arxiv: 2602.08260 · v2 · submitted 2026-02-09 · 📡 eess.SP

Recognition: no theorem link

Towards Optimal Semantic Communications: Reconsidering the Role of Semantic Feature Channels

Yongjeong Oh , Jihong Park , Jinho Choi , Yo-Seb Jeon

Authors on Pith no claims yet

Pith reviewed 2026-05-16 06:14 UTC · model grok-4.3

classification 📡 eess.SP

keywords semantic communicationsemantic featureschannel optimizationmutual informationpower allocationjoint optimizationanalog communicationdigital communication

0 comments

The pith

The semantic feature channel is optimizable and admits a closed-form optimum under linear Gaussian assumptions.

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

The paper models transmission from semantic feature encoder output to decoder input as a single configurable SF channel that includes the physical link and all transceiver processing. It shows that this channel can be shaped through choices such as power allocation and derives its optimal form analytically when the encoder-decoder pair is linear and the source is Gaussian, subject to a mutual-information constraint on reconstruction quality. From this derivation the authors build a joint optimization procedure that designs the encoder, decoder, and SF channel together for both analog and digital semantic communication. They also supply a physical-layer calibration method for real-time power control that adapts to changing conditions. Simulations indicate that the resulting systems outperform conventional fixed-channel designs on task accuracy across varied environments.

Core claim

By redefining the entire path from encoder outputs to decoder inputs as the semantic feature channel and treating its statistics as design variables, the authors obtain a closed-form optimal SF channel under linear encoding, linear decoding, and Gaussian source statistics; this optimum is then used inside a joint optimization framework that simultaneously tunes the encoder-decoder pair and the SF channel, with a calibration procedure that realizes the target statistics via power control at the physical layer.

What carries the argument

The semantic feature (SF) channel, defined as the composite mapping from encoder outputs through the physical channel and all transceiver operations to the decoder inputs, whose statistics are treated as configurable design variables.

If this is right

Joint optimization of encoder, decoder, and SF channel becomes feasible for both analog and digital semantic systems.
Power allocation at the transmitter can be used to realize the derived optimal SF channel statistics.
A physical-layer calibration procedure enables real-time adaptation of the SF channel to varying propagation conditions.
Task performance improves relative to fixed-SF-channel baselines under a range of communication environments.

Where Pith is reading between the lines

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

The same optimization logic may supply useful initializations or regularizers even when nonlinear networks replace the linear encoder-decoder.
The closed-form optimum could serve as a benchmark for testing whether learned semantic feature channels in deep networks approach information-theoretic limits.
Extensions to non-Gaussian sources would require replacing the analytic step with numerical optimization while retaining the joint-design structure.

Load-bearing premise

That the encoder-decoder structure remains linear and the data source is Gaussian, allowing the closed-form derivation of the optimal SF channel.

What would settle it

A direct numerical check, under the same linear encoder-decoder and Gaussian source, showing that the analytically derived SF channel does not achieve the highest mutual information (or task performance) among all channels satisfying the power and reconstruction constraints.

Figures

Figures reproduced from arXiv: 2602.08260 by Jihong Park, Jinho Choi, Yongjeong Oh, Yo-Seb Jeon.

**Figure 2.** Figure 2: MSE curves over the mutual information limit [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 3.** Figure 3: The proposed end-to-end training strategy jointly o [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

**Figure 4.** Figure 4: PSNR curves over the mutual information limit [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗

**Figure 5.** Figure 5: PSNR curves over the mutual information limit [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗

**Figure 6.** Figure 6: PSNR curves over the SNR for single-user analog and di [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗

**Figure 7.** Figure 7: SSIM performance over varying SNRs for multi-user di [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗

**Figure 8.** Figure 8: Selection ratios over the SNR for the user transmitti [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗

read the original abstract

This paper investigates the optimization of transmitting the encoder outputs, termed semantic features (SFs), in semantic communication (SC). We begin by modeling the entire communication process from the encoder output to the decoder input, encompassing the physical channel and all transceiver operations, as the SF channel, thereby establishing an encoder-SF channel-decoder pipeline. In contrast to prior studies that assume a fixed SF channel, we note that the SF channel is configurable, as its characteristics are shaped by various transmission and reception strategies, such as power allocation. Based on this observation, we formulate the SF channel optimization problem under a mutual information constraint between the SFs and their reconstructions, and analytically derive the optimal SF channel under a linear encoder-decoder structure and Gaussian source assumption. Building on this analysis, we propose a joint optimization framework for the encoder-decoder and SF channel applicable to both analog and digital SC systems. To realize the optimized SF channel, we also propose a physical-layer calibration strategy that enables real-time power control and adaptation to varying channel conditions. Simulation results demonstrate that the proposed SF channel optimization achieves superior task performance under various communication environments.

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 derives a closed-form optimal semantic feature channel under linear-Gaussian assumptions and proposes a joint encoder-decoder-channel optimization framework that shows simulation gains.

read the letter

The main thing to know is that this work treats the semantic feature channel as optimizable instead of fixed and delivers an analytical optimum for the linear encoder-decoder plus Gaussian source case. That closed-form result plus the joint framework for analog and digital systems is the concrete addition relative to earlier semantic communications papers that left the channel untouched. They also add a physical-layer calibration step for real-time power control, and the simulations report better task performance under power and bandwidth constraints. The modeling of the full encoder-SF channel-decoder pipeline is clean and the mutual-information constraint is a reasonable way to bound the optimization. The derivation itself looks like a direct extension of standard linear-Gaussian rate-distortion ideas applied to this setting. The central limitation is exactly the one the abstract flags: the closed form holds only under linear structure and Gaussian sources. Typical semantic systems use nonlinear DNN encoders and non-Gaussian data such as images or text, so the analytical optimum does not automatically carry over. Any reported gains in the simulations then rest on the empirical joint optimizer rather than the closed-form result. I would want to see whether the framework still improves performance when the encoder is nonlinear and whether the calibration strategy works outside the assumed regime. This is aimed at researchers already working on semantic communications who want an analytical handle on the physical-layer part. A reader who values closed-form insights under stated assumptions will find it useful; someone looking for immediately deployable nonlinear results may find the gap noticeable. It deserves a serious referee because the framing is coherent, the math is grounded in standard tools, and the joint framework is a clear next step even if the assumptions need more discussion.

Referee Report

2 major / 2 minor

Summary. The paper models the semantic feature (SF) channel in semantic communications as a configurable entity determined by transceiver strategies. It analytically derives a closed-form optimal SF channel under a linear encoder-decoder structure and Gaussian source assumption subject to a mutual-information constraint between SFs and reconstructions. It then proposes a joint optimization framework for the encoder, decoder, and SF channel that applies to both analog and digital systems, introduces a physical-layer calibration strategy for real-time power control, and reports simulation results showing improved task performance across communication environments.

Significance. If the derivation is correct and the joint framework can be shown to retain benefits outside the linear-Gaussian regime, the work would supply a principled interface between semantic encoders and physical channels. The closed-form result under stated assumptions and the explicit joint-optimization plus calibration approach are concrete strengths that could guide subsequent designs in semantic communications.

major comments (2)

[Analytical derivation section (referenced in abstract)] The analytical derivation of the optimal SF channel (stated in the abstract) is obtained under a linear encoder-decoder structure and Gaussian source assumption. This assumption is load-bearing for the claim of optimality; typical semantic-communication pipelines use nonlinear DNN encoders and non-Gaussian sources (images, text). The manuscript must clarify whether the closed-form channel serves only as an initialization or whether the joint optimizer is proven or empirically shown to inherit comparable optimality outside the linear-Gaussian case.
[Joint optimization framework and simulation results] The transition from the linear-Gaussian analysis to the general joint-optimization framework for analog and digital systems requires additional justification. Without explicit discussion or experiments isolating the contribution of the derived optimum versus the joint optimizer, it is unclear whether the reported simulation gains rest on the analytical result or on empirical tuning.

minor comments (2)

[Modeling section] Clarify the precise definition and notation of the SF channel (including how power allocation and transceiver operations enter the model) to avoid ambiguity when moving between the analytical and practical sections.
[Simulation results] In the simulation results, report the specific task metrics, the exact baselines used, and any ablation studies that isolate the SF-channel optimization component.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major point below and will incorporate clarifications and additional experiments in the revised version to better delineate the scope of our analytical results and the contributions of the joint optimization framework.

read point-by-point responses

Referee: The analytical derivation of the optimal SF channel is obtained under a linear encoder-decoder structure and Gaussian source assumption. This assumption is load-bearing for the claim of optimality; typical semantic-communication pipelines use nonlinear DNN encoders and non-Gaussian sources. The manuscript must clarify whether the closed-form channel serves only as an initialization or whether the joint optimizer is proven or empirically shown to inherit comparable optimality outside the linear-Gaussian case.

Authors: We agree that the closed-form derivation holds specifically under the linear-Gaussian assumptions stated in Section III. The result is intended to provide theoretical insight into the structure of an optimal SF channel and to serve as a principled initialization point for the subsequent joint optimization. In the revised manuscript we will add an explicit paragraph in the introduction and a dedicated limitations subsection clarifying that we do not claim or prove optimality for nonlinear encoders or non-Gaussian sources. The joint framework itself is formulated without relying on linearity and is applied in a data-driven manner; our simulations already include DNN-based encoders and show consistent gains, but we will emphasize that these gains are empirical rather than theoretically guaranteed outside the analyzed regime. revision: yes
Referee: The transition from the linear-Gaussian analysis to the general joint-optimization framework for analog and digital systems requires additional justification. Without explicit discussion or experiments isolating the contribution of the derived optimum versus the joint optimizer, it is unclear whether the reported simulation gains rest on the analytical result or on empirical tuning.

Authors: We will strengthen the transition by adding a new subsection (III-D) that explains how the closed-form solution informs the parameterization of the SF channel within the joint optimizer. To isolate contributions, we will include new ablation experiments comparing (i) the joint optimizer initialized with the derived optimal SF channel versus random initialization and (ii) performance with the analytical SF channel fixed versus jointly optimized. These results will be reported in Section V and will help quantify the benefit attributable to the analytical insight versus the general optimization procedure. revision: yes

Circularity Check

0 steps flagged

No significant circularity; analytical derivation is self-contained under explicit external assumptions

full rationale

The paper derives the optimal SF channel in closed form from a mutual-information constraint under the stated linear encoder-decoder structure and Gaussian source assumption. This is a standard constrained optimization result, not a reduction to fitted parameters, self-citations, or definitional tautologies. The subsequent joint optimization framework is presented as building on this independent analytic step rather than presupposing it. No load-bearing self-citation chains or ansatz smuggling are identifiable from the provided text.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Central claim rests on two domain assumptions required for the closed-form derivation; no free parameters or invented entities are stated in the abstract.

axioms (2)

domain assumption Linear encoder-decoder structure
Required to obtain the analytical optimum of the SF channel.
domain assumption Gaussian source assumption
Used to derive the optimal SF channel in closed form.

pith-pipeline@v0.9.0 · 5503 in / 1119 out tokens · 37926 ms · 2026-05-16T06:14:46.427980+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages

[1]

End-to-end learning for task- oriented semantic communications over MIMO channels: An in formation- theoretic framework,

C. Cai, X. Y uan, and Y .-J. A. Zhang, “End-to-end learning for task- oriented semantic communications over MIMO channels: An in formation- theoretic framework,” IEEE J. Sel. Areas Commun. , vol. 43, no. 4, pp. 1292–1307, Apr. 2025

work page 2025
[2]

Semantic communications : Overview, open issues, and future research directions,

X. Luo, H.-H. Chen, and Q. Guo, “Semantic communications : Overview, open issues, and future research directions,” IEEE Wireless Commun. , vol. 29, no. 1, pp. 210–219, Feb. 2022

work page 2022
[3]

Le ss data, more knowledge: Building next generation semantic co mmunication networks,

C. Chaccour, W. Saad, M. Debbah, Z. Han, and H. V . Poor, “Le ss data, more knowledge: Building next generation semantic co mmunication networks,” IEEE Commun. Surveys Tuts. , vol. 27, no. 1, pp. 37–76, Feb. 2025

work page 2025
[4]

Semantic communications for fu ture Internet: 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 fu ture Internet: Fundamentals, applications, and challenges,” IEEE Commun. Surveys Tuts., vol. 25, no. 1, pp. 213–250, ﬁrstquarter 2023

work page 2023
[5]

Semantics-nat ive commu- nication via contextual reasoning,

H. Seo, J. Park, M. Bennis, and M. Debbah, “Semantics-nat ive commu- nication via contextual reasoning,” IEEE Trans. Cogn. Commun. Netw. , vol. 9, no. 3, pp. 604–617, Jun. 2023

work page 2023
[6]

Joint task and data-oriented semantic communications: A deep separate so urce-channel coding scheme,

J. Huang, D. Li, C. Huang, X. Qin, and W. Zhang, “Joint task and data-oriented semantic communications: A deep separate so urce-channel coding scheme,” IEEE Internet Things J. , vol. 11, no. 2, pp. 2255–2272, Jan. 2024

work page 2024
[7]

Separate source channel coding is still what you n eed: An LLM-based rethinking,

T. Ren, R. Li, M.-m. Zhao, X. Chen, G. Liu, Y . Y ang, Z. Zhao, and H. Zhang, “Separate source channel coding is still what you n eed: An LLM-based rethinking,” 2025, arXiv:2501.04285

work page arXiv 2025
[8]

Joint source–channel coding: Fundamentals and recent progress i n practical designs,

D. G¨ und¨ uz, M. A. Wigger, T.-Y . Tung, P . Zhang, and Y . Xia o, “Joint source–channel coding: Fundamentals and recent progress i n practical designs,” Proc. IEEE , vol. 113, no. 9, pp. 888–919, Sep. 2025

work page 2025
[9]

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, Sep. 2019

work page 2019
[10]

DeepJSCC-Q: Constellation constrained deep joint source-channel codi ng,

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

work page 2022
[11]

OFDM-guided deep joint source channel coding for wireless multipath fading channels,

M. Y ang, C. Bian, and H.-S. Kim, “OFDM-guided deep joint source channel coding for wireless multipath fading channels,” IEEE Trans. Cogn. Commun. Netw. , vol. 8, no. 2, pp. 584–599, Jun. 2022

work page 2022
[12]

Vision transformer for adaptive image transmission over MIMO chan nels,

H. Wu, Y . Shao, C. Bian, K. Mikolajczyk, and D. G¨ und¨ uz, “Vision transformer for adaptive image transmission over MIMO chan nels,” in Proc. IEEE Int. Conf. Commun. (ICC) , Rome, Italy, May 2023, pp. 3702– 3707

work page 2023
[13]

Neural joint source-channel coding,

K. Choi, K. Tatwawadi, A. Grover, T. Weissman, and S. Erm on, “Neural joint source-channel coding,” in Proc. Int. Conf. Machine Learning (ICML), Long Beach, CA, USA, Jun. 2019, pp. 1182–1192

work page 2019
[14]

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, Aug. 2023

work page 2023
[15]

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

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

work page 2025
[16]

Universal joint source-cha nnel coding for modulation-agnostic semantic communication,

Y . Huh, H. Seo, and W. Choi, “Universal joint source-cha nnel coding for modulation-agnostic semantic communication,” IEEE J. Sel. Areas Commun., vol. 43, no. 7, pp. 2560–2574, Apr. 2025

work page 2025
[17]

DeepJSCC-1++: Robus t and bandwidth-adaptive wireless image transmission,

C. Bian, Y . Shao, and D. G¨ und¨ uz, “DeepJSCC-1++: Robus t and bandwidth-adaptive wireless image transmission,” in Proc. IEEE Global Commun. Conf. (GLOBECOM) , Kuala Lumpur, Malaysia, Dec. 2023, pp. 3148–3154

work page 2023
[18]

Rate-adapta ble multitask- oriented semantic communication: An extended rate–distor tion theory- based scheme,

F. Liu, Z. Sun, Y . Y ang, C. Guo, and S. Zhao, “Rate-adapta ble multitask- oriented semantic communication: An extended rate–distor tion theory- based scheme,” IEEE Internet Things J. , vol. 11, no. 9, pp. 15 557–15 570, May 2024

work page 2024
[19]

Goldsmith, Wireless communications

A. Goldsmith, Wireless communications. Cambridge, U.K.: Cambridge Univ. Press, 2005

work page 2005
[20]

Blind tr aining for channel-adaptive digital semantic communications,

Y . Oh, J. Park, J. Choi, J. Park, and Y .-S. Jeon, “Blind tr aining for channel-adaptive digital semantic communications,” IEEE Trans. Com- mun., vol. 73, no. 11, pp. 11 274–11 290, Nov. 2025

work page 2025
[21]

Communicat ion-efﬁcient split learning via adaptive feature-wise compression,

Y . Oh, J. Lee, C. G. Brinton, and Y .-S. Jeon, “Communicat ion-efﬁcient split learning via adaptive feature-wise compression,” IEEE Trans. Neural Netw. Learn. Syst. , vol. 36, no. 6, pp. 10 844–10 858, Jun. 2025

work page 2025
[22]

Quantization,

R. Gray and D. Neuhoff, “Quantization,” IEEE Trans. Inf. Theory, vol. 44, no. 6, pp. 2325–2383, Oct. 1998

work page 1998
[23]

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, Oct. 2024

work page 2024
[24]

T. M. Cover, Elements of information theory . John Wiley & Sons, 1999

work page 1999
[25]

Robust i nformation bottleneck for task-oriented communication with digital m odulation,

S. Xie, S. Ma, M. Ding, Y . Shi, M. Tang, and Y . Wu, “Robust i nformation bottleneck for task-oriented communication with digital m odulation,” IEEE J. Sel. Areas Commun. , vol. 41, no. 8, pp. 2577–2591, Aug. 2023

work page 2023
[26]

Deep learning-b ased modulation and power control: A BER perspective,

Y . Oh, J. Park, J. Choi, and Y .-S. Jeon, “Deep learning-b ased modulation and power control: A BER perspective,” IEEE Commun. Lett. , vol. 29, no. 11, pp. 2616–2620, Nov. 2025

work page 2025
[27]

On the general BER expression of one- and two-dimensional amplitude modulations,

K. Cho and D. Y oon, “On the general BER expression of one- and two-dimensional amplitude modulations,” IEEE Trans. Commun., vol. 50, no. 7, pp. 1074–1080, Jul. 2002

work page 2002
[28]

Kellerer, U

H. Kellerer, U. Pferschy, and D. Pisinger, The multiple-choice knapsack problem. Springer Berlin Heidelberg, 2004

work page 2004
[29]

Gradien t-based learning applied to document recognition,

Y . Lecun, L. Bottou, Y . Bengio, and P . Haffner, “Gradien t-based learning applied to document recognition,” Proc. IEEE, vol. 86, no. 11, pp. 2278– 2324, Nov. 1998

work page 1998
[30]

Learning multiple layers of features from tiny images,

A. Krizhevsky and G. Hinton, “Learning multiple layers of features from tiny images,” M.S. thesis, Univ. Toronto, Toronto, ON, Cana da, 2009

work page 2009
[31]

An analysis of single-laye r networks in unsupervised feature learning,

A. Coates, A. Ng, and H. Lee, “An analysis of single-laye r networks in unsupervised feature learning,” in Proc. Int. Conf. Artif. Intell. Statist. (AISTATS), Ft. Lauderdale, FL, USA, Apr. 2011, pp. 215–223

work page 2011
[32]

Adam: A method for stochastic opt imization,

D. P . Kingma and J. Ba, “Adam: A method for stochastic opt imization,” in Proc. Int. Conf. Learn. Represent. (ICLR) , San Diego, CA, USA, May 2015, pp. 1–13

work page 2015
[33]

Learning task-oriented c ommunication for edge inference: An information bottleneck approach,

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

work page 2022
[34]

Im- proving channel resilience for task-oriented semantic com munications: A uniﬁed information bottleneck approach,

S. Lyu, Y . Sun, L. Guo, X. Y uan, F. Fang, L. Zhang, and X. Wa ng, “Im- proving channel resilience for task-oriented semantic com munications: A uniﬁed information bottleneck approach,” IEEE Commun. Lett. , vol. 28, no. 11, pp. 2623–2627, Nov. 2024

work page 2024
[35]

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. Cogn. Commun. Netw. , vol. 11, no. 1, pp. 90–104, Feb. 2025

work page 2025
[36]

Knowl- edge base enabled semantic communication: A generative per spective,

J. Ren, Z. Zhang, J. Xu, G. Chen, Y . Sun, P . Zhang, and S. Cu i, “Knowl- edge base enabled semantic communication: A generative per spective,” IEEE Wireless Commun. , vol. 31, no. 4, pp. 14–22, Aug. 2024

work page 2024
[37]

Enhancing semantic communication with deep generative models: An ove rview,

E. Grassucci, Y . Mitsufuji, P . Zhang, and D. Comminiell o, “Enhancing semantic communication with deep generative models: An ove rview,” in Proc. IEEE Int. Conf. Acoust., Speech, Signal Process. (ICA SSP), Seoul, Korea, Republic of, Apr. 2024, pp. 13 021–13 025

work page 2024
[38]

Application of non-orthogonal multiple access in LT E and 5G networks,

Z. Ding, Y . Liu, J. Choi, Q. Sun, M. Elkashlan, I. Chih-Li n, and H. V . Poor, “Application of non-orthogonal multiple access in LT E and 5G networks,” IEEE Commun. Mag. , vol. 55, no. 2, pp. 185–191, Feb. 2017

work page 2017
[39]

Majorizations and inequalities in matrix the ory,

T. Ando, “Majorizations and inequalities in matrix the ory,” Linear Alge- bra Appl. , vol. 199, pp. 17–67, Mar. 1994

work page 1994