pith. sign in

arxiv: 2502.03117 · v1 · submitted 2025-02-05 · 💻 cs.IT · eess.SP· math.IT

Meta-Learning-Based People Counting and Localization Models Employing CSI from Commodity WiFi NICs

Jihoon Cha , Hwanjin Kim , Junil Choi This is my paper

Pith reviewed 2026-05-23 04:08 UTC · model grok-4.3

classification 💻 cs.IT eess.SPmath.IT
keywords CSIWiFi sensingpeople countinglocalizationmeta-learningchannel state informationindoor sensingpreprocessing
0
0 comments X

The pith

Meta-learning models using preprocessed WiFi CSI count and locate people more accurately across environments than standard training methods.

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

The paper shows how channel state information from ordinary WiFi cards can support indoor people counting and localization despite signal offsets and outside interference. A preprocessing step removes offsets to keep the system fast and filter-free. Pre-trained models are then extended with meta-learning so they adjust to new rooms or setups using limited additional data. Numerical comparisons indicate the meta-learning versions reach higher accuracy than models trained and tested in the usual fixed way. This approach matters if it lets existing WiFi networks perform reliable sensing without new hardware or lengthy retraining for each location.

Core claim

The paper claims that meta-learning-based people counting and localization models employing CSI from commodity WiFi NICs achieve high sensing accuracy compared to other learning schemes that follow simple training and test procedures, after an initial preprocessing step that removes offsets and supports low-latency operation without filtering.

What carries the argument

Meta-learning applied to pre-trained models after CSI offset-removal preprocessing to enable adaptation across measurement environments.

If this is right

  • The preprocessing enables low-latency operation by avoiding any filtering process.
  • The models handle erroneous CSI from interfering signals through the offset removal step.
  • Meta-learning versions adapt to varying environments while standard training and test procedures do not.
  • High sensing accuracy holds for both counting and localization tasks using commodity hardware.

Where Pith is reading between the lines

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

  • The same preprocessing and meta-learning pipeline could apply to other WiFi-based sensing tasks such as activity detection.
  • Fewer labeled samples from new locations may suffice for deployment once meta-training is complete.
  • Existing home or office WiFi networks might support continuous sensing without dedicated sensors if the adaptation works across typical interference levels.

Load-bearing premise

The assumption that meta-learning will reliably adapt the models to different measurement environments without detailed specification of the meta-training procedure, environment diversity, or how interference is handled beyond the initial preprocessing.

What would settle it

A direct comparison test in a new environment with fresh interference patterns where the meta-learning model shows no accuracy gain over a standard pre-trained model using the same preprocessed CSI.

Figures

Figures reproduced from arXiv: 2502.03117 by Hwanjin Kim, Jihoon Cha, Junil Choi.

Figure 1
Figure 1. Figure 1: Structure of a CSI measurement system using com [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: A remote server conducts preprocessing of CSI and [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: A basic CNN module for the proposed people counting an [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: A pre-training-based model using preprocessed CSI a [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: The proposed meta-learning-based people counting [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Experimental environments for CSI measurement. [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: People counting accuracy vs. number of adaptation [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Confusion matrix of Meta-CSI for people counting [PITH_FULL_IMAGE:figures/full_fig_p010_9.png] view at source ↗
Figure 12
Figure 12. Figure 12: RMSE in large room vs. number of adaptation [PITH_FULL_IMAGE:figures/full_fig_p011_12.png] view at source ↗
Figure 11
Figure 11. Figure 11: Confusion matrix of Meta-CSI for localization when [PITH_FULL_IMAGE:figures/full_fig_p011_11.png] view at source ↗
Figure 13
Figure 13. Figure 13: RMSE in open space vs. number of adaptation [PITH_FULL_IMAGE:figures/full_fig_p011_13.png] view at source ↗
Figure 15
Figure 15. Figure 15: CDF of localization error in open space with [PITH_FULL_IMAGE:figures/full_fig_p012_15.png] view at source ↗
read the original abstract

In this paper, we consider people counting and localization systems exploiting channel state information (CSI) measured from commodity WiFi network interface cards (NICs). While CSI has useful information of amplitude and phase to describe signal propagation in a measurement environment of interest, CSI measurement suffers from offsets due to various uncertainties. Moreover, an uncontrollable external environment where other WiFi devices communicate each other induces interfering signals, resulting in erroneous CSI captured at a receiver. In this paper, preprocessing of CSI is first proposed for offset removal, and it guarantees low-latency operation without any filtering process. Afterwards, we design people counting and localization models based on pre-training. To be adaptive to different measurement environments, meta-learning-based people counting and localization models are also proposed. Numerical results show that the proposed meta-learning-based people counting and localization models can achieve high sensing accuracy, compared to other learning schemes that follow simple training and test procedures.

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

1 major / 0 minor

Summary. The manuscript proposes preprocessing of CSI from commodity WiFi NICs to remove offsets for low-latency operation without filtering, followed by pre-training-based models for people counting and localization. It further introduces meta-learning-based variants of these models intended to adapt to varying measurement environments, claiming that numerical results demonstrate higher sensing accuracy relative to standard train/test learning schemes.

Significance. If the meta-learning adaptation is rigorously validated, the work could support more practical deployment of WiFi CSI sensing in dynamic settings by reducing the need for environment-specific retraining. The preprocessing step for offset removal is a potentially useful engineering contribution for low-latency applications, but the overall significance cannot be assessed without the missing methodological details on meta-training.

major comments (1)
  1. The central claim that meta-learning models achieve superior accuracy by adapting to different measurement environments rests on an unspecified meta-training procedure. The abstract provides no count of meta-training tasks, no characterization of environment diversity (room size, furniture layout, number of people, or WiFi traffic load), and no explicit model of residual interference after the proposed preprocessing; without these elements the reported performance gap cannot be attributed to the meta-learning construction rather than limited test conditions.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for highlighting the need for greater specificity on the meta-training procedure. We agree that the current manuscript lacks sufficient detail on task counts, environment diversity, and residual interference modeling, which weakens attribution of the reported gains to meta-learning. We will revise the manuscript to supply these elements.

read point-by-point responses

  1. Referee: The central claim that meta-learning models achieve superior accuracy by adapting to different measurement environments rests on an unspecified meta-training procedure. The abstract provides no count of meta-training tasks, no characterization of environment diversity (room size, furniture layout, number of people, or WiFi traffic load), and no explicit model of residual interference after the proposed preprocessing; without these elements the reported performance gap cannot be attributed to the meta-learning construction rather than limited test conditions.

    Authors: We agree that the meta-training procedure is insufficiently specified. In the revised manuscript we will add: (i) the exact number of meta-training tasks and the meta-learning algorithm (e.g., MAML or Reptile) with hyper-parameters; (ii) a table or subsection characterizing the training environments by room dimensions, furniture configurations, occupant counts, and background WiFi traffic levels; (iii) an explicit description or bound on residual interference after the proposed preprocessing step. These additions will be placed in a new subsection of the methods and referenced in the abstract and numerical-results section so that the performance gap can be properly attributed. We will also update the abstract to mention the meta-training task count. revision: yes

Circularity Check

0 steps flagged

No derivation chain or first-principles claims present; results are purely empirical comparisons

full rationale

The paper proposes CSI preprocessing for offset removal followed by pre-training and meta-learning models for people counting/localization, then reports numerical accuracy results versus simple train/test baselines. No equations, mathematical derivations, uniqueness theorems, fitted parameters renamed as predictions, or self-citation load-bearing steps appear in the provided text. The central claim is an empirical performance gap under meta-learning, which cannot reduce to its inputs by construction because no derivation chain exists to inspect. This is the normal non-finding for an applied ML paper whose value rests on experimental outcomes rather than analytic reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no free parameters, axioms, or invented entities can be identified from the given information.

pith-pipeline@v0.9.0 · 5693 in / 997 out tokens · 22327 ms · 2026-05-23T04:08:34.021942+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

40 extracted references · 40 canonical work pages

  1. [1]

    A Survey of Huma n-Sensing: Methods for Detecting Presence, Count, Location, Track, an d Identity,

    T. Teixeira, G. Dublon, and A. Savvides, “A Survey of Huma n-Sensing: Methods for Detecting Presence, Count, Location, Track, an d Identity,” ACM Computing Surveys , vol. 5, no. 1, pp. 59–69, 2010

    work page 2010

  2. [2]

    Opportunities and Cha llenges of Wireless Human Sensing for the Smart IoT World: A Survey,

    Z. Liu, X. Liu, J. Zhang, and K. Li, “Opportunities and Cha llenges of Wireless Human Sensing for the Smart IoT World: A Survey,” IEEE Network, vol. 33, no. 5, pp. 104–110, 2019

    work page 2019

  3. [3]

    Wireless Sen sing for Human Activity: A Survey,

    J. Liu, H. Liu, Y . Chen, Y . Wang, and C. Wang, “Wireless Sen sing for Human Activity: A Survey,” IEEE Communications Surveys & Tutorials, vol. 22, no. 3, pp. 1629–1645, Oct. 2020

    work page 2020

  4. [4]

    Human Activity Sensing with Wireless Signals: A Survey,

    J. Liu, G. Teng, and F. Hong, “Human Activity Sensing with Wireless Signals: A Survey,” Sensors, vol. 20, no. 4, Feb. 2020

    work page 2020

  5. [5]

    Cou nting People in Crowds with a Real-Time Network of Simple Image Sen sors,

    D. B. Y ang, H. H. Gonz´ alez-Ba˜ nos, and J. G. Guibas, “Cou nting People in Crowds with a Real-Time Network of Simple Image Sen sors,” Proceedings Ninth IEEE International Conference on Comput er Vision, pp. 122–129, Oct. 2003

    work page 2003

  6. [6]

    Human Activity Analysis: A Re view,

    J. Aggarwal and M. S. Ryoo, “Human Activity Analysis: A Re view,” ACM Computing Surveys , vol. 43, no. 3, Apr. 2011

    work page 2011

  7. [7]

    A Review on Video-Based Human Activity Recognit ion,

    S. R. Ke, H. L. U. Thuc, Y . J. Lee, J. N. Hwang, J. H. Y oo, and K. H. Choi, “A Review on Video-Based Human Activity Recognit ion,” Computers, vol. 2, no. 2, pp. 88–131, Jun. 2013

    work page 2013

  8. [8]

    Multi-Person Locali zation via RF Body Reflections,

    F. Adib, Z. Kabelac, and D. Katabi, “Multi-Person Locali zation via RF Body Reflections,” 12th USENIX Symposium on Networked Systems Design and Implementation (NSDI 15) , pp. 279–292, May 2015

    work page 2015

  9. [9]

    Soli: Ubiquitous Gestur e Sensing with Millimeter Wave Radar,

    J. Lien, N. Gillian, M. E. Karagozler, P . Amihood, C. Schw esig, E. Olson, H. Raja, and I. Poupyrev, “Soli: Ubiquitous Gestur e Sensing with Millimeter Wave Radar,” ACM Transactions on Graphics , vol. 35, no. 4, Jul. 2016

    work page 2016

  10. [10]

    Remote Monitoring of Human Vital Signs U sing mm- Wave FMCW Radar,

    M. Alizadeh, G. Shaker, J. C. M. De Almeida, P . P . Morita, and S. Safavi-Naeini, “Remote Monitoring of Human Vital Signs U sing mm- Wave FMCW Radar,” IEEE Access , vol. 7, pp. 54 958–54 968, Apr. 2019

    work page 2019

  11. [11]

    CrossCount: Efficient Device-F ree Crowd Counting by Leveraging Transfer Learning,

    D. Khan and I. W.-H. Ho, “CrossCount: Efficient Device-F ree Crowd Counting by Leveraging Transfer Learning,” IEEE Internet of Things Journal, vol. 10, no. 5, pp. 4049–4058, Mar. 2023

    work page 2023

  12. [12]

    Robust and Practica l WiFi Human Sensing Using On-Device Learning with a Domain Adapti ve Model,

    E. Soltanaghaei, R. A. Sharma, Z. Wang, A. Chittilappil ly, A. Luong, E. Giler, K. Hall, S. Elias, and A. Rowe, “Robust and Practica l WiFi Human Sensing Using On-Device Learning with a Domain Adapti ve Model,” Proceedings of the 7th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transp ortation, pp. 150–159, Nov. 2020

    work page 2020

  13. [13]

    DeepCount: Crowd Counting with Wi-Fi Using Deep Learning,

    Y . Zhao, S. Liu, F. Xue, B. Chen, and X. Chen, “DeepCount: Crowd Counting with Wi-Fi Using Deep Learning,” Journal of Communications and Information Networks , vol. 4, no. 3, pp. 38–52, Sep. 2019

    work page 2019

  14. [14]

    Device -Free Human Activity Recognition Using Commercial WiFi Devices,

    W. Wang, A. X. Liu, M. Shahzad, K. Ling, and S. Lu, “Device -Free Human Activity Recognition Using Commercial WiFi Devices, ” IEEE Journal on Selected Areas in Communications , vol. 35, no. 5, pp. 1118– 1131, May 2017

    work page 2017

  15. [15]

    Towards 3D Human Pose Construction Using Wi Fi,

    W. Jiang, H. Xue, C. Miao, S. Wang, S. Lin, C. Tian, S. Mura li, H. Hu, and L. Su, “Towards 3D Human Pose Construction Using Wi Fi,” Proceedings of the 26th Annual International Conference on Mobile Computing and Networking , pp. 1–14, Sep. 2020

    work page 2020

  16. [16]

    A Survey on Behavior Recognition Using WiFi Channel State Informati on,

    S. Y ousefi, H. Narui, S. Dayal, S. Ermon, and S. V alaee, “A Survey on Behavior Recognition Using WiFi Channel State Informati on,” IEEE Communications Magazine , vol. 55, no. 10, pp. 98–104, Oct. 2017

    work page 2017

  17. [17]

    D eep Transfer Learning for Gesture Recognition with WiFi Signal s,

    Q. Bu, G. Y ang, X. Ming, T. Zhang, J. Feng, and J. Zhang, “D eep Transfer Learning for Gesture Recognition with WiFi Signal s,” Personal and Ubiquitous Computing , pp. 1–12, Jan. 2020

    work page 2020

  18. [18]

    Und erstand- ing WiFi Signal Frequency Features for Position-Independe nt Gesture Sensing,

    K. Niu, F. Zhang, X. Wang, Q. Lv, H. Luo, and D. Zhang, “Und erstand- ing WiFi Signal Frequency Features for Position-Independe nt Gesture Sensing,” IEEE Transactions on Mobile Computing , Mar. 2021

    work page 2021

  19. [19]

    Leveraging Transf er Learning in Multiple Human Activity Recognition Using WiFi Signal,

    S. Arshad, C. Feng, R. Y u, and Y . Liu, “Leveraging Transf er Learning in Multiple Human Activity Recognition Using WiFi Signal,” 2019 IEEE 20th International Symposium on ”A W orld of Wireless, Mobil e and Multimedia Networks” (W oWMoM), pp. 1–10, Jun. 2019

    work page 2019

  20. [20]

    WiFiNet: WiFi-Based Indoor Loc alisation Using CNNs,

    N. Hern´ andez, I. Parra, H. Corrales, R. Izquierdo, A. L . Ballardini, C. Salinas, and I. Garc´ ıa, “WiFiNet: WiFi-Based Indoor Loc alisation Using CNNs,” Expert Systems with Applications , vol. 177, p. 114906, 2021

    work page 2021

  21. [21]

    Reliable Deep Learning Based Localization with CSI Finger prints and Multiple Base Stations,

    A. Foliadis, M. H. C. Garcia, R. A. Stirling-Gallacher, and R. S. Thom¨ a, “Reliable Deep Learning Based Localization with CSI Finger prints and Multiple Base Stations,” arXiv:2111.11839, 2021

    work page arXiv 2021

  22. [22]

    Widar: D ecimeter- Level Passive Tracking via V elocity Monitoring with Commod ity Wi- Fi,

    K. Qian, C. Wu, Z. Y ang, Y . Liu, and K. Jamieson, “Widar: D ecimeter- Level Passive Tracking via V elocity Monitoring with Commod ity Wi- Fi,” Proceedings of the 18th ACM International Symposium on Mobi le Ad Hoc Networking and Computing , pp. 1–10, Jul. 2017

    work page 2017

  23. [23]

    Wi dar2.0: Passive Human Tracking with a Single Wi-Fi Link,

    K. Qian, C. Wu, Y . Zhang, G. Zhang, Z. Y ang, and Y . Liu, “Wi dar2.0: Passive Human Tracking with a Single Wi-Fi Link,” Proceedings of the 16th Annual International Conference on Mobile Systems, Ap plications, and Services , pp. 350–361, Jun. 2018

    work page 2018

  24. [24]

    MultiTrack: Mult i-User Track- ing and Activity Recognition Using Commodity WiFi,

    S. Tan, L. Zhang, Z. Wang, and J. Y ang, “MultiTrack: Mult i-User Track- ing and Activity Recognition Using Commodity WiFi,” Proceedings of the 2019 CHI Conference on Human Factors in Computing System s, pp. 1–12, May 2019

    work page 2019

  25. [25]

    Simultaneous Crowd Estimation in Counting and Localizati on Using WiFi CSI,

    H. Choi, T. Matsui, S. Misaki, A. Miyaji, M. Fujimoto, an d K. Y asumoto, “Simultaneous Crowd Estimation in Counting and Localizati on Using WiFi CSI,” 2021 International Conference on Indoor Positioning and Indoor Navigation (IPIN) , pp. 1–8, 2021

    work page 2021

  26. [26]

    Device-Free Activity Detection and Wireless Localizatio n Based on CNN Using Channel State Information Measurement,

    J. Y an, L. Wan, W. Wei, X. Wu, W. P . Zhu, and D. P . K. Lun, “Device-Free Activity Detection and Wireless Localizatio n Based on CNN Using Channel State Information Measurement,” IEEE Sensors Journal, vol. 21, no. 21, pp. 24 482–24 494, Nov. 2021

    work page 2021

  27. [27]

    Precise Power Delay Profiling wi th Commod- ity Wi-Fi,

    Y . Xie, Z. Li, and M. Li, “Precise Power Delay Profiling wi th Commod- ity Wi-Fi,” IEEE Transactions on Mobile Computing , vol. 18, no. 6, pp. 1342–1355, Jun. 2019

    work page 2019

  28. [28]

    Tensorbeat: Tensor decomp osition for monitoring multiperson breathing beats with commodity wifi,

    X. Wang, C. Y ang, and S. Mao, “Tensorbeat: Tensor decomp osition for monitoring multiperson breathing beats with commodity wifi,” ACM Trans. Intell. Syst. Technol. , vol. 9, no. 1, Sep. 2017

    work page 2017

  29. [29]

    CSI-Based Finger printing for Indoor Localization: A Deep Learning Approach,

    X. Wang, L. Gao, S. Mao, and S. Pandey, “CSI-Based Finger printing for Indoor Localization: A Deep Learning Approach,” IEEE Transactions on V ehicular Technology, vol. 66, no. 1, pp. 763–776, Jan. 2017

    work page 2017

  30. [30]

    Model-Agnostic Meta -Learning for Fast Adaptation of Deep Networks,

    C. Finn, P . Abbeel, and S. Levine, “Model-Agnostic Meta -Learning for Fast Adaptation of Deep Networks,” Proceedings of the 34th Interna- tional Conference on Machine Learning , vol. 70, pp. 1126–1135, Aug. 2017

    work page 2017

  31. [31]

    Optimum rece iver design for wireless broad-band systems using ofdm. i,

    M. Speth, S. Fechtel, G. Fock, and H. Meyr, “Optimum rece iver design for wireless broad-band systems using ofdm. i,” IEEE Transactions on Communications, vol. 47, no. 11, pp. 1668–1677, Nov. 1999

    work page 1999

  32. [32]

    A Survey on Transfer Learning,

    S. J. Pan and Q. Y ang, “A Survey on Transfer Learning,” IEEE Transactions on Knowledge and Data Engineering , vol. 22, no. 10, pp. 1345–1359, Oct. 2010

    work page 2010

  33. [33]

    Detecting R adio Frequency Interference for CSI Measurements on COTS WiFi De vices,

    Y . Zheng, C. Wu, K. Qian, Z. Y ang, and Y . Liu, “Detecting R adio Frequency Interference for CSI Measurements on COTS WiFi De vices,” in 2017 IEEE International Conference on Communications (ICC ), 2017, pp. 1–6

    work page 2017

  34. [34]

    Towards Anti-Interference Human Activity Recognition Based on WiF i Subcar- rier Correlation Selection,

    J. Huang, B. Liu, C. Chen, H. Jin, Z. Liu, C. Zhang, and N. Y u, “Towards Anti-Interference Human Activity Recognition Based on WiF i Subcar- rier Correlation Selection,” IEEE Transactions on V ehicular Technology, vol. 69, no. 6, pp. 6739–6754, Jun. 2020

    work page 2020

  35. [35]

    Deep Channel Learning for Large Intelligent Surfaces Aide d mm-Wave Massive MIMO Systems,

    A. M. Elbir, A. Papazafeiropoulos, P . Kourtessis, and S . Chatzinotas, “Deep Channel Learning for Large Intelligent Surfaces Aide d mm-Wave Massive MIMO Systems,” IEEE Wireless Communications Letters , vol. 9, no. 9, pp. 1447–1451, 2020

    work page 2020

  36. [36]

    Hunger, Floating point operations in matrix-vector calculus

    R. Hunger, Floating point operations in matrix-vector calculus . Munich University of Technology, Inst. for Circuit Theory and Sign al, 2005

    work page 2005

  37. [37]

    On complexity analysis o f super- vised MLP-learning for algorithmic comparisons,

    E. Mizutani and S. E. Dreyfus, “On complexity analysis o f super- vised MLP-learning for algorithmic comparisons,” in IJCNN’01. In- ternational Joint Conference on Neural Networks. Proceedi ngs (Cat. No.01CH37222), vol. 1, Jul. 2001, pp. 347–352

    work page 2001

  38. [38]

    Hardware complexity analys is of deep neural networks and decision tree ensembles for real-time n eural data classification,

    M. Taghavi and M. Shoaran, “Hardware complexity analys is of deep neural networks and decision tree ensembles for real-time n eural data classification,” in 2019 9th International IEEE/EMBS Conference on Neural Engineering (NER) , Mar. 2019, pp. 407–410

    work page 2019

  39. [39]

    Mass ive MIMO channel prediction: Kalman filtering vs. machine learn ing,

    H. Kim, S. Kim, H. Lee, C. Jang, Y . Choi, and J. Choi, “Mass ive MIMO channel prediction: Kalman filtering vs. machine learn ing,” IEEE Transactions on Communications, vol. 69, no. 1, pp. 518–528, Jan. 2021

    work page 2021

  40. [40]

    Massive MIMO channel pre diction via meta-learning and deep denoising: Is a small dataset eno ugh?

    H. Kim, J. Choi, and D. J. Love, “Massive MIMO channel pre diction via meta-learning and deep denoising: Is a small dataset eno ugh?” IEEE Transactions on Wireless Communications , vol. 22, no. 12, pp. 9278– 9290, Dec. 2023

    work page 2023