Pith. sign in

REVIEW 4 major objections 5 minor 52 references

Unsupervised Learning Approach for Beamforming in Cell-Free Integrated Sensing and Communication

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read An unsupervised, distributed teacher–student deep learning scheme designs cell-free ISAC beamformers whose quality is close to the convex-optimization baseline while cutting computation by at least three orders of magnitude.

desk verdict Plausible distributed-DL scheme for cell-free ISAC, but the local-CSI approximation is unquantified and the evaluation lacks statistical rigor; still deserves a serious referee. read the letter →

arxiv 2412.18162 v1 pith:GNGE7BGS submitted 2024-12-24 eess.SP

Mohamed Elrashidy , Mudassir Masood , Ali Arshad Nasir This is my paper · ORCID
classification eess.SP
keywords cell-freemassiveMIMOintegratedsensingandcommunicationbeamformingunsupervisedlearningteacher-studenttrainingdistributeddeepSNRsignal-to-interference-plus-noiseratio
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to show that the joint sensing and communication beamforming problem of cell-free ISAC systems can be solved in real time by a distributed neural network, rather than by a centralized iterative optimizer. The proposed scheme trains one small DNN per access point; each network sees only its own AP's channel state information and outputs that AP's normalized beams for all users and the sensing target. To avoid manually balancing the sensing-versus-communication tradeoff, two teacher models are trained to extremes, one maximizing sensing SNR and the other minimum SINR, and their average scores are used to normalize the loss of a student model whose balance parameter adapts during training. The experiments with a U-net architecture report student performance close to the state-of-the-art convex solver on line-of-sight channels, with the distributed inference run time lower by at least three orders of magnitude.

What carries the argument

The load-bearing mechanism is the distributed teacher–student training unit. Two ensembles of per-AP DNNs are first trained with the teacher loss $L_t^{(\beta)} = -[(1-\beta)g_1 + \beta g_2]$, one with $\beta=0$ (sensing-only) and one with $\beta=1$ (communication-only); their mean scores over the training set supply the reference maxima $\hat{g}_1^{(\max)}$ and $\hat{g}_2^{(\max)}$. A third identical ensemble is trained with the student loss $L_s^{(\lambda)} = -\left[(1-\lambda)\frac{g_1}{\hat{g}_1^{(\max)}} + \lambda\frac{g_2}{\hat{g}_2^{(\max)}}\right]$, where $\lambda$ is updated each mini-batch from the normalized gaps $G_1$ and $G_2$, so the sensing–communication balance is steered automatically rather than set by a fixed hyperparameter. Each DNN outputs its AP's beam set from that AP's local CSI, and the outputs are normalized to the AP power budget before SSNR and SINR are evaluated.

What would settle it

Run the identical teacher–student procedure with more access points (for example $L=4$ and $L=8$): if the student's minimum-SINR gap to the centralized convex baseline grows steadily as $L$ increases, the claim that local-CSI-based distributed networks match the centralized solution is falsified.

Watch

Extended reading notes

Core claim

The central claim is that a teacher–student unsupervised training procedure can produce per-access-point beamformers for cell-free ISAC that are near the quality of a centralized joint SSNR/SINR optimization, while requiring each AP to use only its local CSI at deployment. The method's contribution is that it makes the problem unsupervised and distributed: no labeled beamformer dataset is needed, the DNNs cooperate only through the loss during offline training, and after training each AP can generate its own beams independently. Reported results show the student model matching or approaching the convex baseline's average SINR at two and five users, with a slight shortfall at eight users, and an inference time around a thousandth of the baseline's.

Load-bearing premise

The load-bearing premise is that each AP's neural network can produce near-optimal beamformers using only that AP's local channel information, even though the SINR expression couples every AP's channels and beams through each user's receiver.

Editorial extensions

If this is right

  • After offline training, each AP can compute its own beamformers from local CSI alone, so real-time deployment no longer requires the CPU to solve the optimization or send beam vectors over fronthaul for each channel realization.
  • Because training is unsupervised, generating a dataset does not require repeatedly solving the non-convex optimization to create labels; only channel samples and the SSNR/SINR formulas are needed.
  • The adaptive $\lambda$ rule removes the need for a costly grid search over the tradeoff parameter, and the teacher-provided maxima give an architecture-specific upper bound on what the student can achieve.
  • In the reported line-of-sight settings the U-net student stayed close to the convex baseline as the user count varied, and the measured inference time was roughly three orders of magnitude shorter than the two-step bisection plus SDP baseline.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the strongest untested premise is that local CSI at each AP is enough; since the SINR of every user couples all APs, the interesting stress test is to increase the number of APs and observe whether the student-to-baseline gap widens.
  • Editorial inference: because the teacher scores bound what the student can reach, the method's ceiling is set by architecture capacity; the F5-with-80k-points experiment already hints that bigger filters and more data are the levers, and one could map the ceiling as a function of model size.
  • Editorial inference: the evaluation is restricted to line-of-sight channels and fixed AP/user position distributions; transferring to multipath or moving targets would likely require retraining or an input representation that is less geometry-specific.
  • Editorial inference: the 'at least three orders of magnitude' claim is an implementation-level runtime comparison on a specific CPU/GPU pairing, not an asymptotic statement, so sharper evidence would come from profiling on the actual edge hardware where the AP DNNs would run.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes an unsupervised, distributed, teacher-student deep learning method for joint communication and sensing beamforming in cell-free ISAC systems. It trains one DNN per AP, where the l-th DNN maps only the l-th AP's local CSI to that AP's beamforming vectors, and trains two biased teacher models (SSNR-only and SINR-only) whose average scores are then used to normalize the student loss. The student is deployed per AP at inference time. Experiments with 1D-CNN, CAE, and U-net architectures compare average SSNR/SINR curves against a CVX-based benchmark from [28] and report a runtime comparison claiming at least three orders of magnitude speedup.

Significance. If the central claim holds, the work is practically significant: it offers a path to real-time, decentralized beamforming for cell-free ISAC without relying on a central processor or heavy fronthaul signaling. The unsupervised training avoids generating expensive labeled datasets, and the distributed inference structure is a genuine architectural contribution. The paper also tests multiple CNN architectures and includes a runtime comparison against a conventional solver, which is a strength. However, the significance is conditional because the evaluation does not yet support the headline quantitative claims: there is no independent test set, no statistical variability reporting, and the core distributed-CSI premise is not validated against a centralized baseline or worst-case metrics.

major comments (4)
  1. [Section 5.1, Eqs. (13)-(14) and Eq. (2)] The distributed architecture uses the l-th DNN input as only the l-th AP's local CSI (Eqs. (13)-(14)), while the SINR objective in Eq. (2) couples all APs through the stacked channels h_n and beamformers w_lq. In general, the optimal beamformer at one AP depends on the channels and beamformers of all other APs, so a function of local CSI alone cannot reproduce the centralized optimum. The paper does not quantify the resulting gap: it never compares against a centralized DNN that receives full CSI, and it reports only average SSNR/SINR curves over 200 benchmark points without per-sample or worst-case results. Since the distributed deployment and fronthaul-load reduction claims rest on this approximation, the authors should add a focused numerical study of the local-CSI gap, including per-sample SINR distributions and, ideally, a centralized-DNN baseline.
  2. [Section 8, especially Figs. 7-10, 13-16] The evaluation protocol is not sufficient to support the close-to-CVX performance claim. The CVX benchmark is averaged over only 200 randomly selected points with no variance indicated, and all DL training/validation curves appear to be single runs without error bars or multiple seeds. Moreover, the final reported student performance is selected post hoc from the validation curves using a 94% SINR threshold (Section 8.3), and no separate test set is held out. This makes the reported numbers vulnerable to optimistic selection bias and prevents an assessment of statistical significance. The authors should add error bars over multiple training runs, report performance on an independent test set, and state the model-selection rule before evaluation.
  3. [Section 4, Eq. (11) and Section 5.2] The student loss in Eq. (11) is claimed to be bounded in [-1,0] because the denominators are said to be the 'maximum possible' scores g1(max) and g2(max). However, these values are implemented as sample means of the teacher scores over the training set, not true maxima. A student can exceed the teacher's mean, in which case g1/E[g1(max)] > 1 and the loss can fall below -1. The interpretation of the teacher models as providing strict upper bounds on student performance is therefore not mathematically valid as stated. Either the denominators should be actual maxima (with appropriate robustness handling) or the boundedness and 'maximum attainable' claims should be revised.
  4. [Section 8.4 and Fig. 16b] The abstract's headline 'at least three orders of magnitude' speedup is based on a runtime comparison shown in Fig. 16b, but the figure reports average runtime without confidence intervals, and the text does not specify the number of repeated trials or the variance across runs. A strong quantitative claim of this type needs a more careful timing methodology, including warm-up effects, per-sample runtime distributions, and hardware/software version details. The fairness of comparing distributed DL inference (single AP's DNN) against a centralized solver should also be discussed explicitly, although the maximum-of-two-APs choice in the paper is a reasonable start.
minor comments (5)
  1. [Section 8.3] The sentence beginning 'To address the results shown by Figure 15, we the notation, F n' is missing a verb ('introduce'), and later in the same section 'ahchieved' is a typo for 'achieved'.
  2. [Notation and Section 2] The notation block defines I_N as 'the identity vector of a size N x N', which should be 'identity matrix'.
  3. [Section 8.1 and Fig. 8b] The discussion of CAE overfitting is clear, but the figures would benefit from explicit markers or legends distinguishing training and validation curves, since several curves are close together and hard to read in grayscale.
  4. [Section 5.2 and Eq. (12)] The lambda update rule in Eq. (12) has no convergence analysis or theoretical justification; a brief explanation of why this update should balance the two gaps would improve the presentation, even if the empirical behavior is the main evidence.
  5. [References] Some cited works are arXiv preprints rather than peer-reviewed publications; the authors should update references where stable versions exist, and should double-check the novelty statement in the Introduction in light of the recent cell-free ISAC literature.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the teacher-score normalization in Eq. (11) is a fixed training constant, and the close-to-CVX performance claim is benchmarked against the external solution of [28].

full rationale

The derivation is self-contained against an external baseline. The student loss (Eq. 11) divides g1 and g2 by teacher-derived sample means, but those means are fixed constants computed once before student training; they do not enter the evaluation and do not force the reported SINR/SSNR values. The central performance claim is checked against the CVX-based solution of [28] on 200 points (Section 8), and runtime is measured directly (Fig. 16b). The only self-citation is [18], used as background in the introduction and not load-bearing for any technical step. The local-CSI restriction in Section 5.1 (Eqs. 13-14) is an architectural assumption whose gap to the centralized optimum is not quantified; that is an unquantified approximation and a robustness/correctness concern, not a circular reduction, because none of the paper's equations is defined in terms of the result it is used to predict.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The method introduces no new physical entities or forces. Its load-bearing assumptions are modeling choices: local-CSI sufficiency for per-AP beamformers, the SSNR sensing metric, LOS channels, and the CVX baseline's near-optimality. The free parameters are mostly training hyperparameters and data-derived normalization constants that the central performance claim depends on.

free parameters (5)
  • Teacher normalization constants g1_max and g2_max = Sample means of teacher scores over training set, values not reported numerically
    Enter the student loss in Eq. (11) as denominators and control the balance between SSNR and SINR. They are estimated from the same training data, so they are fitted quantities rather than externally given.
  • Student normalization parameter lambda(0) = 0.5
    Initial value of the student loss balance parameter set by hand in Section 8; lambda then adapts via Eq. (12).
  • Adaptation step size epsilon = 0.01
    Fixed step size for lambda updates in Eq. (12), chosen by hand.
  • Communication power ratio rho = 0.5
    Used in benchmark problem (8) for the CVX baseline; taken from [28] and affects the benchmark operating point.
  • Model selection SINR threshold = 94%
    Post hoc criterion in Section 8.3 used to pick the epoch for final reported student performance; changing it changes the reported SSNR/SINR tradeoff.
assumptions (6)
  • domain assumption Per-AP beamformers can be well approximated as functions of only local CSI at that AP
    Section 5.1 states 'The input to the l-th model is the channel state information (CSI) of the l-th AP.' Since SINR in Eq. (2) couples all APs, optimal beamforming generally needs global CSI; the method's distributed claim rests on this sufficiency.
  • domain assumption The SSNR expression in Eq. (6) is the right sensing performance metric
    The paper adopts the SSNR metric from [28] and states all beams contribute to sensing. If actual radar detection performance differs, the optimized beams may not maximize sensing quality.
  • domain assumption Line-of-sight channel model with steering vectors in Eq. (4) is sufficient
    Section 7 generates both communication and sensing channels using Eq. (4) with no multipath, blockage, or channel estimation error, and CSI is assumed available.
  • domain assumption The CVX-based solution of [28] is a valid near-optimal state-of-the-art baseline
    Section 8 compares against problems (8)-(9) solved with bisection and SDP, treating it as the reference; the paper does not re-derive or question its optimality.
  • ad hoc to paper Teacher sample-mean scores bound student performance for the same architecture
    Section 5.2 assumes g1_max and g2_max from the biased teachers are the maximum attainable scores for the student, which is used to construct the student loss in Eq. (11).
  • ad hoc to paper Lambda adaptation in Eq. (12) converges to a desirable balance
    The update rule with fixed step size is heuristic; no convergence analysis is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unsupervised Learning Approach for Beamforming in Cell-Free Integrated Sensing and Communication." pith.science (2026). https://pith.science/paper/GNGE7BGS

@misc{pith2026241218162,
  author       = {Pith},
  title        = {Pith review of: Unsupervised Learning Approach for Beamforming in Cell-Free Integrated Sensing and Communication},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GNGE7BGS}},
  note         = {Machine review of arXiv:2412.18162}
}
read the original abstract

Cell-free massive multiple input multiple output (MIMO) systems can provide reliable connectivity and increase user throughput and spectral efficiency of integrated sensing and communication (ISAC) systems. This can only be achieved through intelligent beamforming design. While many works have proposed optimization methods to design beamformers for cell-free systems, the underlying algorithms are computationally complex and potentially increase fronthaul link loads. To address this concern, we propose an unsupervised learning algorithm to jointly design the communication and sensing beamformers for cell-free ISAC system. Specifically, we adopt a teacher-student training model to guarantee a balanced maximization of sensing signal to noise ratio (SSNR) and signal to interference plus noise ratio (SINR), which represent the sensing and communication metrics, respectively. The proposed scheme is decentralized, which can reduce the load on the central processing unit (CPU) and the required fronthaul links. To avoid the tradeoff problem between sensing and communication counterparts of the cell-free system, we first train two identical models (teacher models) each biased towards one of the two tasks. A third identical model (a student model) is trained based on the maximum sensing and communication performance information obtained by the teacher models. While the results show that our proposed unsupervised DL approach yields a performance close to the state-of-the-art solution, the proposed approach is more computationally efficient than the state of the art by at least three orders of magnitude.

Figures

Figures reproduced from arXiv: 2412.18162 by the authors.

Figure 1
Figure 1. Cell-free ISAC example setup. by the l-th AP to the q-th agent (i.e., a UE or an ST). Define the subsets N ≜ {1, ..., N} and S ≜ {N +1} to denote the stacked communication beam indices and the stacked sensing beam index in the set Q, respectively. The first N beam indices in Q are reserved for the UEs, whereas the last index of Q is reserved for the ST. Moreover, define the stacked communication channel vector for t… view at source ↗
Figure 2
Figure 2. The proposed training unit for cell-free ISAC beamforming. The shaded boxes [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. The proposed teacher-student training paradigm. [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figures from the paper (13 more)
Figure 4
Figure 4. Figure 4: A 3-layer 1D-CNN example with a filter length of 11. [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]
Figure 5
Figure 5. Figure 5: A convolutional autoencoder with skip-connections. [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]
Figure 6
Figure 6. Figure 6: Different position distribution schemes used for dataset generation at [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]
Figure 7
Figure 7. Figure 7: SSNR teacher performance. (a) 1D-CNN (b) CAE (c) U-net [PITH_FULL_IMAGE:figures/full_fig_p020_7.png]
Figure 8
Figure 8. Figure 8: SINR teacher performance. through training and validation using the three networks. It can be seen from both [PITH_FULL_IMAGE:figures/full_fig_p020_8.png]
Figure 9
Figure 9. Figure 9: Student performance of the CAE. changes caused by changes in the system model, network architecture, train￾ing parameters, or training data points. Such susceptibility to minor changes renders the fixed β training practise impractical. A similar argument can be stated …
Figure 10
Figure 10. Figure 10: Student performance of U-net. It is expected that SINR would increase as N decreases, since the multi￾antenna equipped APs are required to serve less number of users by the same power budget. This explains the differences between the three curves in Figure 13a. On the…
Figure 11
Figure 11. Figure 11: SINR and SSNR at fixed β [PITH_FULL_IMAGE:figures/full_fig_p024_11.png]
Figure 12
Figure 12. Figure 12: SINR and SSNR at fixed λ. tation, only minor changes are expected. Said changes can be observed in Figure 14a, where the SINR curve declines after around 200 epochs, suggest￾ing a slight overfitting during the training process. To address the results shown by [PITH_F…
Figure 13
Figure 13. Figure 13: Student performance of U-net at different number of UEs: (a) SINR and (b) [PITH_FULL_IMAGE:figures/full_fig_p025_13.png]
Figure 14
Figure 14. Figure 14: Student performance at N = 5 for different agent position distribution schemes (c.f., [PITH_FULL_IMAGE:figures/full_fig_p026_14.png]
Figure 15
Figure 15. Figure 15: Student performance at N = 8 for different training and network setups: (a) SINR and (b) SSNR validation curves. correspond to an SINR of at least 1.4798 (i.e., 94% of the maximum SINR). As such, out of ten epochs that satisfy the 94% condition, the maximum SSNR is ac…
Figure 16
Figure 16. Figure 16: (a) Selection strategy results at 94% threshold percentage of SINR for different [PITH_FULL_IMAGE:figures/full_fig_p027_16.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 46 canonical work pages

  1. [28]

    Demirhan, A

    U. Demirhan, A. Alkhateeb, Cell-free isac MIMO systems: Joint sens- ing and communication beamforming, arXiv preprint arXiv:2301.11328 (2023)

    arXiv 2023

  2. [1]

    S. K. Dehkordi, G. Caire, Active sensing schemes for beam-space MIMO radar in ISAC applications, in: 2023 IEEE Wireless Communications and Networking Conference (WCNC), IEEE, 2023, pp. 1–6

    work page 2023

  3. [2]

    P. Gao, L. Lian, J. Yu, Cooperative ISAC with direct localization and rate-splitting multiple access communication: A pareto optimization framework, IEEE Journal on Selected Areas in Communications 41 (5) (2023) 1496–1515

    work page 2023

  4. [3]

    C. Liu, X. Liu, S. Li, W. Yuan, D. W. K. Ng, Deep CLSTM for pre- dictive beamforming in integrated sensing and communication-enabled vehicular networks, Journal of Communications and Information Net- works 7 (3) (2022) 269–277

    work page 2022

  5. [4]

    W. Yuan, Z. Wei, S. Li, J. Yuan, D. W. K. Ng, Integrated sensing and communication-assisted orthogonal time frequency space transmission for vehicular networks, IEEE Journal of Selected Topics in Signal Pro- cessing 15 (6) (2021) 1515–1528

    work page 2021

  6. [5]

    Cheng, D

    X. Cheng, D. Duan, S. Gao, L. Yang, Integrated sensing and commu- nications (ISAC) for vehicular communication networks (VCN), IEEE Internet of Things Journal 9 (23) (2022) 23441–23451

    work page 2022

  7. [6]

    K. Meng, Q. Wu, J. Xu, W. Chen, Z. Feng, R. Schober, A. L. Swindle- hurst, UA V-enabled integrated sensing and communication: Opportu- nities and challenges, IEEE Wireless Communications (2023)

    work page 2023

  8. [7]

    F. Dong, F. Liu, Y. Cui, W. Wang, K. Han, Z. Wang, Sensing as a ser- vice in 6G perceptive networks: A unified framework for ISAC resource allocation, IEEE Transactions on Wireless Communications (2022)

    work page 2022

Show all 52 references
  1. [8]

    F. Dong, F. Liu, Localization as a service in perceptive networks: An ISAC resource allocation framework, in: 2022 IEEE International Con- ference on Communications Workshops (ICC Workshops), IEEE, 2022, pp. 848–853

    2022

  2. [9]

    F. Zeng, J. Yu, J. Li, F. Liu, D. Wang, X. You, Integrated sensing and communication for network-assisted full-duplex cell-free distributed massive MIMO systems, arXiv preprint arXiv:2311.05101 (2023). 29

    2023 arXiv

  3. [10]

    D. Xu, X. Yu, D. W. K. Ng, A. Schmeink, R. Schober, Robust and secure resource allocation for ISAC systems: A novel optimization framework for variable-length snapshots, IEEE Transactions on Communications 70 (12) (2022) 8196–8214

    2022

  4. [11]

    Z. He, W. Xu, H. Shen, D. W. K. Ng, Y. C. Eldar, X. You, Full-duplex communication for ISAC: Joint beamforming and power optimization, IEEE Journal on Selected Areas in Communications (2023)

    2023

  5. [12]

    H. Hua, J. Xu, T. X. Han, Transmit beamforming optimization for in- tegrated sensing and communication, in: 2021 IEEE Global Communi- cations Conference (GLOBECOM), IEEE, 2021, pp. 01–06

    2021

  6. [13]

    Z. Yu, X. Hu, C. Liu, M. Peng, C. Zhong, Location sensing and beam- forming design for irs-enabled multi-user isac systems, IEEE Transac- tions on Signal Processing 70 (2022) 5178–5193

    2022

  7. [14]

    R. Liu, M. Li, A. L. Swindlehurst, Joint beamforming and reflection design for ris-assisted isac systems, in: 2022 30th European Signal Pro- cessing Conference (EUSIPCO), IEEE, 2022, pp. 997–1001

    2022

  8. [15]

    Z. Yu, H. Ren, C. Pan, G. Zhou, B. Wang, M. Dong, J. Wang, Active ris aided isac systems: Beamforming design and performance analysis, IEEE Transactions on Communications (2023)

    2023

  9. [16]

    Zhong, J

    K. Zhong, J. Hu, C. Pan, M. Deng, J. Fang, Joint waveform and beam- forming design for ris-aided isac systems, IEEE Signal Processing Letters 30 (2023) 165–169

    2023

  10. [17]

    Demirhan, A

    U. Demirhan, A. Alkhateeb, Integrated sensing and communication for 6G: Ten key machine learning roles, IEEE Communications Magazine (2023)

    2023

  11. [18]

    Salem, M

    H. Salem, M. Quamar, A. Mansoor, M. Elrashidy, N. Saeed, M. Masood, Data-driven integrated sensing and communication: Recent advances, challenges, and future prospects, arXiv preprint arXiv:2308.09090 (2023)

    2023 arXiv

  12. [19]

    Y. Liu, I. Al-Nahhal, O. A. Dobre, F. Wang, Deep-learning-based chan- nel estimation for IRS-assisted isac system, in: GLOBECOM 2022-2022 IEEE Global Communications Conference, IEEE, 2022, pp. 4220–4225. 30

    2022

  13. [20]

    C. Liu, W. Yuan, S. Li, X. Liu, H. Li, D. W. K. Ng, Y. Li, Learning-based predictive beamforming for integrated sensing and communication in vehicular networks, IEEE Journal on Selected Areas in Communications 40 (8) (2022) 2317–2334

    2022

  14. [21]

    J. Mu, Y. Gong, F. Zhang, Y. Cui, F. Zheng, X. Jing, Integrated sensing and communication-enabled predictive beamforming with deep learning in vehicular networks, IEEE Communications Letters 25 (10) (2021) 3301–3304

    2021

  15. [22]

    X. Liu, H. Zhang, K. Long, A. Nallanathan, V. C. Leung, Distributed unsupervised learning for interference management in integrated sensing and communication systems, IEEE Transactions on Wireless Commu- nications (2023)

    2023

  16. [23]

    Elhoushy, M

    S. Elhoushy, M. Ibrahim, W. Hamouda, Cell-free massive mimo: A sur- vey, IEEE Communications Surveys & Tutorials 24 (1) (2021) 492–523

    2021

  17. [24]

    Nayebi, A

    E. Nayebi, A. Ashikhmin, T. L. Marzetta, H. Yang, B. D. Rao, Precod- ing and power optimization in cell-free massive MIMO systems, IEEE Transactions on Wireless Communications 16 (7) (2017) 4445–4459

    2017

  18. [25]

    Bj¨ ornson, L

    E. Bj¨ ornson, L. Sanguinetti, Making cell-free massive MIMO compet- itive with MMSE processing and centralized implementation, IEEE Transactions on Wireless Communications 19 (1) (2019) 77–90

    2019

  19. [26]

    H. Q. Ngo, A. Ashikhmin, H. Yang, E. G. Larsson, T. L. Marzetta, Cell- free massive MIMO versus small cells, IEEE Transactions on Wireless Communications 16 (3) (2017) 1834–1850

    2017

  20. [27]

    H. Q. Ngo, A. Ashikhmin, H. Yang, E. G. Larsson, T. L. Marzetta, Correction to” Cell-free massive MIMO versus small cells”., IEEE Trans. Wirel. Commun. 19 (5) (2020) 3623–3624

    2020

  21. [29]

    S. Liu, M. Li, Q. Liu, Joint BS mode selection and beamforming design for cooperative cell-free ISAC networks, arXiv preprint arXiv:2305.10800 (2023). 31

    2023 arXiv

  22. [30]

    Behdad, ¨O

    Z. Behdad, ¨O. T. Demir, K. W. Sung, E. Bj¨ ornson, C. Cavdar, Multi-static target detection and power allocation for integrated sens- ing and communication in cell-free massive MIMO, arXiv preprint arXiv:2305.12523 (2023)

    2023 arXiv

  23. [31]

    W. Mao, Y. Lu, C.-Y. Chi, B. Ai, Z. Zhong, Z. Ding, Communication- sensing region for cell-free massive MIMO isac systems, arXiv preprint arXiv:2307.16259 (2023)

    2023 arXiv

  24. [32]

    Elfiatoure, M

    M. Elfiatoure, M. Mohammadi, H. Q. Ngo, M. Matthaiou, Cell-free massive MIMO for ISAC: Access point operation mode selection and power control, arXiv preprint arXiv:2310.09032 (2023)

    2023 arXiv

  25. [33]

    Cao, Q.-Y

    Y. Cao, Q.-Y. Yu, Joint resource allocation for user-centric cell-free integrated sensing and communication systems, IEEE Communications Letters (2023)

    2023

  26. [34]

    L. A. Iliadis, Z. D. Zaharis, S. Sotiroudis, P. Sarigiannidis, G. K. Kara- giannidis, S. K. Goudos, The road to 6G: a comprehensive survey of deep learning applications in cell-free massive MIMO communications systems, EURASIP Journal on Wireless Communications and Network- in...

    2022

  27. [35]

    Zhao, Power allocation in cell-free massive MIMO: Using deep learn- ing methods, Ph.D

    Y. Zhao, Power allocation in cell-free massive MIMO: Using deep learn- ing methods, Ph.D. thesis, Ph. D. dissertation (2022)

    2022

  28. [36]

    Zaher, ¨O

    M. Zaher, ¨O. T. Demir, E. Bj¨ ornson, M. Petrova, Learning-based down- link power allocation in cell-free massive MIMO systems, IEEE Trans- actions on Wireless Communications 22 (1) (2022) 174–188

    2022

  29. [37]

    Nikbakht, A

    R. Nikbakht, A. Jonsson, A. Lozano, Unsupervised-learning power con- trol for cell-free wireless systems, in: 2019 IEEE 30th Annual Interna- tional Symposium on Personal, Indoor and Mobile Radio Communica- tions (PIMRC), IEEE, 2019, pp. 1–5

    2019

  30. [38]

    Hojatian, J

    H. Hojatian, J. Nadal, J.-F. Frigon, F. Leduc-Primeau, Decentralized beamforming for cell-free massive MIMO with unsupervised learning, IEEE Communications Letters 26 (5) (2022) 1042–1046. 32

    2022

  31. [39]

    Zhang, J

    Y. Zhang, J. Zhang, S. Buzzi, H. Xiao, B. Ai, Unsupervised deep learning for power control of cell-free massive MIMO systems, IEEE Transactions on Vehicular Technology (2023)

    2023

  32. [40]

    Rajapaksha, K

    N. Rajapaksha, K. S. Manosha, N. Rajatheva, M. Latva-aho, Unsu- pervised learning-based joint power control and fronthaul capacity al- location in cell-free massive MIMO with hardware impairments, IEEE Wireless Communications Letters (2023)

    2023

  33. [41]

    Rajapaksha, K

    N. Rajapaksha, K. S. Manosha, N. Rajatheva, M. Latva-Aho, Deep learning-based power control for cell-free massive MIMO networks, in: ICC 2021-IEEE International Conference on Communications, IEEE, 2021, pp. 1–7

    2021

  34. [42]

    Zhang, J

    Y. Zhang, J. Zhang, Y. Jin, S. Buzzi, B. Ai, Deep learning-based power control for uplink cell-free massive MIMO systems, in: 2021 IEEE Global Communications Conference (GLOBECOM), IEEE, 2021, pp. 1–6

    2021

  35. [43]

    L. Luo, J. Zhang, S. Chen, X. Zhang, B. Ai, D. W. K. Ng, Downlink power control for cell-free massive MIMO with deep reinforcement learn- ing, IEEE Transactions on Vehicular Technology 71 (6) (2022) 6772– 6777

    2022

  36. [44]

    Liu, C.-Y

    Y.-C. Liu, C.-Y. Ma, Z. He, C.-W. Kuo, K. Chen, P. Zhang, B. Wu, Z. Kira, P. Vajda, Unbiased teacher for semi-supervised object detection, arXiv preprint arXiv:2102.09480 (2021)

    2021 arXiv

  37. [45]

    Vandeghen, A

    R. Vandeghen, A. Cioppa, M. Van Droogenbroeck, Semi-supervised training to improve player and ball detection in soccer, in: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 3481–3490

    2022

  38. [46]

    J. Xu, X. Wang, P. Zhu, X. You, Privacy-preserving channel estima- tion in cell-free hybrid massive MIMO systems, IEEE Transactions on Wireless Communications 20 (6) (2021) 3815–3830

    2021

  39. [47]

    H. Song, T. Goldstein, X. You, C. Zhang, O. Tirkkonen, C. Studer, Joint channel estimation and data detection in cell-free massive MU-MIMO systems, IEEE Transactions on Wireless Communications 21 (6) (2021) 4068–4084. 33

    2021

  40. [48]

    Amadid, M

    J. Amadid, M. Boulouird, A. Belhabib, A. Zeroual, On channel esti- mation for rician fading with the phase-shift in cell-free massive mimo system, Wireless Personal Communications (2022) 1–21

    2022

  41. [49]

    Ronneberger, P

    O. Ronneberger, P. Fischer, T. Brox, U-net: Convolutional networks for biomedical image segmentation, in: Medical Image Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, Proceedings, Part III 18, Spring...

    2015

  42. [50]

    C. Qi, W. Ci, J. Zhang, X. You, Hybrid beamforming for millimeter wave MIMO integrated sensing and communications, IEEE Communications Letters 26 (5) (2022) 1136–1140

    2022

  43. [51]

    Y. Zhuo, Z. Wang, Performance analysis of ISAC system under corre- lated communication-sensing channel, IEEE Transactions on Vehicular Technology (2023)

    2023

  44. [52]

    H. Wang, J. Yang, FBUNet: Full convolutional network based on fu- sion block architecture for biomedical image segmentation, Journal of Medical and Biological Engineering 41 (2021) 185–202. 34

    2021

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.