Pith. sign in

REVIEW 4 major objections 6 minor 51 references

Distributed Gossip-GAN for Low-overhead CSI Feedback Training in FDD mMIMO-OFDM Systems

T0 review · 4 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a gossip-trained GAN can substitute for centralized training of CSI-feedback autoencoders in FDD massive MIMO, matching accuracy while cutting uplink overhead.

desk verdict A credible engineering combination of gossip learning and GANs for CSI feedback; the parameter-averaging step is the load-bearing assumption and it is not adequately validated. read the letter →

arxiv 2509.10490 v1 pith:45XFHDAM submitted 2025-08-31 eess.SP cs.AIcs.ITmath.IT

Yuwen Cao , Guijun Liu , Tomoaki Ohtsuki , Howard H. Yang , Tony Q. S. Quek This is my paper
classification eess.SPcs.AIcs.ITmath.IT
keywords CSIfeedbackmassiveMIMOgenerativeadversarialnetworkgossiplearningcatastrophicforgettingdeepautoencoderFDDmMIMO-OFDM
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

This paper tries to establish that the neural network used to compress channel state information (CSI) in frequency-division duplex massive MIMO systems can be trained without shipping raw channel data to the base station. Its Gossip-GAN scheme has each participating user train a small generative adversarial network on locally collected channels, periodically exchange generator and discriminator weights with a few peers over device-to-device links, and average the received weights into the local model. One finished generator is then sent to the base station, which uses it to synthesize a large dataset for training the compressor-reconstructor autoencoder. The authors report that this scheme matches the feedback accuracy of centralized training on the DeepMIMO and COST2100 datasets, requires about ten times less computation per epoch than a centrally trained GAN, and lets a mobile user revisit an old environment without retraining by storing one small generator per environment instead of raw data.

What carries the argument

The load-bearing object is the averaged-weight Gossip-GAN: each user trains a generator and discriminator on local CSI, then every fixed interval sends the weights to a peer selected by a fixed topology, and when enough models arrive merges them by uniform averaging (Eqs. 15-16). The paper relies on this averaged model to be a valid GAN approximating the aggregate channel distribution; it then selects one trained generator, transmits its 0.455M parameters to the base station, and samples a synthetic dataset of size 10^4 to train the CsiNet autoencoder. For catastrophic forgetting, the same generator is stored per environment and the mixed dataset from all stored generators is used for traini

What would settle it

Train ten GANs on disjoint local CSI subsets, merge weights by Eq. (15), and compare the merged generator's output distribution against the true mixture distribution using a distribution distance such as Wasserstein distance; if the merged generator is far from the mixture, or if the autoencoder trained on its synthetic data fails to reach the reported NMSE range on held-out channels, the averaging step is the point of failure.

Watch

Extended reading notes

Core claim

The central discovery is that a fully distributed gossip-learning GAN, with no central server, can capture the channel distribution well enough that a single user's generator, forwarded to the base station, produces synthetic CSI on which a deep autoencoder achieves feedback accuracy close to training on true centralized data. The authors show this in two DeepMIMO scenarios and with COST2100, find performance improves with more participating users, and demonstrate the anti-forgetting property: after moving from a sparse to a dense area, combining the stored old-environment generator with the new one yields NMSE of -14.75 dB in the old area and -18.51 dB in the new one, versus -0.33 dB withou

Load-bearing premise

Everything downstream depends on the assumption that averaging the weights of peer-trained GANs yields a single generator whose synthetic samples still look like real channels from the combined user region; the paper does not prove this and borrows the averaging step from a position paper.

Editorial extensions

If this is right

  • Training a CSI-feedback autoencoder no longer requires uploading raw channel data to the base station; only the parameters of one generator are transmitted, and that overhead stays constant as antenna arrays scale up.
  • The distributed gossip strategy cuts per-epoch training computation and memory by roughly a factor of ten versus centralized GAN training, making generative training feasible on resource-constrained users.
  • Users returning to a previously seen environment keep useful feedback accuracy because each environment is represented by a compact generator, not raw data, and the memory cost drops from 39.06M to 1.82M units versus retraining.
  • The scheme is architecture-agnostic: replacing CsiNet with CRNet, CsiQNet, or DeepCMC preserves or improves NMSE.
  • Fully connected gossip topology approximates centralized GAN performance, and accuracy improves as the number of participating users grows from 5 to 10.

Reading between the lines

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

  • If uniform weight averaging of GANs is as valid as the paper assumes, the same gossip training recipe should transfer to other generative models such as VAEs or diffusion models for channel synthesis, with the same bandwidth savings.
  • The generator-per-environment memory strategy is essentially a parameter-based continual-learning mechanism; it could be combined with rehearsal-free techniques like elastic weight consolidation to retain more than two environments without storing a generator per scene.
  • Because only synthetic data reach the base station, the scheme could be paired with differential privacy on local gradients to blunt the adversarial-machine-learning risks the authors list as future work.
  • The reported bandwidth advantage grows with antenna count, so the scheme should become more attractive in larger mMIMO arrays; a direct scaling study would quantify that crossover.
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 / 6 minor

Summary. The paper proposes Gossip-GAN, a fully distributed framework for training CSI feedback autoencoders in FDD mMIMO-OFDM systems. Each UE collects a small local CSI dataset and trains a local GAN (a WGAN-GP/CTGAN variant), then gossips generator/discriminator weights with neighboring UEs; the averaged generator is sent to the BS, which synthesizes a large fake CSI dataset used to train a CsiNet-type autoencoder. The paper also applies the framework to catastrophic forgetting by storing past-scenario generators and replaying their samples when training on new scenarios. Experiments on DeepMIMO and COST2100 datasets compare NMSE against centralized GAN training, no-connection distributed GAN, federated learning, and true CSI training, and report comparable or slightly worse NMSE with lower uplink overhead and memory cost.

Significance. If the central claims hold, the framework is a meaningful step toward privacy-preserving, low-overhead distributed training for CSI feedback, and it demonstrates compatibility with several modern autoencoder architectures (CsiNet, CRNet, CsiQNet, DeepCMC). The paper includes useful empirical elements: public DeepMIMO data, repeated experiments with 95% confidence intervals, a comparison against federated GAN, and a memory-cost comparison for catastrophic forgetting. However, the load-bearing model-averaging step (Eqs. 15–16) is supported only by an IID-shard experiment and no convergence or distributional analysis, so the significance is currently conditional on additional validation in heterogeneous settings.

major comments (4)
  1. [§III-B, Eqs. (15)–(16)] Uniform averaging of generator and discriminator weights across gossip peers is the load-bearing step of the proposed framework, but it is supported only by a citation to a position paper [46]. GAN training is a non-convex min-max problem, and averaging two trained WGAN/CTGAN models is not generally guaranteed to produce a generator that approximates the aggregate CSI distribution; mode collapse or inconsistent BatchNorm statistics are real risks. The paper provides no convergence analysis, no aggregation-error bound, and no direct distributional check (e.g., MMD, FID, or generated-sample NMSE before autoencoder training) of the merged generator. I request either theoretical justification or a direct empirical validation of the merged model, especially before the downstream autoencoder training is shown to depend on it.
  2. [§IV-B, Fig. 8] The experimental setup uses K=10 UEs, each receiving 500 samples randomly drawn from the same dense-area or sparse-area dataset. This makes the local shards approximately IID, which is the most favorable regime for uniform parameter averaging. The framework is motivated by UEs at different locations, BS sectors, or mobility conditions, where local CSI distributions differ. No experiment varies the degree of non-IID-ness across UEs. Since uniform averaging is known to degrade under heterogeneity, the claim that Gossip-GAN 'can achieve similar CSI feedback accuracy as centralized training' is not yet established for the target deployment scenarios. Please add experiments with controlled non-IID shards (e.g., UEs assigned disjoint row ranges of DeepMIMO or different BSs) and report per-UE distribution shift and final NMSE.
  3. [Algorithm 1, lines 12–16] In the ONRECEIVEMODEL function, received models are saved to a buffer, and merging is triggered when the buffer size reaches npeers. The buffer is never cleared after MERGE_SAVED_MODELS, so a literal implementation would include stale models in every subsequent merge, rather than averaging the most recent npeers peer models. This is an algorithmic bug or an under-specification that affects the correctness and reproducibility of the gossip procedure. Please correct the pseudo-code (e.g., clear the buffer after merging) and also define npeers explicitly for both topologies (for Topology 2, presumably npeers=K−1) and specify the communication interval Δ used in the simulations.
  4. [§IV-B, Fig. 10 and S=1e4] The number of synthetic samples S is set to 1.0e4 after observing that the test NMSE saturates at that point in Fig. 10. Selecting a hyperparameter based on the test-set performance curve is a form of test-set leakage and makes the reported NMSE optimistic. S should be chosen on a validation split (or justified by a separate model-selection procedure), and the sensitivity of the final NMSE to S should be reported. This is load-bearing because the claim of 'similar accuracy to centralized training' depends on the chosen S; without a clean selection protocol, the reader cannot assess whether the reported numbers are tuned to the test set.
minor comments (6)
  1. [Abstract and §IV-A] The abstract states that results are obtained with 'real-world datasets,' but DeepMIMO is a ray-tracing simulation dataset and COST2100 is a stochastic channel-model dataset. Please rephrase to 'simulation-based channel datasets' or clarify what is meant by real-world.
  2. [§I] The introduction states 'Section III provides the simulation results' and 'Section IV concludes our paper,' but the actual structure has simulations in Section IV and conclusions in Section V. Please correct the section references.
  3. [References] Reference [10] is identical to reference [2]. Please remove the duplicate.
  4. [§IV-B and Table IV] The claim of reduced uplink bandwidth is not quantified against a concrete raw-CSI upload baseline. Please provide a table or calculation showing the number of bits for raw CSI (e.g., 32×32×2 complex values per sample × 5000 samples) versus the transmitted model parameters (0.455M floats), and also account for the D2D gossip traffic that occurs during training.
  5. [§III-B, Fig. 5] For Topology 2, the text says 'each UE communicates with all the other UEs simultaneously,' but Algorithm 1 does not define npeers for this case. Please state explicitly that npeers=K−1 for Topology 2 and clarify whether the save/merge buffer is reset after each aggregation round.
  6. [§IV-C, Table VI] The table layout is confusing: 'After training on' is used as a column header, but the rows mix the training environment and the evaluation environment. Please reformat the table so that rows clearly indicate (training environment, evaluation environment) pairs, and add a sentence explaining the catastrophic-forgetting phenomenon quantitatively.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claims are tested against external real-world data and centralized baselines.

full rationale

The paper's main claims (CSI feedback accuracy comparable to centralized training, catastrophic forgetting mitigation, low uplink overhead) are each validated by experiments against external benchmarks: real DeepMIMO and COST2100 channel data, a centrally trained GAN, true-CSI DAE training, and a federated GAN baseline. The generator-to-synthetic-data-to-autoencoder pipeline is evaluated by NMSE on held-out real CSI, so the evaluation is not reducible to the training inputs. The averaging rule in Eqs. (15)-(16) is a load-bearing design assumption taken from a cited position paper, but it is not derived from the paper's own fitting procedure; if averaging failed, the reported results would simply not hold. It is a correctness/robustness concern (especially under non-IID data), not a circularity. The choice of S=1e4 in Section IV-B is a hyperparameter selected from a validation curve, not a predicted quantity derived from itself. The only self-citations are incidental architecture/background references (e.g., [25], [36]) and do not support the central claim. An implementation bug (Algorithm 1 never clearing the saved received-model buffer) and the lack of convergence analysis for the merge step are noted as correctness risks, not circularity.

Assumptions & free parameters 8 free parameters · 5 assumptions · 0 invented entities

The framework relies on several domain assumptions and the unproven averaging step. The free parameters are standard hyperparameters or post hoc choices. No new theoretical entities are introduced.

free parameters (8)
  • Number of synthetic samples S = 1.0e4
    Chosen from saturation curve in Fig. 10; post hoc selection.
  • Number of UEs K = 10
    Set in simulations; performance improves with K (Fig. 12); no optimization.
  • Local data per UE = 500
    Arbitrary choice; total 5000 matched to centralized baseline.
  • GAN loss hyperparameters λ1, λ2, M' = 10, 2, 0.2
    From CTGAN [30]; not tuned here.
  • Dropout probability = 0.5
    Selected in TABLE III experiments; performance varies with it.
  • Communication interval = every 10 epochs
    Set in experiments; no analysis.
  • Learning rate = 0.001 for AE and GAN
    Standard choice; not justified.
  • Compression ratio γ = 1/16 default
    Used for most experiments; also tests 1/32-1/128.
assumptions (5)
  • ad hoc to paper Uniform averaging of GAN parameters (Eqs. 15-16) yields a valid aggregated model.
    Borrowed from position paper [46]; no convergence proof; central to the method.
  • domain assumption The K selected UEs observe the same channel distribution as the target deployment area.
    Assumed throughout; enables using one generator to serve all UEs in the region (Section III-D).
  • standard math WGAN gradient penalty and consistency regularization (CTGAN) stabilize training on small per-UE datasets.
    Adopted from [30]; the paper relies on it for generator quality.
  • domain assumption The channel model from DeepMIMO (Eqs. 2-6) accurately represents practical FDD mMIMO-OFDM channels.
    Used to generate all simulation data; limited to specific scenarios.
  • domain assumption Ignoring adversarial ML threats is acceptable.
    Stated in footnote 4; acknowledged in conclusions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Distributed Gossip-GAN for Low-overhead CSI Feedback Training in FDD mMIMO-OFDM Systems." pith.science (2026). https://pith.science/paper/45XFHDAM

@misc{pith2026250910490,
  author       = {Pith},
  title        = {Pith review of: Distributed Gossip-GAN for Low-overhead CSI Feedback Training in FDD mMIMO-OFDM Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/45XFHDAM}},
  note         = {Machine review of arXiv:2509.10490}
}
read the original abstract

The deep autoencoder (DAE) framework has turned out to be efficient in reducing the channel state information (CSI) feedback overhead in massive multiple-input multipleoutput (mMIMO) systems. However, these DAE approaches presented in prior works rely heavily on large-scale data collected through the base station (BS) for model training, thus rendering excessive bandwidth usage and data privacy issues, particularly for mMIMO systems. When considering users' mobility and encountering new channel environments, the existing CSI feedback models may often need to be retrained. Returning back to previous environments, however, will make these models perform poorly and face the risk of catastrophic forgetting. To solve the above challenging problems, we propose a novel gossiping generative adversarial network (Gossip-GAN)-aided CSI feedback training framework. Notably, Gossip-GAN enables the CSI feedback training with low-overhead while preserving users' privacy. Specially, each user collects a small amount of data to train a GAN model. Meanwhile, a fully distributed gossip-learning strategy is exploited to avoid model overfitting, and to accelerate the model training as well. Simulation results demonstrate that Gossip-GAN can i) achieve a similar CSI feedback accuracy as centralized training with real-world datasets, ii) address catastrophic forgetting challenges in mobile scenarios, and iii) greatly reduce the uplink bandwidth usage. Besides, our results show that the proposed approach possesses an inherent robustness.

Figures

Figures reproduced from arXiv: 2509.10490 by the authors.

Figure 1
Figure 1. The central dataset training scenario for CSI feedba [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. An overview of the autoencoder CSI feedback framewor [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The working diagram of GAN model. be understood as the discriminator’s accuracy in determining whether the CSI is true or fake. Using alternating training, the generator tries to generate data similar to true CSI: maximize the Loss, while the discriminator tries to judge the truth of the data: minimize the Loss. This adversarial approach will eventually allow the generator to model the real CSI distribution Pr(H). T… view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: The workflow diagram of the proposed Gossip-GAN CSI tr [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: The diagrams of our proposed two different network to [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: The schematic diagram of Gossip-GAN used to solve cat [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: “O1 28” an outdoor scenario of two streets and one intersec￾tion at operating frequencies 28 GHz in DeepMIMO. and share similar channel characteristics. During the offline training phase, the proposed Gossip-GAN framework selects a subset of K UEs to train a CSI feedba…
Figure 9
Figure 9. Figure 9: The NMSE performance comparison of the proposed fram [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 10
Figure 10. Figure 10: NMSE performance vs. the number of fake data. [PITH_FULL_IMAGE:figures/full_fig_p010_10.png]
Figure 12
Figure 12. Figure 12: NMSE (dB) performance of the proposed framework in t [PITH_FULL_IMAGE:figures/full_fig_p011_12.png]
Figure 13
Figure 13. Figure 13: Comparison of NMSE performance after coupling the [PITH_FULL_IMAGE:figures/full_fig_p011_13.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

51 extracted references · 44 canonical work pages

  1. [46]

    Gossiping gans : Position paper,

    C. Hardy, E. Le Merrer, and B. Sericola, “Gossiping gans : Position paper,” in Proc. W orkshop Distrib. Infrastructures Deep Learn. , 2018, pp. 25–28

    work page 2018

  2. [1]

    An overview of massive MIMO: Benefits and challenges,

    L. Lu, G. Y . Li, A. L. Swindlehurst, A. Ashikhmin, and R. Zh ang, “An overview of massive MIMO: Benefits and challenges,” IEEE J. Sel. Topics Signal Process. , vol. 8, no. 5, pp. 742–758, 2014

    2014

  3. [2]

    Massive MIMO for next generation wireless systems,

    E. G. Larsson, O. Edfors, F. Tufvesson, and T. L. Marzetta , “Massive MIMO for next generation wireless systems,” IEEE Commun. Mag. , vol. 52, no. 2, pp. 186–195, 2014

    2014

  4. [3]

    On an intelligent hie rarchical routing strategy for ultra-dense free space optical low ear th orbit satellite networks,

    B. Mao, X. Zhou, J. Liu, and N. Kato, “On an intelligent hie rarchical routing strategy for ultra-dense free space optical low ear th orbit satellite networks,” IEEE J. Sel. Areas Commun. , vol. 42, no. 5, pp. 1219–1230, 2024

    work page 2024

  5. [4]

    A speculative study on 6G,

    F. Tariq, M. R. A. Khandaker, K.-K. Wong, M. A. Imran, M. Be nnis, and M. Debbah, “A speculative study on 6G,” IEEE Wireless Commun. , vol. 27, no. 4, pp. 118–125, 2020

    work page 2020

  6. [5]

    Statist ical 3-D beamforming for large-scale MIMO downlink systems over ric ian fading channels,

    X. Li, S. Jin, H. A. Suraweera, J. Hou, and X. Gao, “Statist ical 3-D beamforming for large-scale MIMO downlink systems over ric ian fading channels,” IEEE Trans. Commun. , vol. 64, no. 4, pp. 1529–1543, 2016

    work page 2016

  7. [6]

    Prospective multiple antenna technologies for beyo nd 5G,

    J. Zhang, E. Bj¨ ornson, M. Matthaiou, D. W. K. Ng, H. Y ang, and D. J. Love, “Prospective multiple antenna technologies for beyo nd 5G,” IEEE J. Sel. Areas Commun. , vol. 38, no. 8, pp. 1637–1660, 2020

    work page 2020

  8. [7]

    On a novel high accuracy posit ioning with intelligent reflecting surface and unscented kalman filter f or intelligent transportation systems in B5G,

    Y . Zhu, B. Mao, and N. Kato, “On a novel high accuracy posit ioning with intelligent reflecting surface and unscented kalman filter f or intelligent transportation systems in B5G,” IEEE J. Sel. Areas Commun. , vol. 42, no. 1, pp. 68–77, 2024

    work page 2024

Show all 51 references
  1. [8]

    On a hierarchical content caching and asynchronous updating s cheme for non-terrestrial network-assisted connected automated ve hicles,

    B. Mao, Y . Liu, H. Guo, Y . Xun, J. Wang, J. Liu, and N. Kato, “ On a hierarchical content caching and asynchronous updating s cheme for non-terrestrial network-assisted connected automated ve hicles,” IEEE J. Sel. Areas Commun. , vol. 43, no. 1, pp. 64–74, 2025

    2025

  2. [9]

    Large-scale antenn a systems with hybrid analog and digital beamforming for millimeter w ave 5G,

    S. Han, C.-l. I, Z. Xu, and C. Rowell, “Large-scale antenn a systems with hybrid analog and digital beamforming for millimeter w ave 5G,” IEEE Commun. Mag. , vol. 53, no. 1, pp. 186–194, 2015. 13

    2015

  3. [10]

    Massive MIMO for next generation wireless systems,

    E. G. Larsson, O. Edfors, F. Tufvesson, and T. L. Marzett a, “Massive MIMO for next generation wireless systems,” IEEE Commun. Mag. , vol. 52, no. 2, pp. 186–195, 2014

    2014

  4. [11]

    Next generation 5G wir eless networks: A comprehensive survey,

    M. Agiwal, A. Roy, and N. Saxena, “Next generation 5G wir eless networks: A comprehensive survey,” IEEE Commun. Surveys Tutor . , vol. 18, no. 3, pp. 1617–1655, 2016

    2016

  5. [12]

    6G wireless communications networks: A compreh ensive survey,

    M. Alsabah, M. A. Naser, B. M. Mahmmod, S. H. Abdulhussai n, M. R. Eissa, A. Al-Baidhani, N. K. Noordin, S. M. Sait, K. A. Al-Uta ibi, and F. Hashim, “6G wireless communications networks: A compreh ensive survey,” IEEE Access , vol. 9, pp. 148 191–148 243, 2021

    2021

  6. [13]

    Compres- sive sensing techniques for next-generation wireless comm unications,

    Z. Gao, L. Dai, S. Han, I. Chih-Lin, Z. Wang, and L. Hanzo, “Compres- sive sensing techniques for next-generation wireless comm unications,” IEEE Wireless Commun. , vol. 25, no. 3, pp. 144–153, 2018

    2018

  7. [14]

    Compre ssed sensing for wireless communications: Useful tips and trick s,

    J. W. Choi, B. Shim, Y . Ding, B. Rao, and D. I. Kim, “Compre ssed sensing for wireless communications: Useful tips and trick s,” IEEE Commun. Surveys Tutor ., vol. 19, no. 3, pp. 1527–1550, 2017

    2017

  8. [15]

    Comparison of exp licit CSI feedback schemes for 5G new radio,

    R. Ahmed, K. Jayasinghe, and T. Wild, “Comparison of exp licit CSI feedback schemes for 5G new radio,” in Proc. IEEE 89th V eh. Technol. Conf., 2019, pp. 1–5

    2019

  9. [16]

    A review of codebooks for CSI feedback in 5G new radio and beyond,

    Z. Qin and H. Yin, “A review of codebooks for CSI feedback in 5G new radio and beyond,” arXiv preprint arXiv:2302.09222 , 2023

    2023 arXiv

  10. [17]

    Deep learning for mas sive MIMO CSI feedback,

    C.-K. Wen, W.-T. Shih, and S. Jin, “Deep learning for mas sive MIMO CSI feedback,” IEEE Wireless Commun. Lett., vol. 7, no. 5, pp. 748–751, 2018

    2018

  11. [18]

    Multi-resolution CSI feedb ack with deep learning in massive MIMO system,

    Z. Lu, J. Wang, and J. Song, “Multi-resolution CSI feedb ack with deep learning in massive MIMO system,” in Proc. IEEE Int. Conf. Commun. (ICC), 2020, pp. 1–6

    2020

  12. [19]

    Kno wledge- distillation-aided lightweight neural network for massiv e mimo CSI feedback,

    H. Tang, J. Guo, M. Matthaiou, C.-K. Wen, and S. Jin, “Kno wledge- distillation-aided lightweight neural network for massiv e mimo CSI feedback,” in Proc. IEEE 94th V eh. Technol. Conf. , 2021, pp. 1–5

    2021

  13. [20]

    Transnet: Full attention ne twork for CSI feedback in FDD massive MIMO system,

    Y . Cui, A. Guo, and C. Song, “Transnet: Full attention ne twork for CSI feedback in FDD massive MIMO system,” IEEE Wireless Commun. Lett., vol. 11, no. 5, pp. 903–907, 2022

    2022

  14. [21]

    Dilated convolution based CSI feedback compression for massive MIM O sys- tems,

    S. Tang, J. Xia, L. Fan, X. Lei, W. Xu, and A. Nallanathan, “Dilated convolution based CSI feedback compression for massive MIM O sys- tems,” IEEE Trans. V eh. Technol. , vol. 71, no. 10, pp. 11 216–11 221, 2022

    2022

  15. [22]

    Dig-CSI: A distributed and generative model assisted CSI feedback tr aining framework,

    Z. Du, Z. Liu, H. Li, S. Fan, X. Gu, and L. Zhang, “Dig-CSI: A distributed and generative model assisted CSI feedback tr aining framework,” IEEE Wireless Commun. Lett. , pp. 1–1, 2024

    2024

  16. [23]

    ChannelGAN: Deep l earning- based channel modeling and generating,

    H. Xiao, W. Tian, W. Liu, and J. Shen, “ChannelGAN: Deep l earning- based channel modeling and generating,” IEEE Wireless Commun. Lett. , vol. 11, no. 3, pp. 650–654, 2022

    2022

  17. [24]

    A review on gener ative adversarial networks: Algorithms, theory, and applicatio ns,

    J. Gui, Z. Sun, Y . Wen, D. Tao, and J. Y e, “A review on gener ative adversarial networks: Algorithms, theory, and applicatio ns,” IEEE Trans. Knowl. Data Eng. , vol. 35, no. 4, pp. 3313–3332, 2023

    2023

  18. [25]

    Downlink CSI feedback algorithm with deep transfer learni ng for FDD massive MIMO systems,

    J. Zeng, J. Sun, G. Gui, B. Adebisi, T. Ohtsuki, H. Gacani n, and H. Sari, “Downlink CSI feedback algorithm with deep transfer learni ng for FDD massive MIMO systems,” IEEE Trans. Cogn. Commun. Networking , vol. 7, no. 4, pp. 1253–1265, 2021

    2021

  19. [26]

    Model tran smission- based online updating approach for massive MIMO CSI feedbac k,

    B. Zhang, H. Li, X. Liang, X. Gu, and L. Zhang, “Model tran smission- based online updating approach for massive MIMO CSI feedbac k,” IEEE Commun. Lett. , vol. 27, no. 6, pp. 1609–1613, 2023

    2023

  20. [27]

    User-centric onli ne gossip training for autoencoder-based CSI feedback,

    J. Guo, Y . Zuo, C.-K. Wen, and S. Jin, “User-centric onli ne gossip training for autoencoder-based CSI feedback,” IEEE J. Sel. Topics Signal Process., vol. 16, no. 3, pp. 559–572, 2022

    2022

  21. [28]

    Unsupervis ed online learning in deep learning-based massive MIMO CSI feedback,

    Y . Cui, J. Guo, C.-K. Wen, S. Jin, and S. Han, “Unsupervis ed online learning in deep learning-based massive MIMO CSI feedback, ” IEEE Commun. Lett. , vol. 26, no. 9, pp. 2086–2090, 2022

  22. [29]

    Continuous onlin e learning- based CSI feedback in massive MIMO systems,

    X. Zhang, J. Wang, Z. Lu, and H. Zhang, “Continuous onlin e learning- based CSI feedback in massive MIMO systems,” IEEE Commun. Lett. , vol. 28, no. 3, pp. 557–561, 2024

    2024

  23. [30]

    Improving the improved training of wasserstein gans: A consistency term and its dua l effect,

    X. Wei, B. Gong, Z. Liu, W. Lu, and L. Wang, “Improving the improved training of wasserstein gans: A consistency term and its dua l effect,” arXiv preprint arXiv:1803.01541 , 2018

    2018 arXiv

  24. [31]

    Decentralize d learning works: An empirical comparison of gossip learning and federated le arning,

    I. Heged˝ us, G. Danner, and M. Jelasity, “Decentralize d learning works: An empirical comparison of gossip learning and federated le arning,” J. Parallel Distrib. Comput. , vol. 148, pp. 109–124, 2021

    2021

  25. [32]

    Gossip lear ning with linear models on fully distributed data,

    R. Orm’andi, I. Heged˝ us, and M. Jelasity, “Gossip lear ning with linear models on fully distributed data,” Concurrency Computat.: Pract. Exper ., vol. 25, no. 4, pp. 556–571, 2013

    2013

  26. [33]

    Lea rning to continuously optimize wireless resource in episodicall y dynamic en- vironment,

    H. Sun, W. Pu, M. Zhu, X. Fu, T.-H. Chang, and M. Hong, “Lea rning to continuously optimize wireless resource in episodicall y dynamic en- vironment,” in Proc. IEEE Int. Conf. Acoust., Speech Signal Processing (ICASSP), 2021, pp. 4945–4949

    2021

  27. [34]

    Parameterization of the cost 2100 MIMO cha nnel model in indoor scenarios,

    J. Poutanen, K. Haneda, L. Liu, C. Oestges, F. Tufvesson , and P . V ainikainen, “Parameterization of the cost 2100 MIMO cha nnel model in indoor scenarios,” in Proc. of the 5th European Conference on Antennas and Propagation (EUCAP) , 2011, pp. 3606–3610

    2011

  28. [35]

    DeepMIMO: A generic deep learning datas et for mil- limeter wave and massive MIMO applications,

    A. Alkhateeb, “DeepMIMO: A generic deep learning datas et for mil- limeter wave and massive MIMO applications,” in Proc. of Information Theory and Applications W orkshop (ITA) , San Diego, CA, Feb. 2019

    2019

  29. [36]

    Fully convolutional neural network-based CSI limited feedback f or FDD massive MIMO systems,

    G. Fan, J. Sun, G. Gui, H. Gacanin, B. Adebisi, and T. Ohts uki, “Fully convolutional neural network-based CSI limited feedback f or FDD massive MIMO systems,” IEEE Trans. Cogn. Commun. Networking , vol. 8, no. 2, pp. 672–682, 2022

    2022

  30. [37]

    Activatio n functions in deep learning: A comprehensive survey and benchmark,

    S. R. Dubey, S. K. Singh, and B. B. Chaudhuri, “Activatio n functions in deep learning: A comprehensive survey and benchmark,” Neurocom- puting, vol. 503, pp. 92–108, 2022

    2022

  31. [38]

    Communication-ef ficient personalized federated edge learning for massive MIMO CSI f eedback,

    Y . Cui, J. Guo, C.-K. Wen, and S. Jin, “Communication-ef ficient personalized federated edge learning for massive MIMO CSI f eedback,” IEEE Trans. Wireless Commun. , pp. 1–1, 2023

    2023

  32. [39]

    Study on channel model for frequencies from 0.5 to 100 GHz,

    G. TR 38.901, “Study on channel model for frequencies from 0.5 to 100 GHz,” Mar. 2017. [Online]. Available: https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx?specifica

    2017

  33. [40]

    Cross-entropy loss func tions: Theoretical analysis and applications,

    A. Mao, M. Mohri, and Y . Zhong, “Cross-entropy loss func tions: Theoretical analysis and applications,” in Proc. Int. conf. Mach. Learn. (ICML), 2023, pp. 23 803–23 828

    2023

  34. [41]

    Improved training of wasserstein gans,

    I. Gulrajani, F. Ahmed, M. Arjovsky, V . Dumoulin, and A. C. Courville, “Improved training of wasserstein gans,” Advances in neural information processing systems, vol. 30, 2017

    2017

  35. [42]

    Automatic different iation in pytorch,

    A. Paszke, S. Gross, S. Chintala, G. Chanan, E. Y ang, Z. D eVito, Z. Lin, A. Desmaison, L. Antiga, and A. Lerer, “Automatic different iation in pytorch,” 2017

    2017

  36. [43]

    Compariso n of methods generalizing max-and average-pooling,

    F. Bieder, R. Sandk”uhler, and P . C. Cattin, “Compariso n of methods generalizing max-and average-pooling,” arXiv preprint arXiv:2103.01746, 2021

    2021 arXiv

  37. [44]

    Deep learning for effi cient CSI feedback in massive MIMO: Adapting to new environments and s mall datasets,

    Z. Liu, L. Wang, L. Xu, and Z. Ding, “Deep learning for effi cient CSI feedback in massive MIMO: Adapting to new environments and s mall datasets,” IEEE Trans. Wireless Commun. , pp. 1–1, 2024

    2024

  38. [45]

    Decen tralized federated learning of deep neural networks on non-iid data,

    N. Onoszko, G. Karlsson, O. Mogren, and E. L. Zec, “Decen tralized federated learning of deep neural networks on non-iid data, ” arXiv preprint arXiv:2107.08517, 2021

    2021 arXiv

  39. [47]

    Torchviz,

    Szagoruyko, “Torchviz,” 2021, accessed: 2024-07-27. [Online]. Available: https://github.com/szagoruyko/pytorchviz

    2021

  40. [48]

    Federated generative adversarial learning,

    C. Fan and P . Liu, “Federated generative adversarial learning,” CoRR, vol. abs/2005.03793, 2020. [Online]. Available: https://arxiv.org/abs/2005.03793

    2005 arXiv

  41. [49]

    Gossip learni ng as a decentral- ized alternative to federated learning,

    I. Heged˝ us, G. Danner, and M. Jelasity, “Gossip learni ng as a decentral- ized alternative to federated learning,” in Proc. IFIP Int. Conf. Distrib. Appl. Interoperable Syst. , 2019, pp. 74–90

    2019

  42. [50]

    An efficient deep learning framework for low rate massive MIMO CSI reporting,

    Z. Liu, L. Zhang, and Z. Ding, “An efficient deep learning framework for low rate massive MIMO CSI reporting,” IEEE Trans. Commun., vol. 68, no. 8, pp. 4761–4772, 2020

    2020

  43. [51]

    Distributed d eep convo- lutional compression for massive MIMO CSI feedback,

    M. B. Mashhadi, Q. Y ang, and D. G¨ und¨ uz, “Distributed d eep convo- lutional compression for massive MIMO CSI feedback,” IEEE Trans. Wireless Commun., vol. 20, no. 4, pp. 2621–2633, 2021

    2021

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.