pith. sign in

arxiv: 2606.22038 · v1 · pith:2UKQXIBPnew · submitted 2026-06-20 · 📡 eess.SP

Full-Domain Coupler: A Wireless Native Neural Backbone for Channel Representation and Deduction

Zirui Chen , Ziqing Xing , Zhaoyang Zhang , Hongning Ruan , Yuzhi Yang , Zhaohui Yang , Chongwen Huang , Merouane Debbah This is my paper

Pith reviewed 2026-06-26 11:22 UTC · model grok-4.3

classification 📡 eess.SP
keywords wireless AIchannel state informationrepresentation learningneural backbonemulti-domain fusionchannel deductionCSI tensor
0
0 comments X

The pith

Coupler decomposes CSI representation into domain-specific layers then couples them via dimension-staggered cascades.

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

The paper proposes Coupler as a neural backbone built specifically for wireless channel state information tensors rather than adapting general networks. It breaks representation learning into separate time, space, and frequency domains on a layer-by-layer basis and then recombines the domain features with a staggered cascade structure. This design targets the redundancy that arises when wireless AI simply stitches existing architectures together. A reader would care because the result is claimed to be both more parameter-efficient and better at fusing information across physical domains for tasks such as channel estimation and prediction.

Core claim

Coupler leverages the physical insights of channel tensors to decompose representation learning into individual domains on a layer-by-layer basis, and then couples the learned domain-specific features through a dimension-staggered cascade. This full-domain interleaved learning architecture enables superior parameter efficiency and fine-grained multi-domain feature fusion.

What carries the argument

The dimension-staggered cascade that interleaves features from domain-specific learners (CMLPs for space and frequency, plus convolution/attention/gating for time).

If this is right

  • Produces multiple lightweight schemes by pairing CMLPs with optional temporal mechanisms for diverse channel tasks.
  • Delivers measurable gains on channel estimation, interpolation, prediction, and feedback.
  • Retains performance when tested on real-world measured CSI data.
  • Offers a candidate basic architecture for larger wireless foundation models.

Where Pith is reading between the lines

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

  • The same layer-wise domain split might apply to other tensor-structured signals that have separable physical dimensions.
  • If the cascade coupling proves stable, it could reduce the need for heavy attention mechanisms in wireless models.

Load-bearing premise

The physical insights of channel tensors permit effective decomposition of representation learning into individual domains on a layer-by-layer basis.

What would settle it

An experiment in which a conventional network without per-domain layer decomposition matches or exceeds Coupler on both accuracy and parameter count for the same channel deduction tasks would falsify the claimed advantage.

Figures

Figures reproduced from arXiv: 2606.22038 by Chongwen Huang, Hongning Ruan, Merouane Debbah, Yuzhi Yang, Zhaohui Yang, Zhaoyang Zhang, Ziqing Xing, Zirui Chen.

Figure 1
Figure 1. Figure 1: Illustration of channel deduction for a mobile user. The current channel is deduced from past channels and the present [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The overall structure of vanilla CDNet [42], serving as a representative multi-stage decoupled architecture. operates on the specific physical domain at each stage. Consequently, it cannot effectively interweave the dynamic temporal evolution with the spatial-frequency characteristics at a deep feature level, failing to fully exploit the intrinsic multi-domain coupling effect dictated by physical propagati… view at source ↗
Figure 3
Figure 3. Figure 3: Overall architecture of the Coupler backbone, consisting of cascaded temporal, spatial, and frequency feature mixing modules. The spatial and frequency mixing modules are implemented using CMLPs, while the temporal mixing module admits three different realizations, corresponding to ConvCoupler, ACoupler, and AFCoupler, respectively. With IC-2D layers P 2D pre and P 2D post, Coupler can be readily applied t… view at source ↗
Figure 4
Figure 4. Figure 4: Using ‘O1’ scenario in DeepMIMO dataset [49] as experimental scenario, and collecting training and testing datasets from it. A. Experiment Settings 1) Communication Scenario: In this work, we first conduct simulation experiments using the raytracing￾based DeepMIMO dataset [49]. Specifically, we generate wireless channel data in a typical outdoor com￾munication scenario ‘O1’, as shown in [PITH_FULL_IMAGE:f… view at source ↗
Figure 5
Figure 5. Figure 5: NMSE of proposed Couplers and benchmarks under various estimated present channel sizes. The sizes of the estimated present partial channels are shown in the legend, and the size of the full channel is 32 antennas × 32 subcarriers. In each box, the central line indicates the median, the box edges represent the 25th and 75th percentiles, and the whiskers extend to the furthest data points within 1.5 times th… view at source ↗
Figure 6
Figure 6. Figure 6: Cumulative probability distribution of errors between the acquired channel and the true channel (NMSE). alone cannot sufficiently exploit the 2D spatial-frequency channel structure after flattening. This is because ACDNet sandwiches the temporal Transformer module between two spatial-frequency feature mixing modules, allowing the spatial-frequency structure to be preliminarily mapped before temporal intera… view at source ↗
Figure 7
Figure 7. Figure 7: NMSE convergence curves on the test set during training. a significantly shorter training time to attain satisfactory performance, which offers advantages for time￾sensitive applications, such as online learning and fine-tuning. 3) Acquired CSI Accuracy versus Number of Past Channels: The advantage of channel deduction over channel estimation primarily lies in extracting additional information from past ch… view at source ↗
Figure 8
Figure 8. Figure 8: Impact of the number of available past channels on channel deduction accuracy. The size of estimated partial channel through pilots is 3 antennas × 3 subcarriers. mechanisms, respectively. These mechanisms enable them to learn global channel structure from historical data, thereby making them robust to mobility pattern variations. 4) Robustness against Channel Disturbances: In practical scenarios, influenc… view at source ↗
Figure 9
Figure 9. Figure 9: Robustness evaluation of different models. The size of estimated partial channel through pilots is 3 antennas × 3 subcarriers [PITH_FULL_IMAGE:figures/full_fig_p027_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: NMSE of proposed Couplers and benchmarks under various numbers of training sub-areas. partial channel for channel acquisition, as it provides critical small-scale fast-fading information. Among the proposed models, AFCoupler and ConvCoupler demonstrate superior robustness, while ACoupler exhibits relatively higher sensitivity. This is attributed to the ACoupler’s reliance on fine-grained feature correlati… view at source ↗
Figure 11
Figure 11. Figure 11: Continuous channel acquisition for a mobile user using Couplers in the autoregressive scenario and the ideal case without error propagation. 6) Continuous Channel Acquisition with Couplers: Following the autoregressive deployment intro￾duced in Section III-C, we utilize the trained Coupler models to provide continuous channel acquisition services for a mobile user, aiming to evaluate the models’ effective… view at source ↗
Figure 12
Figure 12. Figure 12: Illustration of the DICHASUS-015x dataset [52]. case without error propagation, ACoupler exhibits substantial performance degradation in autoregression. Even in the ideal case, the NMSE of the channels obtained by ACoupler shows noticeable fluctuations. As discussed in Section III-B3, ACoupler relies on fine-grained temporal similarities to select highly relevant historical channel features. While this me… view at source ↗
Figure 13
Figure 13. Figure 13: Channel acquisition performance of different models under varying sizes of the estimated present channel. The models are tested using real-world channel measurements from trajectories DICHASUS-0152 to 0157 [52]. indoor mobile communication scenario featuring scatterers such as metal plates, whiteboards, and walls, as shown in [PITH_FULL_IMAGE:figures/full_fig_p030_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Autoregressive channel acquisition performance of AFCoupler on a new trajectory DICHASUS-0158 [52]. exploit the nonlinear feature correlations of channels across spatial, temporal, and frequency domains, thereby achieving significant performance improvements. Among the channel deduction models, ConvLSTM exhibits poor performance due to a mismatch be￾tween its network architecture and the characteristics o… view at source ↗
read the original abstract

Data representation is a fundamental issue in deep learning. However, as wireless data scales and deeply couples across many physical domains such as time, space, and frequency, existing wireless artificial intelligence (AI) technologies lack dedicated representation solutions. Instead, they mainly rely on stitching general-purpose networks, a tool-driven paradigm that inevitably results in structural redundancy and bottlenecks in information flow. To fill this gap, this paper proposes Coupler, a wireless native-AI neural backbone designed for representation learning of channel state information (CSI)--the pivotal data in wireless systems. Leveraging the revealed physical insights of channel tensors, Coupler decomposes representation learning into individual domains on a layer-by-layer basis, and then couples the learned domain-specific features through a dimension-staggered cascade. This full-domain interleaved learning architecture enables superior parameter efficiency and fine-grained multi-domain feature fusion. Based on this backbone, we use the complex-domain multilayer perceptrons (CMLPs) as spatial and frequency domain learners, while employing three optional mechanisms--convolution, attention, or gating--to capture temporal dependencies. This results in a series of efficient channel learning schemes with diverse functionalities and extreme lightweights, showcasing the compactness, versatility and flexibility of Coupler. We evaluate these schemes on channel deduction, a general representation task encompassing channel estimation, interpolation, prediction, and feedback. Extensive experimental evaluations validate their significant performance gains and robust applicability even for real-world measured data, demonstrating the potential of Coupler as a promising basic architecture in the design of wireless foundation models.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The paper proposes Coupler, a wireless-native neural backbone for CSI representation learning. It decomposes representation learning into individual physical domains (space, frequency, time) on a layer-by-layer basis based on channel tensor insights, then couples the domain-specific features via a dimension-staggered cascade. The architecture employs complex-domain MLPs for spatial/frequency domains and optional mechanisms (convolution, attention, or gating) for temporal dependencies, yielding lightweight schemes evaluated on the general channel deduction task (estimation, interpolation, prediction, feedback) with reported gains in parameter efficiency, multi-domain fusion, and performance on both simulated and real measured data.

Significance. If the central claims hold, the work supplies a dedicated architecture that directly addresses structural redundancy in general-purpose networks for wireless data, offering a potential foundational component for wireless foundation models. Strengths include the explicit grounding in channel tensor structure, the resulting parameter efficiency, the flexibility across temporal mechanisms, and the inclusion of real-world measured data validation.

minor comments (2)
  1. [Abstract] The abstract states that the architecture 'enables superior parameter efficiency' but does not quantify the parameter counts or FLOPs relative to the stitched general-purpose baselines; adding these numbers in §4 or Table 1 would strengthen the efficiency claim.
  2. Notation for the dimension-staggered cascade and the exact layer-by-layer decomposition is introduced in the abstract without a forward reference to the corresponding diagram or equations; a brief pointer would improve readability.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive and accurate summary of our work, the recognition of its significance as a potential foundational architecture, and the recommendation for minor revision. No specific major comments or requested changes were listed in the report.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper proposes Coupler as a new neural backbone architecture explicitly motivated by physical properties of channel tensors, with domain decomposition and staggered coupling presented as design choices rather than derived predictions. No equations or claims in the abstract reduce a result to a fitted parameter or self-citation by construction; the architecture is described as enabling efficiency and fusion, with evaluation on channel deduction tasks treated as external validation. The derivation chain remains self-contained against the stated physical insights without load-bearing self-referential steps.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim rests on the assumption about channel tensor properties and introduces the Coupler as a new entity without additional free parameters specified in the abstract.

axioms (1)
  • domain assumption Channel tensors possess physical insights that permit layer-by-layer domain decomposition
    This is the basis for the Coupler design as stated in the abstract.
invented entities (1)
  • Coupler neural backbone no independent evidence
    purpose: To provide a wireless-native representation learning architecture for CSI
    Newly proposed architecture without external validation mentioned.

pith-pipeline@v0.9.1-grok · 5836 in / 1284 out tokens · 35060 ms · 2026-06-26T11:22:08.679332+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

53 extracted references · 2 linked inside Pith

  1. [1]

    Parameter-efficient channel deduction based on full-domain interleaved learning,

    Z. Xing, Z. Chen, Z. Zhanget al., “Parameter-efficient channel deduction based on full-domain interleaved learning,” in IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), 2026 (accepted)

    2026

  2. [2]

    Framework and overall objectives of the future development of IMT for 2030 and beyond,

    International Telecommunication Union, “Framework and overall objectives of the future development of IMT for 2030 and beyond,” ITU-R, Tech. Rep. Recommendation M.2160-0, Nov 2023

    2030

  3. [3]

    ComAI: The convergence of communication and artificial intelligence,

    P. Zhang, K. Niu, X. Wanget al., “ComAI: The convergence of communication and artificial intelligence,”IEEE Commun. Surveys Tuts., vol. 28, pp. 2163–2197, 2026

    2026

  4. [4]

    Toward wisdom-evolutionary and primitive-concise 6G: A new paradigm of semantic communication networks,

    P. Zhang, W. Xu, H. Gaoet al., “Toward wisdom-evolutionary and primitive-concise 6G: A new paradigm of semantic communication networks,”Engineering, vol. 8, pp. 60–73, 2022

    2022

  5. [5]

    Big AI models for 6G wireless networks: Opportunities, challenges, and research directions,

    Z. Chen, Z. Zhang, and Z. Yang, “Big AI models for 6G wireless networks: Opportunities, challenges, and research directions,”IEEE Wireless Commun., vol. 31, no. 5, pp. 164–172, 2024

    2024

  6. [6]

    Toward a 6G AI-native air interface,

    J. Hoydis, F. A. Aoudia, A. Valcarceet al., “Toward a 6G AI-native air interface,”IEEE Commun. Mag., vol. 59, no. 5, pp. 76–81, 2021

    2021

  7. [7]

    Way to build native AI-driven 6G air interface: Principles, roadmap, and outlook,

    P. Zhang, K. Niu, Y . Liuet al., “Way to build native AI-driven 6G air interface: Principles, roadmap, and outlook,”IEEE Trans. Netw. Sci. Eng., vol. 13, pp. 3551–3565, 2026

    2026

  8. [8]

    A task-driven design approach for 6G AI-native architecture,

    X. Wang, L. Lu, Q. Liet al., “A task-driven design approach for 6G AI-native architecture,”Engineering, vol. 56, pp. 87–103, 2026

    2026

  9. [9]

    Situation-aware resource allocation for multi-dimensional intelligent multiple access: A proactive deep learning framework,

    Y . Liu, X. Wang, J. Meiet al., “Situation-aware resource allocation for multi-dimensional intelligent multiple access: A proactive deep learning framework,”IEEE J. Sel. Areas Commun., vol. 39, no. 1, pp. 116–130, 2020

    2020

  10. [10]

    Intelligent beam management based on deep reinforcement learning in high-speed railway scenarios,

    Y . Qiao, Y . Niu, X. Zhanget al., “Intelligent beam management based on deep reinforcement learning in high-speed railway scenarios,”IEEE Trans. V eh. Technol., vol. 73, no. 3, pp. 3917–3931, 2023

    2023

  11. [11]

    The Agentic-AI core: An AI-empowered, mission-oriented core network for next-generation mobile telecommunications,

    X. Li, W. Shi, H. Zhanget al., “The Agentic-AI core: An AI-empowered, mission-oriented core network for next-generation mobile telecommunications,”Engineering, vol. 56, pp. 104–119, 2026

    2026

  12. [12]

    Towards wireless native big AI model: the mission and approach differ from large language model,

    Z. Chen, Z. Zhang, C. Liuet al., “Towards wireless native big AI model: the mission and approach differ from large language model,”Sci. China Inf. Sci., vol. 68, no. 7, p. 170303, 2025

    2025

  13. [13]

    Wireless environmental information theory: A new paradigm toward 6G online and proactive environment intelligence communication,

    J. Zhang, L. Yu, S. Liuet al., “Wireless environmental information theory: A new paradigm toward 6G online and proactive environment intelligence communication,”Engineering, vol. 56, pp. 186–200, 2026. June 23, 2026 DRAFT 34

    2026

  14. [14]

    Acquisition of channel state information for mmwave massive MIMO: Traditional and machine learning-based approaches,

    C. Qi, P. Dong, W. Maet al., “Acquisition of channel state information for mmwave massive MIMO: Traditional and machine learning-based approaches,”Sci. China Inf. Sci., vol. 64, no. 8, p. 181301, 2021

    2021

  15. [15]

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

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

    2022

  16. [16]

    From data-driven learning to physics-inspired inferring: A novel mobile MIMO channel prediction scheme based on neural ODE,

    Z. Xiao, Z. Zhang, Z. Chenet al., “From data-driven learning to physics-inspired inferring: A novel mobile MIMO channel prediction scheme based on neural ODE,”IEEE Trans. Wireless Commun., vol. 23, no. 7, pp. 7186–7199, 2023

    2023

  17. [17]

    A hybrid inference architecture incorporating neural network with belief propagation for AI receivers,

    Y . Yang, Z. Zhang, Z. Chenet al., “A hybrid inference architecture incorporating neural network with belief propagation for AI receivers,”IEEE Trans. Wireless Commun., vol. 24, no. 8, pp. 6365–6381, 2025

    2025

  18. [18]

    Multi-sources fusion learning for multi-points NLOS localization in OFDM system,

    B. Wang, Z. Shuai, C. Huanget al., “Multi-sources fusion learning for multi-points NLOS localization in OFDM system,” IEEE J. Sel. Topics Signal Process., vol. 18, no. 7, pp. 1339–1350, 2024

    2024

  19. [19]

    Analogical learning for cross-scenario generalization: Framework and application to intelligent localization,

    Z. Chen, Z. Zhang, Z. Xinget al., “Analogical learning for cross-scenario generalization: Framework and application to intelligent localization,”arXiv preprint arXiv:2504.08811, 2025

    arXiv 2025

  20. [20]

    Multi-view wireless sensing via conditional generative learning: Framework and model design,

    Z. Xing, Z. Zhang, Z. Chenet al., “Multi-view wireless sensing via conditional generative learning: Framework and model design,”IEEE Trans. Wireless Commun., vol. 25, pp. 12 691–12 708, 2026

    2026

  21. [21]

    Deep learning-based channel estimation,

    M. Soltani, V . Pourahmadi, A. Mirzaeiet al., “Deep learning-based channel estimation,”IEEE Commun. Lett., vol. 23, no. 4, pp. 652–655, 2019

    2019

  22. [22]

    Deep residual learning meets OFDM channel estimation,

    L. Li, H. Chen, H.-H. Changet al., “Deep residual learning meets OFDM channel estimation,”IEEE Wireless Commun. Lett., vol. 9, no. 5, pp. 615–618, May 2020

    2020

  23. [23]

    Deep learning based channel estimation algorithm for fast time-varying MIMO-OFDM systems,

    Y . Liao, Y . Hua, and Y . Cai, “Deep learning based channel estimation algorithm for fast time-varying MIMO-OFDM systems,”IEEE Commun. Lett., vol. 24, no. 3, pp. 572–576, Mar. 2020

    2020

  24. [24]

    Deep learning for TDD and FDD massive MIMO: Mapping channels in space and frequency,

    M. Alrabeiah and A. Alkhateeb, “Deep learning for TDD and FDD massive MIMO: Mapping channels in space and frequency,” inAsilomar Conference on Signals, Systems, and Computers (Asilomar), 2019, pp. 1465–1470

    2019

  25. [25]

    Channel mapping based on interleaved learning with complex-domain MLP-mixer,

    Z. Chen, Z. Zhang, Z. Yanget al., “Channel mapping based on interleaved learning with complex-domain MLP-mixer,” IEEE Wireless Commun. Lett., vol. 13, no. 5, pp. 1369–1373, 2024

    2024

  26. [26]

    Deep learning based channel extrapolation for large-scale antenna systems: Opportunities, challenges and solutions,

    S. Zhang, Y . Liu, F. Gaoet al., “Deep learning based channel extrapolation for large-scale antenna systems: Opportunities, challenges and solutions,”IEEE Wireless Commun., vol. 28, no. 6, pp. 160–167, 2021

    2021

  27. [27]

    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, 2018

    2018

  28. [28]

    Convolutional neural network-based multiple-rate compressive sensing for massive MIMO CSI feedback: Design, simulation, and analysis,

    J. Guo, C.-K. Wen, S. Jinet al., “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

  29. [29]

    SALDR: Joint self-attention learning and dense refine for massive MIMO CSI feedback with multiple compression ratio,

    X. Song, J. Wang, J. Wanget al., “SALDR: Joint self-attention learning and dense refine for massive MIMO CSI feedback with multiple compression ratio,”IEEE Wireless Commun. Lett., vol. 10, no. 9, pp. 1899–1903, Sep. 2021

    1903

  30. [30]

    A lightweight deep network for efficient CSI feedback in massive MIMO systems,

    Y . Sun, W. Xu, L. Lianget al., “A lightweight deep network for efficient CSI feedback in massive MIMO systems,”IEEE Wireless Commun. Lett., vol. 10, no. 8, pp. 1840–1844, Jun. 2021

    2021

  31. [31]

    Viewing channel as sequence rather than image: A 2-D Seq2Seq approach for efficient MIMO-OFDM CSI feedback,

    Z. Chen, Z. Zhang, Z. Xiaoet al., “Viewing channel as sequence rather than image: A 2-D Seq2Seq approach for efficient MIMO-OFDM CSI feedback,”IEEE Trans. Wireless Commun., vol. 22, no. 11, pp. 7393–7407, Nov. 2023

    2023

  32. [32]

    Deep learning for fading channel prediction,

    W. Jiang and H. D. Schotten, “Deep learning for fading channel prediction,”IEEE Open J. Commun. Soc., vol. 1, pp. 320–332, Mar. 2020

    2020

  33. [33]

    Channel state information prediction for 5G wireless communications: A deep learning approach,

    C. Luo, J. Ji, Q. Wanget al., “Channel state information prediction for 5G wireless communications: A deep learning approach,”IEEE Trans. Netw. Sci. Eng., vol. 7, no. 1, pp. 227–236, Jan.-Mar. 2020. June 23, 2026 DRAFT 35

    2020

  34. [34]

    Accurate channel prediction based on Transformer: Making mobility negligible,

    H. Jiang, M. Cui, D. W. K. Nget al., “Accurate channel prediction based on Transformer: Making mobility negligible,” IEEE J. Sel. Areas Commun., vol. 40, no. 9, pp. 2717–2732, 2022

    2022

  35. [35]

    Mobile MIMO channel prediction with ODE-RNN: A physics-inspired adaptive approach,

    Z. Xiao, Z. Zhang, Z. Chenet al., “Mobile MIMO channel prediction with ODE-RNN: A physics-inspired adaptive approach,” inIEEE Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), 2022, pp. 1301–1307

    2022

  36. [36]

    C-GRBFnet: A physics-inspired generative deep neural network for channel representation and prediction,

    Z. Xiao, Z. Zhang, C. Huanget al., “C-GRBFnet: A physics-inspired generative deep neural network for channel representation and prediction,”IEEE J. Sel. Areas Commun., vol. 40, no. 8, pp. 2282–2299, Aug. 2022

    2022

  37. [37]

    CSI-based positioning in massive MIMO systems using convolutional neural networks,

    S. Bast, A. Guevara, and S. Pollin, “CSI-based positioning in massive MIMO systems using convolutional neural networks,” inIEEE V ehicular Technology Conference (VTC-Spring), 2020, pp. 1–5

    2020

  38. [38]

    Machine-learning assisted outdoor localization via sector-based fog massive MIMO,

    H. Pirzadeh, C. Wang, and H. Papadopoulos, “Machine-learning assisted outdoor localization via sector-based fog massive MIMO,” inIEEE International Conference on Communications (ICC), 2019, pp. 1–6

    2019

  39. [39]

    Deep learning-based multi-user positioning in wireless FDMA cellular networks,

    Z. Chen, Z. Zhang, Z. Xiaoet al., “Deep learning-based multi-user positioning in wireless FDMA cellular networks,” IEEE J. Sel. Areas Commun., vol. 41, no. 12, pp. 3848–3862, 2023

    2023

  40. [40]

    Attention-aided outdoor localization in commercial 5G NR systems,

    G. Tian, D. Pjani ´c, X. Caiet al., “Attention-aided outdoor localization in commercial 5G NR systems,”IEEE Trans. Mach. Learn. Commun. Netw., vol. 2, pp. 1678–1692, 2024

    2024

  41. [41]

    Synergizing global pattern learning and time order characterization in mobile channel prediction: An RWKV-based approach,

    Z. Wang, Z. Chen, R. Liet al., “Synergizing global pattern learning and time order characterization in mobile channel prediction: An RWKV-based approach,” 2026. [Online]. Available: https://arxiv.org/abs/2606.16170

    arXiv 2026

  42. [42]

    Channel deduction: A new learning framework to acquire channel from outdated samples and coarse estimate,

    Z. Chen, Z. Zhang, Z. Yanget al., “Channel deduction: A new learning framework to acquire channel from outdated samples and coarse estimate,”IEEE J. Sel. Areas Commun., vol. 43, no. 3, pp. 944–958, 2025

    2025

  43. [43]

    Spatial channel deduction: Acquiring channel from approximate position and coarse estimate,

    Z. Chen, Z. Zhang, Z. Xinget al., “Spatial channel deduction: Acquiring channel from approximate position and coarse estimate,” in2025 IEEE 36th International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC). Istanbul, Turkiye: IEEE, 2025, pp. 1–6

    2025

  44. [44]

    Multi-input multi-output fading channel tracking and equalization using Kalman estimation,

    C. Komninakis, C. Fragouli, A. H. Sayedet al., “Multi-input multi-output fading channel tracking and equalization using Kalman estimation,”IEEE Trans. Signal Process., vol. 50, no. 5, pp. 1065–1076, 2002

    2002

  45. [45]

    Deep learning-based CSI feedback approach for time-varying massive MIMO channels,

    T. Wang, C.-K. Wen, S. Jinet al., “Deep learning-based CSI feedback approach for time-varying massive MIMO channels,” IEEE Wireless Commun. Lett., vol. 8, no. 2, pp. 416–419, 2018

    2018

  46. [46]

    Geometry-aided channel deduction: A robust channel acquisition framework utilizing coarse scenario prompt,

    H. Ruan, Z. Zhang, Z. Chenet al., “Geometry-aided channel deduction: A robust channel acquisition framework utilizing coarse scenario prompt,”arXiv preprint arXiv:2605.07425, 2026

    Pith/arXiv arXiv 2026

  47. [47]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmaret al., “Attention is all you need,”Advances in neural information processing systems, vol. 30, 2017

    2017

  48. [48]

    An Attention free Transformer,

    S. Zhai, W. Talbott, N. Srivastavaet al., “An Attention free Transformer,”arXiv preprint arXiv:2105.14103, 2021

    arXiv 2021

  49. [49]

    DeepMIMO: A generic deep learning dataset for millimeter wave and massive MIMO applications,

    A. Alkhateeb, “DeepMIMO: A generic deep learning dataset for millimeter wave and massive MIMO applications,” in Proceedings of the Information Theory and Applications Workshop (ITA), San Diego, CA, USA, 2019, pp. 1–8

    2019

  50. [50]

    Convolutional LSTM network: A machine learning approach for precipitation nowcasting,

    X. Shi, Z. Chen, H. Wanget al., “Convolutional LSTM network: A machine learning approach for precipitation nowcasting,” Advances in neural information processing systems, vol. 28, 2015

    2015

  51. [51]

    Adam: A method for stochastic optimization,

    D. Kingma and J. Ba, “Adam: A method for stochastic optimization,”arXiv preprint arXiv:1412.6980, 2014

    Pith/arXiv arXiv 2014

  52. [52]

    A distributed massive MIMO channel sounder for “big CSI data

    F. Euchner, M. Gauger, S. Dörneret al., “A distributed massive MIMO channel sounder for “big CSI data”-driven machine learning,” in25th International ITG Workshop on Smart Antennas (WSA), French Riviera, France, 2021, pp. 1–6

    2021

  53. [53]

    Channel estimation for OFDM systems based on comb-type pilot arrangement in frequency selective fading channels,

    M. H. Hsieh and C. Wei, “Channel estimation for OFDM systems based on comb-type pilot arrangement in frequency selective fading channels,”IEEE Trans. Consum. Electron., vol. 44, no. 1, pp. 217–225, 1998. June 23, 2026 DRAFT

    1998