Attention-Infused Autoencoder for Massive MIMO CSI Compression

Kangzhi Lou; Xiping Wu

REVIEW 3 major objections 5 minor 31 references

Attention-Infused Autoencoder for Massive MIMO CSI Compression

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read An attention-fused autoencoder compresses massive MIMO channel feedback more accurately than prior learned methods, and mixed training lets one model span indoor and outdoor scenarios.

desk verdict Solid incremental AE paper with clean ablations, but the headline gain over ACRNet is likely a bit-depth artifact until quantization is equalized. read the letter →

arxiv 2504.12440 v1 pith:QTKB6RER submitted 2025-04-16 eess.SP

Kangzhi Lou , Xiping Wu This is my paper · ORCID

classification eess.SP

keywords CSIcompressionmassiveMIMOautoencoderattentionmechanismchannelstateinformationfeedbackcross-scenariogeneralizationmixedtraining6G

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 proposes AiANet, an autoencoder for compressing downlink channel state information in massive MIMO systems. It claims that by fusing channel-wise and spatial attention in parallel, adding locally-aware self-attention, and gating dense skip connections, AiANet reconstructs CSI with lower normalized mean squared error than existing learned compressors, up to 3.42 dB better than ACRNet at a 1/4 compression ratio. The paper further claims that a mixed-training scheme, where indoor and outdoor samples are trained together, gives the model genuine cross-scenario generalizability, unlike separate training and direct transfer. If correct, this offers a route to a single universal CSI compressor instead of multiple scenario-specific models.

What carries the argument

Three modules carry the argument. Hybrid Attention-Gated Fusion (HAGF) computes channel and spatial masks in parallel and cross-gates the main and residual feature streams, so the network can adaptively weight frequency-selective versus scattering-rich structure. Locally-Aware Self-Attention (LASA) adds a depthwise-convolution local path to a scaled dot-product global attention path and fuses them with a learnable gate, capturing both fine delay-bin structure and long-range spatial dependence. Gated Dense Connections (GDC) route each encoder block's input to all later blocks through trainable sigmoid gates, allowing the network to suppress or amplify multi-scale feature flow. The decoder mirrors the attention modules once before a sigmoid output.

What would settle it

Run AiANet and ACRNet on the same indoor and outdoor data but quantize the codeword to one to four bits or add uplink noise; if the NMSE advantage at a 1/4 ratio falls below the reported spread, or if mixed-trained AiANet no longer beats scenario-specific ACRNet models, the central generalization claim would be contradicted.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that an encoder-decoder built from hybrid attention-gated fusion, locally-aware self-attention, and gated dense connections combines two things earlier autoencoder compressors treated separately: accurate reconstruction within a single scenario and usable performance across scenarios. In the reported tables, AiANet reaches -35.44 dB NMSE indoors at a 1/4 ratio versus -32.02 dB for ACRNet, and under mixed training it reaches -21.09 dB indoors and -13.85 dB outdoors, versus -16.33 dB and -10.07 dB for ACRNet under the same scheme. The authors interpret the mixed-training result as evidence that the model learns shared CSI structure across channel environments rather than overfitting one scenario.

Load-bearing premise

The performance gap is measured on a simulated indoor/outdoor channel dataset with a noise-free, quantization-free feedback path, so the claim that one model generalizes across real environments rests on those simulation conditions carrying over to practice.

Editorial extensions

If this is right

If the mixed-training result holds, operators could deploy one encoder-decoder for both indoor and outdoor coverage instead of storing several scenario-specific models.
The NMSE gains translate into more accurate angular-delay CSI at the base station, which should improve precoding and spectral efficiency in FDD massive MIMO systems.
Training on a balanced mix of scenarios appears to buy cross-scenario robustness at the cost of some single-scenario peak accuracy, a trade-off the paper quantifies at each compression ratio.
The accuracy advantage over ACRNet narrows as compression becomes more aggressive, suggesting the attention machinery buys less when the codeword is very small.

Reading between the lines

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

The paper leaves quantization and feedback noise out of the experiments; since ACRNet was designed as a binarized network, a fair one-bit or few-bit codeword comparison could narrow or change the reported margins.
The same HAGF/LASA modules could be applied to related channel-learning tasks, such as CSI prediction, positioning, or direct precoder mapping, where fusing global and local radio structure matters.
A testable extension would be to train one AiANet on mixed urban and rural channels or across different frequency bands; the paper's logic predicts smaller per-scenario degradation than retraining separate compressors.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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

Colleague,

Quick take: this is a competent incremental AE paper with a real comparison-fairness problem that could change the headline number. ACRNet is a binarized network; the experiments here evaluate full-precision AiANet against ACRNet's native quantized/binarized results without holding bit rate constant. The 3.42 dB gain looks like a bit-depth advantage until shown otherwise.

What's genuinely good: the architecture combination—HAGF, LASA, GDC—is new, and the ablations are clean. Each component is removed or replaced independently, and the tables show consistent gains across compression ratios. The mixed-training scheme is a useful addition; it's the first cross-scenario result in this comparison set, and it shows real generalization gains over separate training. The writing is clear, the benchmark is standard, and the reported numbers are internally consistent (the 3.42 dB matches Table V).

Soft spots, in order of severity:

1. Quantization fairness. The system model says feedback is "often treated as noise-free in initial designs," but ACRNet's design is explicitly binarized/quantized. Every experiment uses a noiseless, unquantized codeword. If the ACRNet entries are its binarized results and AiANet uses 32-bit floats, the comparison is not apples-to-apples—at equal η, AiANet sends many more bits per codeword. This is the load-bearing issue. The authors need to either quantize AiANet to the same bit rate, or run ACRNet in full precision, or report performance across bit budgets.

2. No seeds or error bars. Every table is a single run. At η=1/32 and 1/64, the margins over ACRNet shrink to roughly 0.1-0.8 dB, which is within typical training noise for these models. Without multiple seeds, the small-margin claims are not established.

3. Missing baselines. FSAMNet and SwinCFNet are cited in the introduction and listed in Table I, but not in any performance comparison. Given the paper claims to outperform "existing AE-based methods such as ACRNet," omitting the two most recent attention-based baselines is a gap.

4. Minor inconsistency. Section IV says AdamW weight decay 1e-4; Section V-A says weight decay disabled (λ=0). One of these is wrong, and it matters for reproducibility.

The central claim—that the proposed modules help—is probably true; the ablations support it. But the headline gain over ACRNet is not yet verified under equal bit rates. This paper deserves a serious referee: the architecture is reasonable, the study is useful, and the comparison issue is fixable. I'd send it to review, but the revision must include quantization-aware evaluation and error bars.

It's worth a place in your reading group for the cross-scenario training discussion alone.

Referee Report

3 major / 5 minor

Summary. The paper proposes AiANet, an attention-infused autoencoder for massive MIMO CSI compression, combining a Hybrid Attention-Gated Fusion (HAGF) module, a Locally-Aware Self-Attention (LASA) module, and Gated Dense Connections (GDC). The authors evaluate AiANet on the COST2100 indoor/outdoor dataset at compression ratios 1/4 to 1/64 against CsiNet, CsiNetPlus, CRNet, DCRNet, and ACRNet, reporting NMSE and cosine similarity under intra-scenario, cross-scenario, and mixed-training protocols. The central claims are that AiANet outperforms existing autoencoder-based methods by up to 3.42 dB in NMSE and exhibits superior cross-scenario generalizability when trained on mixed indoor/outdoor data.

Significance. If the reported gains hold, AiANet would be a meaningful advance in learned CSI compression: it targets the practically important problem of cross-scenario generalization, uses standard public benchmarks, reports clean component-wise ablations (Section IV), and provides parameter/FLOP comparisons (Table VIII). The proposed modules are clearly described and the intra-scenario numbers are internally consistent with the stated 3.42 dB gain at eta=1/4 indoor. However, the central quantitative comparison is not yet established under equal feedback bit rates, because ACRNet is a binarized/quantized baseline while AiANet is evaluated with full-precision codewords, and because no statistical confidence is reported for any of the tables. The significance of the work is therefore conditional on resolving these issues.

major comments (3)

[Section V-B, Table V; Section II; Table I] The headline 3.42 dB improvement over ACRNet at eta=1/4 indoor is not an apples-to-apples comparison as presented. Table I and reference [13] describe ACRNet as a binarized network with quantization, whereas all experiments in Section V evaluate a noise-free, quantization-free codeword path; Section II itself states that the codeword is 'quantized, modulated, and transmitted' but 'often treated as noise-free in initial designs.' If the ACRNet entries are native binarized/quantized results and AiANet transmits full-precision codewords, the comparison does not hold bit rate or quantization level constant, and the reported gain could be substantially a bit-depth effect. The authors should either evaluate ACRNet under the same full-precision conditions, implement a common quantization scheme for both models, or clearly report that the comparison is against binarized ACRNet at a different effective feedback bit rate. This issue also affects the mixed-training comparison in Table VII.
[Tables V-VII] All performance tables report single-run NMSE values with no seed information, error bars, or confidence intervals. This is particularly problematic at the high compression ratios where the margins shrink to a few tenths of a dB: for example, in Table V at eta=1/64 the indoor gain over ACRNet is 0.91 dB and the outdoor gain is 0.21 dB, and in Table VII at eta=1/64 the mixed-training gains are 0.80 dB (indoor) and 0.36 dB (outdoor). Without multiple random seeds and a measure of variance, these small margins cannot be distinguished from training noise, so the claim of consistent superiority at high compression is not statistically supported. I recommend reporting mean and standard deviation over at least five independent runs, or at minimum providing a clear statement of the number of runs and seed dependence.
[Section V-B2, Table VI] The abstract's claim of 'superior cross-scenario generalizability' is not supported at the higher compression ratios. Table VI shows that at eta=1/64 the advantage over ACRNet is 0.07 dB for indoor-to-outdoor and 0.03 dB for outdoor-to-indoor (-11.56 vs -11.63 and -13.33 vs -13.36 dB, respectively), and even at eta=1/32 the gains are only 0.35 and 0.49 dB. Given the absence of error bars, these differences are within plausible run-to-run variation. The paper's own text acknowledges that the advantage 'narrows remarkably' with increasing compression, so the generalizability claim should be restricted to lower compression ratios or supported with statistical evidence across the full range.

minor comments (5)

[Section III-A, Eq. (7), Algorithm 2] The attention computation is written inconsistently: Eq. (7) uses Q^T K with a Reshape operation, while Algorithm 2 uses Q K^T, and the dimensions of the softmax argument are not made clear. Please unify the notation and specify the exact tensor shapes.
[Section III-A, Eq. (11), Fig. 2] The gating parameters are denoted gamma_ij in the Fig. 2 caption and W in Eq. (11), while Section III-A3 also introduces alpha via Eq. (12). The relationship between these symbols should be clarified to avoid confusion.
[Section III-A, Eq. (12)] The text states that the gate parameters are 'initialized to zero' and then says the initial gate values are approximately 0.5 after sigmoid; this is consistent, but the wording could be tightened to make clear that alpha is initialized to zero, not the gate value itself.
[Section II] There is a typo in the introduction: 'volumn' should read 'volume'.
[Fig. 10] The red arrows indicating visible differences may be difficult to discern in grayscale or small print; please use enhanced markers or provide a zoomed inset.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper's claims are empirical comparisons against external benchmarks on public data; the only self-reference is a non-load-bearing footnote to the authors' conference version.

full rationale

No circularity found. The paper's central claims are empirical comparisons against external benchmarks (CsiNet, CsiNetPlus, CRNet, DCRNet, ACRNet) on the public COST2100-derived dataset, using standard reconstruction metrics NMSE and cosine similarity (Eqs. (13) and (14)). These metrics are not defined in terms of AiANet's own encoder, decoder, or training procedure, and no reported number is forced by a fitted parameter renamed as a prediction. The architectural components (HAGF, LASA, GDC) are defined independently and ablated against explicit alternatives; the reported gains are measured on held-out test splits rather than on training data. The mixed-training generalization claim is evaluated on separate indoor and outdoor test sets after training on a balanced mixture, which is a standard held-out evaluation, and the separate-training cross-scenario table (Table VI) provides an additional transfer test that is not contaminated by the mixed-training labels. The only self-reference is the footnote citing the authors' WCNC conference version [1], which is not used to justify any architectural choice, numerical result, or theoretical uniqueness claim, so it is not load-bearing. Potential concerns such as quantization and bit-depth fairness relative to the binarized ACRNet baseline, and the absence of seeds or error bars, are benchmark-validity and reproducibility risks rather than circularity; they do not change the score.

Assumptions & free parameters 2 free parameters · 4 assumptions · 3 invented entities

The central claim is empirical and rests on trained network weights whose behavior is measured only on the standard COST2100 dataset with a fixed 32-antenna, noise-free-feedback setup. The three architectural modules are the paper's inventions and have no independent evidence beyond its own ablations. The training protocol itself contains an internal inconsistency in the reported weight decay, and no new physical entities are introduced.

free parameters (2)

Learnable network weights and gate scalars (alpha, beta1, beta2, omega, gamma, gating coefficients W) = Fitted on the COST2100 training split via AdamW
The reported NMSE numbers come entirely from training these parameters on the standard dataset; they are ordinary neural-network weights, not fitted physical constants, but the gains are inseparable from the training protocol.
Training hyperparameters (learning rate 1e-4, batch size 200, 1000 epochs, kernel size choices, attention reduction… = Chosen by the authors
Reported in Sections IV and V-A; the weight-decay setting is inconsistent across the two sections (1e-4 versus lambda=0), which is an unresolved detail of the protocol.

assumptions (4)

domain assumption The COST2100 channel model with default settings is a representative benchmark for indoor and outdoor massive MIMO channels.
Stated in Section III-C; the entire cross-scenario generalization claim is measured on these two simulated scenarios only, with no real-channel validation.
domain assumption Truncation to the first Nc=32 angular-delay rows preserves the essential CSI information.
Section II, adopted from CsiNet [11]; the discarded rows are assumed to be near zero based on maximum expected delay spread.
domain assumption The feedback codeword is transmitted without quantization or noise.
Section II treats the uplink as 'often treated as noise-free in initial designs'; none of the simulations include quantization or feedback errors, which limits the practical deployment claims.
domain assumption Minimizing MSE reconstruction error (Eq. 5) is the correct objective for downstream precoding performance.
The conclusion equates NMSE gains with higher data rates and spectral efficiency without an end-to-end precoding or rate evaluation.

invented entities (3)

HAGF (Hybrid Attention-Gated Fusion)
purpose: Adaptive channel-spatial feature fusion in encoder and decoder
Validated only by the paper's own ablation study (Table II); no external replication or independent benchmark exists for this module.
LASA (Locally-Aware Self-Attention)
purpose: Joint global and local spatial context modeling
Validated only by the paper's own ablation study (Table III); no independent evidence outside this work.
GDC (Gated Dense Connections)
purpose: Adaptive multi-scale feature propagation between encoder blocks
Validated only by the paper's own ablation study (Table IV); no independent evidence outside this work.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Attention-Infused Autoencoder for Massive MIMO CSI Compression." pith.science (2026). https://pith.science/paper/QTKB6RER

@misc{pith2026250412440,
  author       = {Pith},
  title        = {Pith review of: Attention-Infused Autoencoder for Massive MIMO CSI Compression},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QTKB6RER}},
  note         = {Machine review of arXiv:2504.12440}
}

read the original abstract

As the number of multiple-input multiple-output (MIMO) antennas increases drastically with the development towards 6G systems, channel state information (CSI) compression becomes crucial for mitigating feedback overhead. In recent years, learning models such as autoencoders (AE) have been studied for CSI compression, aiming to eliminate model assumptions and reduce compression loss. However, current learning methods are often designed and trained mainly for individual channel scenarios, with limited generalizability across different scenarios, of which the channel characteristics are prominently discrepant. Motivated by this, we propose a novel AE-based learning method named attention-infused autoencoder network (AiANet), which can parallelly and adaptively extract channel-wise and spatial features of CSI with an attention fusion mechanism. In addition, a locally-aware self-attention mechanism is developed to extract both global and local spatial patterns, to better capture the unique CSI features of different scenarios. Moreover, a mixed-training scheme is introduced to enable the proposed AiANet to gain generalizability across indoor and outdoor scenarios. Results show that when trained and tested in the same scenario, AiANet can substantially outperform the existing AE-based methods such as ACRNet, with an improvement of up to 3.42 dB in terms of normalized mean squared error (NMSE). With the mixed-training scheme, AiANet exhibits superior cross-scenario generalizability compared to the benchmark methods which are trained in one scenario and misused in another.

Figures

Figures reproduced from arXiv: 2504.12440 by the authors.

**Figure 2.** Architecture of the AiANet encoder. The encoder processes the input through four sequential [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** The structure diagram of the EncoderBlock with [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (6 more)

**Figure 4.** Figure 4: Architecture of the Hierarchical Attention Gating [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

**Figure 5.** Figure 5: Architecture of the Locally-Aware Self-Attention (LASA) module. It integrates features via three paths:(1) An upper [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

**Figure 6.** Figure 6: The structure diagram of the decoder in AiANet. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

**Figure 7.** Figure 7: Convergence comparison for attention module ablation [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]

**Figure 8.** Figure 8: Convergence behavior (NMSE in dB vs. training [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]

**Figure 10.** Figure 10: Visual comparison of CSI reconstruction under mixed [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 28 canonical work pages

[13]

Binarized aggregated network with quantization: Flexible deep learning deployment for CSI feedback in massive MIMO systems,

Z. Lu, X. Zhang, H. He, J. Wang, and J. Song, “Binarized aggregated network with quantization: Flexible deep learning deployment for CSI feedback in massive MIMO systems,” IEEE Trans. Wireless Commun. , vol. 21, no. 7, pp. 5514–5525, Jul. 2022

work page 2022
[1]

Attention-infused autoencoder for massive mimo csi compression,

K. Lou, X. Wu, and H. Ji, “Attention-infused autoencoder for massive mimo csi compression,” in Proc. IEEE Wireless Commun. Netw. Conf. (WCNC), Milan, Italy, Mar. 2025, to be published

work page 2025
[2]

Bj ¨ornson, J

E. Bj ¨ornson, J. Hoydis, and L. Sanguinetti, Massive MIMO Networks: Spectral, Energy, and Hardware Efficiency . Boston, USA: Now Publishers Inc., 2017

work page 2017
[3]

Massive mimo in the ul/dl of cellular networks: How many antennas do we need?

J. Hoydis, S. ten Brink, and M. Debbah, “Massive mimo in the ul/dl of cellular networks: How many antennas do we need?” IEEE J. Sel. Areas Commun., vol. 31, no. 2, pp. 160–171, Feb. 2013

work page 2013
[4]

Near-optimal signal recovery from random projections: Universal encoding strategies?

E. J. Candes and T. Tao, “Near-optimal signal recovery from random projections: Universal encoding strategies?” IEEE Trans. Inf. Theory , vol. 52, no. 12, pp. 5406–5425, Dec. 2006

work page 2006
[5]

Atomic decomposition by basis pursuit,

S. S. Chen, D. L. Donoho, and M. A. Saunders, “Atomic decomposition by basis pursuit,” SIAM Rev., vol. 43, no. 1, pp. 129–159, Dec. 2001

work page 2001
[6]

Orthogonal matching pursuit: recursive function approximation with applications to wavelet decomposition,

Y . Pati, R. Rezaiifar, and P. Krishnaprasad, “Orthogonal matching pursuit: recursive function approximation with applications to wavelet decomposition,” in Proc. 27th Asilomar Conf. Signals, Syst. and Com- put., Pacific Grove, CA, USA, Nov. 1993, pp. 40–44 vol.1

work page 1993
[7]

An iterative thresholding al- gorithm for linear inverse problems with a sparsity constraint,

I. Daubechies, M. Defrise, and C. De Mol, “An iterative thresholding al- gorithm for linear inverse problems with a sparsity constraint,” Commun. Pure Appl. Math. , vol. 57, no. 11, pp. 1413–1457, Nov. 2004

work page 2004

Show all 31 references

[8]

Compressed sensing,

D. L. Donoho, “Compressed sensing,” IEEE Trans. Inf. Theory, vol. 52, no. 4, pp. 1289–1306, Apr. 2006

2006
[9]

User’s Guide for TV AL3: TV Minimiza- tion by Augmented Lagrangian and Alternating Direction Algorithms,

C. Li, W. Yin, and Y . Zhang, “User’s Guide for TV AL3: TV Minimiza- tion by Augmented Lagrangian and Alternating Direction Algorithms,” Department of Computational and Applied Mathematics, Rice Univer- sity, Houston, TX, USA, Tech. Rep. TR09-02, 2009

2009
[10]

From denoising to compressed sensing,

C. A. Metzler, A. Maleki, and R. G. Baraniuk, “From denoising to compressed sensing,” IEEE Trans. Inf. Theory, vol. 62, no. 9, pp. 5117– 5144, Sept. 2016

2016
[11]

Deep learning for massive MIMO CSI feedback,

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

2018
[12]

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,” in Proc. IEEE Int. Conf. Commun. (ICC), Dublin, Ireland, Jun. 2020, pp. 1–6

2020
[14]

Dilated Convolution Based CSI Feedback Compression for Massive MIMO JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. XX, APRIL 2025 13 Systems,

S. Tang, J. Xia, L. Fan, X. Lei, W. Xu, and A. Nallanathan, “Dilated Convolution Based CSI Feedback Compression for Massive MIMO JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. XX, APRIL 2025 13 Systems,” IEEE Trans. Veh. Technol., vol. 71, no. 10, pp. 11 216–11 221, Oct. 2022

2025
[15]

Enhancing deep learning performance of massive MIMO CSI feedback,

S. Ji and M. Li, “Enhancing deep learning performance of massive MIMO CSI feedback,” in Proc. IEEE Int. Conf. Commun. (ICC) , Rome, Italy, May 2023, pp. 4949–4954

2023
[16]

Exploiting bi-directional channel reciprocity in deep learning for low rate massive mimo csi feedback,

Z. Liu, L. Zhang, and Z. Ding, “Exploiting bi-directional channel reciprocity in deep learning for low rate massive mimo csi feedback,” IEEE Wireless Commun. Lett. , vol. 8, no. 3, pp. 889–892, Jun. 2019

2019
[17]

Distributed deep con- volutional compression for massive mimo csi feedback,

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

2021
[18]

Fusion of self-attention mechanism for csi feedback in massive mimo systems,

T. Zhang, Y . Xu, and C. Zhou, “Fusion of self-attention mechanism for csi feedback in massive mimo systems,” ICT Express, vol. 11, no. 1, pp. 124–128, Feb. 2025

2025
[19]

Swin transformer- based csi feedback for massive mimo,

J. Cheng, W. Chen, J. Xu, Y . Guo, L. Li, and B. Ai, “Swin transformer- based csi feedback for massive mimo,”arXiv preprint arXiv:2401.06435, 2024

2024 arXiv
[20]

Channel estimation for massive mimo using gaussian-mixture bayesian learning,

C.-K. Wen, S. Jin, K.-K. Wong, J.-C. Chen, and P. Ting, “Channel estimation for massive mimo using gaussian-mixture bayesian learning,” IEEE Trans. Wireless Commun. , vol. 14, no. 3, pp. 1356–1368, Mar. 2015

2015
[21]

Learning In- ternal Representations by Error Propagation,

D. E. Rumelhart, G. E. Hinton, and R. J. Williams, “Learning In- ternal Representations by Error Propagation,” in Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Volume 1: Foundations, D. E. Rumelhart, J. L. McClelland, and P. R. Group, Eds. C...

1986
[22]

Cbam: Convolutional block attention module,

S. Woo, J. Park, J.-Y . Lee, and I. S. Kweon, “Cbam: Convolutional block attention module,” in Proc. Eur. Conf. Comput. Vis. (ECCV) , Munich, Germany, Sept. 2018, pp. 3–19

2018
[23]

Densenet for dense flow,

Y . Zhu and S. Newsam, “Densenet for dense flow,” in Proc. IEEE Int. Conf. Image Process. (ICIP) , Beijing, China, Sept. 2017, pp. 790–794

2017
[24]

The COST 2100 MIMO Channel Model,

L. Liu et al., “The COST 2100 MIMO Channel Model,” IEEE Wireless Commun., vol. 19, no. 6, pp. 92–99, Dec. 2012

2012
[25]

Squeeze-and-excitation networks,

J. Hu, L. Shen, and G. Sun, “Squeeze-and-excitation networks,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Salt Lake City, UT, USA, Jun. 2018, pp. 7132–7141

2018
[26]

Spa- tial Transformer Networks,

M. Jaderberg, K. Simonyan, A. Zisserman, and K. Kavukcuoglu, “Spa- tial Transformer Networks,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), Montreal, QC, Canada, Dec. 2015, pp. 2017–2025

2015
[27]

Understanding the Difficulty of Training Deep Feedforward Neural Networks,

X. Glorot and Y . Bengio, “Understanding the Difficulty of Training Deep Feedforward Neural Networks,” inProc. 13th Int. Conf. Artif. Intell. Stat. (AISTATS), ser. Proceedings of Machine Learning Research (PMLR), vol. 9, Chia Laguna Resort, Sardinia, Italy, May 2010, pp. 249–256

2010
[28]

Decoupled Weight Decay Regularization,

I. Loshchilov and F. Hutter, “Decoupled Weight Decay Regularization,” in Proc. Int. Conf. Learn. Represent. (ICLR) , New Orleans, LA, USA, May 2019

2019
[29]

SGDR: Stochastic Gradient Descent with Warm Restarts,

——, “SGDR: Stochastic Gradient Descent with Warm Restarts,” in Proc. Int. Conf. Learn. Represent. (ICLR) , Toulon, France, Apr. 2017

2017
[30]

Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification,

K. He, X. Zhang, S. Ren, and J. Sun, “Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification,” in Proc. IEEE Int. Conf. Comput. Vis. (ICCV) , Santiago, Chile, Dec. 2015, pp. 1026–1034

2015
[31]

Convolutional Neural Network- Based Multiple-Rate Compressive Sensing for Massive MIMO CSI Feedback: Design, Simulation, and Analysis,

J. Guo, C.-K. Wen, S. Jin, and G. Y . Li, “Convolutional Neural Network- Based Multiple-Rate Compressive Sensing for Massive MIMO CSI Feedback: Design, Simulation, and Analysis,” IEEE Trans. Wireless Commun., vol. 19, no. 4, pp. 2827–2840, Apr. 2020

2020

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [13]

Binarized aggregated network with quantization: Flexible deep learning deployment for CSI feedback in massive MIMO systems,

Z. Lu, X. Zhang, H. He, J. Wang, and J. Song, “Binarized aggregated network with quantization: Flexible deep learning deployment for CSI feedback in massive MIMO systems,” IEEE Trans. Wireless Commun. , vol. 21, no. 7, pp. 5514–5525, Jul. 2022

work page 2022

[2] [1]

Attention-infused autoencoder for massive mimo csi compression,

K. Lou, X. Wu, and H. Ji, “Attention-infused autoencoder for massive mimo csi compression,” in Proc. IEEE Wireless Commun. Netw. Conf. (WCNC), Milan, Italy, Mar. 2025, to be published

work page 2025

[3] [2]

Bj ¨ornson, J

E. Bj ¨ornson, J. Hoydis, and L. Sanguinetti, Massive MIMO Networks: Spectral, Energy, and Hardware Efficiency . Boston, USA: Now Publishers Inc., 2017

work page 2017

[4] [3]

Massive mimo in the ul/dl of cellular networks: How many antennas do we need?

J. Hoydis, S. ten Brink, and M. Debbah, “Massive mimo in the ul/dl of cellular networks: How many antennas do we need?” IEEE J. Sel. Areas Commun., vol. 31, no. 2, pp. 160–171, Feb. 2013

work page 2013

[5] [4]

Near-optimal signal recovery from random projections: Universal encoding strategies?

E. J. Candes and T. Tao, “Near-optimal signal recovery from random projections: Universal encoding strategies?” IEEE Trans. Inf. Theory , vol. 52, no. 12, pp. 5406–5425, Dec. 2006

work page 2006

[6] [5]

Atomic decomposition by basis pursuit,

S. S. Chen, D. L. Donoho, and M. A. Saunders, “Atomic decomposition by basis pursuit,” SIAM Rev., vol. 43, no. 1, pp. 129–159, Dec. 2001

work page 2001

[7] [6]

Orthogonal matching pursuit: recursive function approximation with applications to wavelet decomposition,

Y . Pati, R. Rezaiifar, and P. Krishnaprasad, “Orthogonal matching pursuit: recursive function approximation with applications to wavelet decomposition,” in Proc. 27th Asilomar Conf. Signals, Syst. and Com- put., Pacific Grove, CA, USA, Nov. 1993, pp. 40–44 vol.1

work page 1993

[8] [7]

An iterative thresholding al- gorithm for linear inverse problems with a sparsity constraint,

I. Daubechies, M. Defrise, and C. De Mol, “An iterative thresholding al- gorithm for linear inverse problems with a sparsity constraint,” Commun. Pure Appl. Math. , vol. 57, no. 11, pp. 1413–1457, Nov. 2004

work page 2004