterized by transmit power and compression ratio to model pair-dependent interference.

axioms (1)

domain assumption Interference in the learned feature space depends jointly on the user pair, transmit power, and compression ratio
Invoked to bridge the learned transceiver design with network-level sum-rate optimization.

invented entities (1)

dual-conditioned similarity modulator (DC-SimM) no independent evidence
purpose: Gates cross-user feature fusion according to inter-user semantic similarity in the Swin Transformer transceiver
New component introduced to condition the transceiver on semantic similarity.

pith-pipeline@v0.9.0 · 5602 in / 1442 out tokens · 68665 ms · 2026-05-10T17:01:33.399369+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

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

[1]

A vision of 6G wireless sy stems: Applications, trends, technologies, and open research pro blems,

W. Saad, M. Bennis, and M. Chen, “A vision of 6G wireless sy stems: Applications, trends, technologies, and open research pro blems,” IEEE Netw., vol. 34, no. 3, pp. 134–142, 2019

2019
[2]

Edge artiﬁcial int elligence for 6G: Vision, enabling technologies, and applications,

K. B. Letaief, Y . Shi, J. Lu, and J. Lu, “Edge artiﬁcial int elligence for 6G: Vision, enabling technologies, and applications,” IEEE J. Sel. Areas Commun., vol. 40, no. 1, pp. 5–36, 2021

2021
[3]

The roadmap to 6G: AI empowered wireless networks,

K. B. Letaief, W. Chen, Y . Shi, J. Zhang, and Y .-J. A. Zhang , “The roadmap to 6G: AI empowered wireless networks,” IEEE Commun. Mag., vol. 57, no. 8, pp. 84–90, 2019

2019
[4]

Edge learning for B5G networks with distributed signal pro cessing: Semantic communication, edge computing, and wireless sens ing,

W. Xu, Z. Y ang, D. W. K. Ng, M. Levorato, Y . C. Eldar, and M. D ebbah, “Edge learning for B5G networks with distributed signal pro cessing: Semantic communication, edge computing, and wireless sens ing,” IEEE J. Sel. Topics Signal Process. , vol. 17, no. 1, pp. 9–39, 2023

2023
[5]

A new path to integrated learning and communication (ILAC): Large AI models leveraging hyperdimensional computing,

W. Xu, Z. Y ang, D. W. K. Ng, R. Schober, H. V . Poor, Z. Zhang, and X. Y ou, “A new path to integrated learning and communication (ILAC): Large AI models leveraging hyperdimensional computing,” IEEE Trans. on Commun. , vol. 74, pp. 4948–4973, 2026

2026
[6]

ComAI: The convergence of communication and artiﬁcial intelligence,

P . Zhang, K. Niu, X. Wang, Y . Liu, Z. Liang, C. Dong, J. Dai, X. Xu, W. Xu, Z. Zhang et al. , “ComAI: The convergence of communication and artiﬁcial intelligence,” IEEE Commun. Surveys Tuts. , 2025

2025
[7]

Semantic communications for future in ternet: Fundamentals, applications, and challenges,

W. Y ang, H. Du, Z. Q. Liew, W. Y . B. Lim, Z. Xiong, D. Niyato, X. Chi, X. Shen, and C. Miao, “Semantic communications for future in ternet: Fundamentals, applications, and challenges,” IEEE Commun. Surveys Tuts., vol. 25, no. 1, pp. 213–250, 2022

2022
[8]

Task- oriented communications for 6G: Vision, principles, and te chnologies,

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

2023
[9]

Agentic AI-Empowered Wireless Agent Networks With Semantic-Aware Collaboration via ILAC

Z. Zhao, J. Wang, Z. Y ang, K. Y ang, Z. Zhang, M. Chen, and K. Huang, “Agentic AI-empowered wireless agent networks with semant ic-aware collaboration via ILAC,” arXiv preprint arXiv:2604.02381 , 2026. [Online]. Available: https://arxiv.org/abs/2604.02381

work page internal anchor Pith review Pith/arXiv arXiv 2026
[10]

Deep jo int source- channel coding for wireless image transmission,

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

2019
[11]

Nonlinear transform source-channel coding for semantic communicati ons,

J. Dai, S. Wang, K. Tan, Z. Si, X. Qin, K. Niu, and P . Zhang, “Nonlinear transform source-channel coding for semantic communicati ons,” IEEE J. Sel. Areas Commun. , vol. 40, no. 8, pp. 2300–2316, 2022

2022
[12]

Swi nJSCC: Taming swin transformer for deep joint source-channel codi ng,

K. Y ang, S. Wang, J. Dai, X. Qin, K. Niu, and P . Zhang, “Swi nJSCC: Taming swin transformer for deep joint source-channel codi ng,” IEEE Trans. on Cogn. Commun. Netw. , 2024

2024
[13]

Semantic feature multiple access (SFMA) over wireless net works,

J. Wang, Z. Y ang, C. Huang, Z. Zhang, M. Shikh-Bahaei, an d M. Chen, “Semantic feature multiple access (SFMA) over wireless net works,” in Proc. IEEE INFOCOM Wkshps. , 2025, pp. 1–6

2025
[14]

Generative AI empowered semantic feature multipl e access (SFMA) over wireless networks,

J. Wang, Y . Y ang, Z. Y ang, C. Huang, M. Chen, Z. Zhang, and M. Shikh- Bahaei, “Generative AI empowered semantic feature multipl e access (SFMA) over wireless networks,” IEEE Trans. on Cogn. Commun. Netw., 2025

2025
[15]

Generative joint source-channel coding for semantic image transmission,

E. Erdemir, T.-Y . Tung, P . L. Dragotti, and D. G¨ und¨ uz,“Generative joint source-channel coding for semantic image transmission,” IEEE J. Sel. Areas Commun. , vol. 41, no. 8, pp. 2645–2657, 2023

2023
[16]

Deep learning-based superpo sition coded modulation for hierarchical semantic communications over broadcast channels,

Y . Bo, S. Shao, and M. Tao, “Deep learning-based superpo sition coded modulation for hierarchical semantic communications over broadcast channels,” IEEE Trans. Commun. , vol. 73, no. 2, pp. 1186–1200, 2024

2024
[17]

Task-oriented multi-user semantic communications,

H. Xie, Z. Qin, X. Tao, and K. B. Letaief, “Task-oriented multi-user semantic communications,” IEEE J. Sel. Areas Commun. , vol. 40, no. 9, pp. 2584–2597, 2022

2022
[18]

Non- orthogonal multiple access enhanced multi-user semantic communicati on,

W. Li, H. Liang, C. Dong, X. Xu, P . Zhang, and K. Liu, “Non- orthogonal multiple access enhanced multi-user semantic communicati on,” IEEE Trans. on Cogn. Commun. Netw. , vol. 9, no. 6, pp. 1438–1453, 2023

2023
[19]

DeepMA: End-to-end deep multiple access for wi reless image transmission in semantic communication,

W. Zhang, K. Bai, S. Zeadally, H. Zhang, H. Shao, H. Ma, an d V . C. Leung, “DeepMA: End-to-end deep multiple access for wi reless image transmission in semantic communication,” IEEE Trans. on Cogn. Commun. Netw., vol. 10, no. 2, pp. 387–402, 2023

2023
[20]

Semantic feature division multiple access for mul ti-user digital interference networks,

S. Ma, C. Zhang, B. Shen, Y . Wu, H. Li, S. Li, G. Shi, and N. A l- Dhahir, “Semantic feature division multiple access for mul ti-user digital interference networks,” IEEE Trans. Wireless Commun. , vol. 23, no. 10, pp. 15 230–15 244, 2024

2024
[21]

Optimi zation of image transmission in cooperative semantic communication networks,

W. Zhang, Y . Wang, M. Chen, T. Luo, and D. Niyato, “Optimi zation of image transmission in cooperative semantic communication networks,” IEEE Trans. Wireless Commun. , vol. 23, no. 2, pp. 861–873, 2023

2023
[22]

Compression ratio allocation for probabilis tic semantic communication with RSMA,

Z. Zhao, Z. Y ang, Y . Hu, C. Zhu, M. Shikh-Bahaei, W. Xu, Z. Zhang, and K. Huang, “Compression ratio allocation for probabilis tic semantic communication with RSMA,” IEEE Trans. Commun. , 2025

2025
[23]

Adaptable semanti c compression and resource allocation for task-oriented communications ,

C. Liu, C. Guo, Y . Y ang, and N. Jiang, “Adaptable semanti c compression and resource allocation for task-oriented communications ,” IEEE Trans. on Cogn. Commun. Netw. , vol. 10, no. 3, pp. 769–782, 2023

2023
[24]

IRS-enhanced s ecure semantic communication networks: Cross-layer and context -awared re- source allocation,

L. Wang, W. Wu, F. Zhou, Z. Qin, and Q. Wu, “IRS-enhanced s ecure semantic communication networks: Cross-layer and context -awared re- source allocation,” IEEE Trans. Wireless Commun. , vol. 24, no. 1, pp. 494–508, 2024

2024
[25]

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

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

2022
[26]

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

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

2023
[27]

Impact of user pairing on 5G nonorthogonal multiple-access downlink transmissions,

Z. Ding, P . Fan, and H. V . Poor, “Impact of user pairing on 5G nonorthogonal multiple-access downlink transmissions,” IEEE Trans. V eh. Technol., vol. 65, no. 8, pp. 6010–6023, 2016

2016
[28]

Dynamic user clu stering and power allocation for uplink and downlink non-orthogona l multiple access (NOMA) systems,

M. S. Ali, H. Tabassum, and E. Hossain, “Dynamic user clu stering and power allocation for uplink and downlink non-orthogona l multiple access (NOMA) systems,” IEEE Access, vol. 5, pp. 18 446–18 461, 2017

2017
[29]

Join t power allocation and decoding order selection for NOMA systems: O utage- optimal strategies,

M. Y ang, J. Chen, Z. Ding, Y . Liu, L. Lv, and L. Y ang, “Join t power allocation and decoding order selection for NOMA systems: O utage- optimal strategies,” IEEE Trans. Wireless Commun. , vol. 23, no. 1, pp. 290–304, 2023

2023
[30]

Optimal us er pairing for downlink non-orthogonal multiple access (NOMA),

L. Zhu, J. Zhang, Z. Xiao, X. Cao, and D. O. Wu, “Optimal us er pairing for downlink non-orthogonal multiple access (NOMA),” IEEE Wireless Commun. Lett. , vol. 8, no. 2, pp. 328–331, 2018

2018
[31]

Swin transformer: Hierarchical vision transforme r using shifted windows,

Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vision transforme r using shifted windows,” in Proc. IEEE/CVF Int. Conf. Comput. Vis. (ICCV) , 2021, pp. 10 012–10 022

2021
[32]

BPG image format,

F. Bellard, “BPG image format,” 2014, [Online]. Availa ble: https://bellard.org/bpg/

2014
[33]

NR; physical layer procedures for data,

3GPP, “NR; physical layer procedures for data,” 3rd Gen eration Partner- ship Project (3GPP), Technical Speciﬁcation (TS) 38.214, 2 018, version 15.0.0, Release 16