Model Splitting Enhanced Communication-Efficient Federated Learning for CSI Feedback

Gaojie Chen; Haijun Zhang; Victor C. M. Leung; Xiaoyi Fan; Xiping Hu; Yanjie Dong

REVIEW 3 major objections 4 minor 29 references

Model Splitting Enhanced Communication-Efficient Federated Learning for CSI Feedback

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Splitting the CSI-feedback autoencoder between user devices and the base station lets federated training reach the same test loss with less than half the exchanged parameters.

desk verdict CSILocal's headline communication saving is an artifact of counting one boundary tensor against full-model baselines; the paper has a useful system design but its central claim reverses when you count what its own Remark 2 lists. read the letter →

arxiv 2506.04113 v1 pith:AITNCP3X submitted 2025-06-04 eess.SP

Yanjie Dong , Haijun Zhang , Gaojie Chen , Xiaoyi Fan , Victor C. M. Leung , Xiping Hu This is my paper

classification eess.SP

keywords CSIfeedbackfederatedlearningmodelsplittingcommunicationefficiencypipelineparallelismmassiveMIMOautoencodercompression

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 prove that federated training of a CSI-feedback autoencoder can be made communication-efficient by splitting the model so each user device keeps the encoder and decoder head while the base station holds the decoder tail. Instead of exchanging full model parameters, devices send only the small smashed-data tensors at the split boundaries during forward and backward passes. The authors report that this reaches the same testing loss (about 1.3e-3) with 19.66 billion exchanged parameters, versus 47.06 billion for FedAvg and 38.24 billion for FedProx. They also show that splitting the decoder tail across multiple GPUs at the base station cuts per-iteration wall-clock time. If correct, this lowers the communication bottleneck that limits federated CSI feedback in massive MIMO systems.

What carries the argument

The load-bearing object is the three-way model split with boundary data exchange: encoder $F_{n,1}$ and decoder head $F_{n,3}$ stay at each UE, and decoder tail $F_2$ sits at the BTS. During forward passes the UEs upload smashed data $F_{n,1}(H_{n,m})$ and receive $F_2(F_{n,1}(H_{n,m}))$; during backward passes they exchange the corresponding gradients. Because the boundary dimensions satisfy $c_1 \ll \min\{d_1,d_2\}$ and $c_2 \ll \min\{d_2,d_3\}$, the communication volume is governed by the smashed-data sizes rather than the model sizes. The pipeline module further splits $F_2$ into sub-models on multiple GPUs to overlap forward and backward computation across micro-batches.

What would settle it

Measure the true bytes and wall-clock time on a live wireless testbed with heterogeneous user devices, counting both uplink smashed data and downlink decoder-tail outputs plus protocol overhead; if at equal test loss the total transmitted bytes and training latency are not below the full-parameter-exchange baselines, the communication-efficiency claim is falsified.

Watch

Extended reading notes

Core claim

The central claim is that CSILocal, a model-splitting algorithm for federated CSI feedback, significantly reduces the number of parameters exchanged between UEs and the BTS during training while keeping reconstruction accuracy essentially unchanged. The CSI autoencoder is divided into three modules: the encoder $F_{n,1}$ and decoder head $F_{n,3}$ reside at each UE, and the decoder tail $F_2$ is shared at the BTS. Communication is limited to the boundary activation tensors and their gradients, whose dimensions $c_1$ and $c_2$ are chosen much smaller than the parameter counts of the modules. In numerical experiments with 10 UEs, CSILocal reaches a testing loss of about $1.3\times10^{-3}$ with 19.66 billion exchanged parameters, compared with 47.06 billion for FedAvg and 38.24 billion for FedProx. A pipeline module that splits the decoder tail across GPUs reduces the wall-clock duration per iteration by up to about 22% at batch size 800.

Load-bearing premise

The central assumption is that all user devices have identical processing capabilities, so stragglers can be ignored, and that communication cost is faithfully measured by the number of exchanged parameters rather than actual transmitted bits, latency, or the energy of sending smashed data back to devices.

Editorial extensions

If this is right

At equal testing loss, CSILocal exchanges about 58% fewer parameters than FedAvg and about 49% fewer than FedProx in the reported 10-UE setup.
The communication cost per iteration scales with the boundary dimensions $c_1$ and $c_2$, not with the total model size, so larger models do not automatically imply heavier federated traffic.
Pipeline parallelism in the decoder tail shortens per-iteration wall-clock time, with the reported reduction growing with mini-batch size and number of UEs.
Higher encoding dimensions compensate for non-IID-induced reconstruction degradation, at the price of larger smashed-data exchanges.
Keeping the original CSI matrices at each UE preserves local data privacy in the sense that raw CSI is not transmitted.

Reading between the lines

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

The communication accounting counts exchanged parameters, not transmitted bits; a fairer comparison would include the downlink smashed data sent from the BTS to each UE, digital-modulation overhead, and retransmissions, which could shrink or reverse the reported gap.
The privacy argument that raw CSI never leaves the UE does not automatically cover the smashed-data tensors, which are learned features of the channel and may encode location or identity information; a membership-inference test on the boundary activations would probe this.
The same split-and-exchange pattern could be applied to other split points or to hierarchical aggregators, and comparing against gradient-compression baselines (quantization, sparsification) would show whether model splitting or compression is the dominant source of the savings.

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.

Desk Editor's Note

read the letter

Yeah, I read CSILocal. The headline claim—19.66B exchanged parameters versus 47.06B for FedAvg—doesn't survive contact with the paper's own Remark 2. Remark 2 says each iteration exchanges four boundary tensors per sample: encoder output up, decoder-tail output down, and two gradient tensors in the backward pass. With N=10, batch 800, c1=c2=256, one forward tensor alone is 10*800*256*20,000 = 40.96B over the full 20k iterations. The quoted 19.66B works out to exactly one tensor per sample for about 9,600 iterations. So they counted one direction, one tensor, and compared that against full-model parameter counts for the baselines. Count all four tensors and the total is around 78B at 9,600 iterations, or 164B at 20k—worse than FedAvg's 47.06B. The central communication-efficiency claim is an artifact.

That's a shame because the paper has some genuine content. The model-splitting architecture for CSI feedback is a reasonable extension of split learning to a domain where it hasn't been applied. The pipeline parallelism at the BTS is real, and the wallclock measurements (4% to 22% speedup depending on batch size) are concrete. The convergence plots across indoor/outdoor/non-IID are useful. The authors also cite the relevant split-learning and personalized-FL literature.

But the accounting error is load-bearing. The paper's own Remark 2 contradicts the numbers in Section IV.D. Beyond that, the evaluation has other gaps: no error bars, a single operating point (testing loss 1.3e-3) for the big comparison, idealized assumptions about identical UEs and error-free digital transmission, and no code. The pipeline speedup is measured on a small scale (2 micro-batches).

Who's this for? Researchers working on communication-efficient federated CSI feedback would want to look at the system design, but they'd have to redo the communication accounting before trusting any of the conclusions. I would not cite this as evidence of communication efficiency. I'd still send it to a serious referee, because the error is subtle enough that an expert could catch it and the authors might have a defensible version (e.g., with quantization or asymmetric counting) that could be fixed. But my own verdict: the main claim fails as written.

Referee Report

3 major / 4 minor

Summary. The paper proposes CSILocal, a split-learning/federated training method for deep-learning-based CSI feedback in massive MIMO. The autoencoder is divided into an encoder and decoder head kept at each UE and a decoder tail kept at the BTS; UEs and the BTS exchange only boundary activation and gradient tensors rather than full model parameters. The decoder tail is further split across multiple BTS GPUs with pipeline parallelism to reduce wall-clock training time. Experiments on CsiNet-style data compare CSILocal with FedAvg, FedProx, and personalized variants under indoor, outdoor, and non-IID settings, reporting testing loss versus exchanged parameters and per-iteration pipeline timings.

Significance. If the communication-efficiency claim were correct, the paper would offer a practical way to train CSI feedback networks in federated settings with reduced uplink traffic, and the pipeline parallelism at the BTS is a sensible engineering idea. The paper does provide numerical comparisons with six benchmark algorithms, includes a non-IID data setup, and reports measured pipeline timings. However, the central quantitative claim is undermined by an inconsistent accounting of exchanged tensors: the reported 19.66 billion exchanged parameters does not match the paper's own description of the boundary tensors exchanged per iteration. The pipeline speedup percentages in Table I are also miscalculated. Because the main contribution is communication efficiency, these issues are load-bearing and the manuscript in its current form does not support its principal claims.

major comments (3)

[Section IV.D, Remark 2] The headline number "19.66 billion exchanged parameters" is not consistent with the paper's own specification. Remark 2 states that for each sample and iteration, four boundary tensors are exchanged: the encoder output of dimension c1, the decoder-tail output of dimension c2, and two matching gradient tensors in the backward pass. With N=10 UEs, a per-UE mini-batch size of 800, c1=c2=256, and K=20,000 iterations (Section IV.A), a single forward boundary tensor alone carries 10 x 800 x 256 x 20,000 = 4.096 x 10^10 scalars, and all four boundary tensors carry 1.6384 x 10^11 scalars. The reported 19.66 x 10^9 equals 10 x 800 x 256 x 9,600, i.e., one scalar per sample per UE for only 9,600 iterations, which does not correspond to any stated counting rule. Even the downlink decoder-tail output alone over 20,000 iterations is 4.096 x 10^10 scalars, exceeding the 3.824 x 10^10 claimed for FedProx. No quantization, compression, or subsampling of boundary tensors is described, so the discrepancy cannot be attributed to such mechanisms. The comparison against 47.06 billion and 38.24 billion parameters is therefore unsupported.
[Section II.B and Section IV.D] The communication-cost metric counts numbers of exchanged parameters or scalars but does not account for actual bits or wireless resource consumption. In a real link, every exchanged tensor, including the BTS-to-UE transmission of the decoder-tail output [F2 composed with F_{n,1}(H_{n,m})], consumes time-frequency resources and energy. The paper's metric assigns zero communication cost to this downlink transmission and does not include the overhead of digital transmission even though Section II.B assumes digital transmission under channel capacity. Since the paper's central contribution is "communication efficiency," the comparison should report total uplink and downlink bits per round with a stated precision or quantization format, and should include the decoder-tail output and backward gradient tensors in the accounting.
[Section IV.B, Table I] The claimed wall-clock reductions are computed with the wrong denominator. For indoor data, mini-batch 800, and encoding dimension 256, the duration drops from 0.1160 s to 0.0952 s per iteration; the relative reduction is (0.1160 - 0.0952)/0.1160 = 17.9%, not the stated 21.85%. The paper uses the pipeline time as the denominator, which inflates the percentages. The same issue affects the other entries, e.g., 0.0583 to 0.0560 is a 3.9% reduction, not 4.11%. The conclusion that pipeline parallelism reduces training time is directionally correct, but the reported magnitudes are inaccurate and should be corrected.

minor comments (4)

[Algorithm 1, line 8; Section IV.A] There are several typos and textual slips: "calcualtes" should be "calculates," "normlized" should be "normalized," and Section IV.D contains the duplicated phrase "can can."
[Section II.B] The term "smashed data" is used repeatedly but is never formally defined; the authors should define it as the boundary activation tensors exchanged at the split points, and consistently use that terminology.
[Equation (2)] The local objective f_n(θ_{n,1}, θ_2, θ_{n,3}; H_{n,k}) uses index k internally while the outer sum runs over m; please align the indexing to avoid confusion.
[Section III and Figure 3] The relationship between mini-batches, micro-batches, and the four-GPU pipeline is described only briefly; a clearer statement of how data are divided and how gradient aggregation across micro-batches is performed would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the communication-efficiency claim is an empirical comparison against external benchmarks, and self-citations are background-only.

full rationale

The paper's central claim is that the proposed CSILocal algorithm reduces the number of exchanged parameters during federated CSI feedback training compared to FedAvg and FedProx. This claim is supported by numerical experiments using the CsiNet dataset and standard federated baselines, not by a derivation from a fitted parameter or a self-citation chain. The model-splitting architecture is explicitly described (encoder/decoder tail/decoder head), and the boundary dimensions are chosen as hyperparameters. The reported exchanged-parameter counts are empirical results read from convergence curves at a fixed testing loss. No prediction is constructed from a fit to the data, and no uniqueness theorem or prior work by the authors is invoked to forbid alternatives. The self-citations in the paper ([13], [15]) are used only as background references for downstream tasks and precoder design; they are not load-bearing for the communication-efficiency claim. While the communication accounting may warrant scrutiny (e.g., whether all boundary tensors are counted consistently relative to baseline parameter counts), that is a question of measurement consistency, not circularity. The derivation chain does not reduce to its inputs by definition or fit.

Assumptions & free parameters 1 free parameters · 2 assumptions · 0 invented entities

The main free parameters are the boundary dimensions c1 and c2 and the pipeline micro-batch count, all chosen by the authors. There are no invented entities. The domain assumptions (negligible transmission errors and identical UE capabilities) are strong but standard in academic papers on federated learning.

free parameters (1)

c1 and c2 (boundary dimensions) = 256 or 512
The boundary dimensions c1 and c2 are choices that control the communication cost and reconstruction accuracy. They are not fitted but are hyperparameters set by the authors. They directly affect the amount of smashed data exchanged, so they are key to the claimed communication reduction.

assumptions (2)

domain assumption The communication link between each UE and BTS experiences negligible error rate under digital transmission (Section II.B).
This assumption is stated to justify ignoring transmission errors. It is an idealization that may not hold in real wireless systems.
domain assumption All UEs have identical processing capabilities and stragglers can be ignored (Section II.B).
This simplifies the analysis but is unrealistic in practice. Stragglers can dominate federated learning wallclock time.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Model Splitting Enhanced Communication-Efficient Federated Learning for CSI Feedback." pith.science (2026). https://pith.science/paper/AITNCP3X

@misc{pith2026250604113,
  author       = {Pith},
  title        = {Pith review of: Model Splitting Enhanced Communication-Efficient Federated Learning for CSI Feedback},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AITNCP3X}},
  note         = {Machine review of arXiv:2506.04113}
}

read the original abstract

Recent advancements have introduced federated machine learning-based channel state information (CSI) compression before the user equipments (UEs) upload the downlink CSI to the base transceiver station (BTS). However, most existing algorithms impose a high communication overhead due to frequent parameter exchanges between UEs and BTS. In this work, we propose a model splitting approach with a shared model at the BTS and multiple local models at the UEs to reduce communication overhead. Moreover, we implant a pipeline module at the BTS to reduce training time. By limiting exchanges of boundary parameters during forward and backward passes, our algorithm can significantly reduce the exchanged parameters over the benchmarks during federated CSI feedback training.

Figures

Figures reproduced from arXiv: 2506.04113 by the authors.

**Figure 2.** An illustration of the pipeline parallelism for [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 4.** An illustration of used deep neural network that have encoder, [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗

Figures from the paper (2 more)

**Figure 5.** Figure 5: The impact of compression ratios on the CSI reconstruction [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]

**Figure 6.** Figure 6: The convergence of testing loss over the communication overhead for the different environments and encoding dimensions. [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

29 extracted references · 23 canonical work pages

[1]

Reconfigurable-intelligent-surface-assisted B5G/6G wireless communications: Challenges, solution, and future opportunities,

Z. Chen, G. Chen, J. Tang, S. Zhang, D. K. So, O. A. Dobre, K.- K. Wong, and J. Chambers, “Reconfigurable-intelligent-surface-assisted B5G/6G wireless communications: Challenges, solution, and future opportunities,”IEEE Commun. Mag., vol. 61, no. 1, pp. 16–22, Jan. 2023

work page 2023
[2]

Edge learning for B5G networks with distributed signal processing: Semantic communication, edge computing, and wireless sensing,

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

work page 2023
[3]

LLM4CP: Adapting large language models for channel prediction,

B. Liu, X. Liu, S. Gao, X. Cheng, and L. Yang, “LLM4CP: Adapting large language models for channel prediction,”J. Commun. Inform. Netw., vol. 9, no. 2, pp. 113–125, June 2024

work page 2024
[4]

Massive-MIMO iterative channel estimation and decoding (MICED) in the uplink,

D. Verenzuela, E. Bj ¨ornson, X. Wang, M. Arnold, and S. ten Brink, “Massive-MIMO iterative channel estimation and decoding (MICED) in the uplink,”IEEE Trans. Commun., vol. 68, no. 2, pp. 854–870, Feb. 2020

work page 2020
[5]

A matching- based pilot assignment algorithm for cell-free massive MIMO networks,

Y . Gao, H. Hu, J. Chen, X. Wang, X. Chu, and J. Zhang, “A matching- based pilot assignment algorithm for cell-free massive MIMO networks,” IEEE Trans. Veh. Technol., vol. 73, no. 1, pp. 1453–1457, Jan. 2024

work page 2024
[6]

Overview of deep learning- based CSI feedback in massive MIMO systems,

J. Guo, C.-K. Wen, S. Jin, and G. Y . Li, “Overview of deep learning- based CSI feedback in massive MIMO systems,”IEEE Trans. Commun., vol. 70, no. 12, pp. 8017–8045, Dec. 2022

2022
[7]

Massive MIMO in 5G: How beamforming, codebooks, and feedback enable larger arrays,

R. M. Dreifuerst and R. W. Heath, “Massive MIMO in 5G: How beamforming, codebooks, and feedback enable larger arrays,”IEEE Commun. Mag., vol. 61, no. 12, pp. 18–23, Dec. 2023

work page 2023
[8]

TSG-RAN WG1 #89: R1-1709232 WF on type I and II CSI codebooks,

3GPP, “TSG-RAN WG1 #89: R1-1709232 WF on type I and II CSI codebooks,” Tech. Rep., 2017

work page 2017

Show all 29 references

[9]

Deep learning and compressive sensing-based CSI feedback in FDD massive MIMO systems,

P. Liang, J. Fan, W. Shen, Z. Qin, and G. Y . Li, “Deep learning and compressive sensing-based CSI feedback in FDD massive MIMO systems,”IEEE Trans. Veh. Technol., vol. 69, no. 8, pp. 9217–9222, Aug. 2020

2020
[10]

Deep learning for massive MIMO CSI feedback,

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

2018
[11]

CLNet: Complex input lightweight neural network designed for massive MIMO CSI feedback,

S. Ji and M. Li, “CLNet: Complex input lightweight neural network designed for massive MIMO CSI feedback,”IEEE Wireless Commun. Lett., vol. 10, no. 10, pp. 2318–2322, Oct. 2021

2021
[12]

Interleave frequency division multiplexing,

Y . Chi, L. Liu, Y . Ge, X. Chen, Y . Li, and Z. Zhang, “Interleave frequency division multiplexing,”IEEE Wireless Commun. Lett., vol. 13, no. 7, pp. 1963–1967, July 2024

1963
[13]

An online zero-forcing precoder for weighted sum-rate maximization in green CoMP systems,

Y . Dong, H. Zhang, J. Li, F. R. Yu, S. Guo, and V . C. M. Leung, “An online zero-forcing precoder for weighted sum-rate maximization in green CoMP systems,”IEEE Trans. Wireless Commun., vol. 21, no. 9, pp. 7566–7581, Sept. 2022

2022
[14]

Connectivity analysis for V2I communications in cognitive vehicular networks,

X. Li, L. Liu, R. Zhou, N. Zhang, C. Wu, M. Atiquzzaman, and M. Guizani, “Connectivity analysis for V2I communications in cognitive vehicular networks,”IEEE Trans. Veh. Technol., to be published, 2024, doi: 10.1109/TVT.2024.3481300

2024
[15]

Accelerating wireless federated learning via Nesterov’s momen- tum and distributed principal component analysis,

Y . Dong, L. Wang, J. Wang, X. Hu, H. Zhang, F. R. Yu, and V . C. M. Leung, “Accelerating wireless federated learning via Nesterov’s momen- tum and distributed principal component analysis,”IEEE Trans. Wireless Commun., vol. 23, no. 6, pp. 5938–5952, June 2024

2024
[16]

Transformer-empowered 6G intelligent networks: From massive MIMO processing to semantic communication,

Y . Wang, Z. Gao, D. Zheng, S. Chen, D. G ¨und¨uz, and H. V . Poor, “Transformer-empowered 6G intelligent networks: From massive MIMO processing to semantic communication,”IEEE Wireless Commun., vol. 30, no. 6, pp. 127–135, Dec. 2023

2023
[17]

Foundation models and intelligent decision-making: Progress, challenges, and perspectives,

J. Huanget al., “Foundation models and intelligent decision-making: Progress, challenges, and perspectives,”The Innovation, vol. 6, no. 6, May 2025

2025
[18]

Dual-propagation-feature fusion enhanced neural CSI compression for massive MIMO,

S. Zhang, W. Xu, S. Jin, X. You, D. W. K. Ng, and L.-C. Wang, “Dual-propagation-feature fusion enhanced neural CSI compression for massive MIMO,”IEEE Trans. Commun., vol. 71, no. 9, pp. 5182–5198, Sept. 2023

2023
[19]

Deep learning for distributed channel feedback and multiuser precoding in FDD massive MIMO,

F. Sohrabi, K. M. Attiah, and W. Yu, “Deep learning for distributed channel feedback and multiuser precoding in FDD massive MIMO,” IEEE Trans. Wireless Commun., vol. 20, no. 7, pp. 4044–4057, July 2021

2021
[20]

Data-driven deep learning based hybrid beamforming for aerial massive MIMO-OFDM systems with implicit CSI,

Z. Gao, M. Wu, C. Hu, F. Gao, G. Wen, D. Zheng, and J. Zhang, “Data-driven deep learning based hybrid beamforming for aerial massive MIMO-OFDM systems with implicit CSI,”IEEE J. Sel. Areas Commun., vol. 40, no. 10, pp. 2894–2913, Oct. 2022

2022
[21]

Distributed deep convo- lutional compression for massive MIMO CSI feedback,

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

2021
[22]

Deep learning- based CSI feedback for RIS-assisted multi-user systems,

J. Guo, X. Yang, C.-K. Wen, S. Jin, and G. Y . Li, “Deep learning- based CSI feedback for RIS-assisted multi-user systems,”arXiv preprint arXiv:2003.03303v4, Mar. 2024

2003 arXiv
[23]

Communication-efficient personalized federated edge learning for massive MIMO CSI feedback,

Y . Cui, J. Guo, C.-K. Wen, and S. Jin, “Communication-efficient personalized federated edge learning for massive MIMO CSI feedback,” IEEE Trans. Wireless Commun., vol. 23, no. 7, pp. 7362–7375, July 2024

2024
[24]

User-centric online gossip training for autoencoder-based CSI feedback,

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

2022
[25]

Wireless federated learning over resource-constrained networks: Digital versus analog transmissions,

J. Yao, W. Xu, Z. Yang, X. You, M. Bennis, and H. V . Poor, “Wireless federated learning over resource-constrained networks: Digital versus analog transmissions,”IEEE Trans. Wireless Commun., vol. 23, no. 10, pp. 14 020–14 036, Oct. 2024

2024
[26]

A simple convergence proof of Adam and Adagrad,

A. D ´efossez, L. Bottou, F. Bach, and N. Usunier, “A simple convergence proof of Adam and Adagrad,”Trans. Mach. Learn. Research, 2022

2022
[27]

Gpipe: Efficient training of giant neural networks using pipeline parallelism,

Y . Huang, Y . Cheng, A. Bapna, O. Firat, D. Chen, M. Chen, H. Lee, J. Ngiam, Q. V . Le, Y . Wuet al., “Gpipe: Efficient training of giant neural networks using pipeline parallelism,” inProc. NeurIPS, vol. 32, Vancouver, BC, Canada, Dec. 2019

2019
[28]

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

Z. Lu, J. Wang, and J. Song, “Multi-resolution CSI feedback with deep learning in massive MIMO system,” inProc. IEEE ICC, Dublin, Ireland, June 2020, pp. 1–6

2020
[29]

Federated learning with partial model personalization,

K. Pillutla, K. Malik, A.-R. Mohamed, M. Rabbat, M. Sanjabi, and L. Xiao, “Federated learning with partial model personalization,” in Proc. ICML, Baltimore, MD, USA, July 2022, pp. 17 716–17 758

2022

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Reconfigurable-intelligent-surface-assisted B5G/6G wireless communications: Challenges, solution, and future opportunities,

Z. Chen, G. Chen, J. Tang, S. Zhang, D. K. So, O. A. Dobre, K.- K. Wong, and J. Chambers, “Reconfigurable-intelligent-surface-assisted B5G/6G wireless communications: Challenges, solution, and future opportunities,”IEEE Commun. Mag., vol. 61, no. 1, pp. 16–22, Jan. 2023

work page 2023

[2] [2]

Edge learning for B5G networks with distributed signal processing: Semantic communication, edge computing, and wireless sensing,

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

work page 2023

[3] [3]

LLM4CP: Adapting large language models for channel prediction,

B. Liu, X. Liu, S. Gao, X. Cheng, and L. Yang, “LLM4CP: Adapting large language models for channel prediction,”J. Commun. Inform. Netw., vol. 9, no. 2, pp. 113–125, June 2024

work page 2024

[4] [4]

Massive-MIMO iterative channel estimation and decoding (MICED) in the uplink,

D. Verenzuela, E. Bj ¨ornson, X. Wang, M. Arnold, and S. ten Brink, “Massive-MIMO iterative channel estimation and decoding (MICED) in the uplink,”IEEE Trans. Commun., vol. 68, no. 2, pp. 854–870, Feb. 2020

work page 2020

[5] [5]

A matching- based pilot assignment algorithm for cell-free massive MIMO networks,

Y . Gao, H. Hu, J. Chen, X. Wang, X. Chu, and J. Zhang, “A matching- based pilot assignment algorithm for cell-free massive MIMO networks,” IEEE Trans. Veh. Technol., vol. 73, no. 1, pp. 1453–1457, Jan. 2024

work page 2024

[6] [6]

Overview of deep learning- based CSI feedback in massive MIMO systems,

J. Guo, C.-K. Wen, S. Jin, and G. Y . Li, “Overview of deep learning- based CSI feedback in massive MIMO systems,”IEEE Trans. Commun., vol. 70, no. 12, pp. 8017–8045, Dec. 2022

2022

[7] [7]

Massive MIMO in 5G: How beamforming, codebooks, and feedback enable larger arrays,

R. M. Dreifuerst and R. W. Heath, “Massive MIMO in 5G: How beamforming, codebooks, and feedback enable larger arrays,”IEEE Commun. Mag., vol. 61, no. 12, pp. 18–23, Dec. 2023

work page 2023

[8] [8]

TSG-RAN WG1 #89: R1-1709232 WF on type I and II CSI codebooks,

3GPP, “TSG-RAN WG1 #89: R1-1709232 WF on type I and II CSI codebooks,” Tech. Rep., 2017

work page 2017