Input-Correlated Supervision Noise Limits the Benefits of OTA Training for Learned Receivers

Dani Korpi; Harri Holma; Olav Tirkkonen; Riku Luostari

REVIEW 3 major objections 4 minor 19 references

Input-Correlated Supervision Noise Limits the Benefits of OTA Training for Learned Receivers

T0 review · 3 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper shows that increasing over-the-air training data monotonically improves an end-to-end neural receiver, but degrades a neural channel estimator past an optimum because its labels inherit noise from the input.

desk verdict A careful OTA measurement study showing a real asymmetry between end-to-end and channel-estimator receivers as OTA fraction increases, with a plausible but not fully pinned-down causal mechanism. read the letter →

arxiv 2608.12918 v1 pith:L4ZO6JB4 submitted 2026-08-13 eess.SP

Riku Luostari , Dani Korpi , Olav Tirkkonen , Harri Holma This is my paper · ORCID

classification eess.SP

keywords over-the-airtrainingneuralreceiverschannelestimationsupervisionnoiseOFDMend-to-endlearninglabelmeasurementcampaign

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 asks whether measured over-the-air (OTA) data actually helps learned wireless receivers and shows the answer depends on what the network is trained to predict. Training a capacity-matched end-to-end receiver and a neural channel estimator on mixtures of simulated and measured 5.88 GHz OFDM data, it finds that the end-to-end receiver improves monotonically as the measured fraction grows, while the channel estimator peaks at an intermediate fraction and degrades under fully measured training. The cause is the supervision target: OTA channel labels are built by dividing the received signal by the known transmitted symbol, so the label noise is the same noise realization present at the network input. An MSE-trained estimator converges to the observation-derived target rather than the clean channel, and additional measurements reduce variance but cannot remove the residual bias. CRC-validated decoded bits, by contrast, provide effectively error-free supervision, which is why end-to-end training keeps gaining from measured data.

What carries the argument

The load-bearing object is the OTA channel-label construction $H^{\mathrm{gt}}_{ij}=y_{ij}/x_{ij}$, which extends the least-squares pilot estimate to the full time-frequency grid. Its error term $n_{ij}/x_{ij}$ shares the noise realization with the network input $Y$, so under MSE training the estimator converges to the conditional expectation of the label given the input and retains a bias that extra data cannot remove; this is what the paper calls input-correlated supervision noise. The controlled intervention is the same separable time-frequency average used by decision-directed estimation (kernels $k_f=5$, $k_t=3$), applied to the labels only while input, architecture, and optimization stay fixed. That isolates the supervision mechanism from measurement diversity.

What would settle it

A decisive test: in a simulator with unlimited samples, corrupt synthetic channel labels as $H + \alpha n/x$ with the same noise realization $n$ that appears in the input, and sweep $\alpha$ from 0 (independent noise) to 1 (fully correlated). If the fully correlated case reproduces the channel estimator's peak-and-degrade pattern at high OTA fractions and the independent case does not, the paper's mechanism is confirmed; if both degrade equally, the cause is elsewhere.

Watch

Extended reading notes

Core claim

The central claim is that OTA training data is not uniformly beneficial: it consistently helps an end-to-end receiver (CM DeepRx) while limiting a capacity-matched channel estimator (ACM-E) at high measured-data fractions. The paper attributes this asymmetry to input-correlated supervision noise. At each resource element, the OTA channel label is $H^{\mathrm{gt}}_{ij}=y_{ij}/x_{ij}=H_{ij}+n_{ij}/x_{ij}$, where $n_{ij}$ is the same noise realization already present in the input $Y$. Because the MSE optimum is the conditional expectation of the target given the input, the correlated error term contributes a non-zero bias that does not vanish as the dataset grows. Bit-level supervision from CRC-validated decoded bits does not inherit input noise, so the end-to-end receiver keeps improving with OTA fraction. A controlled experiment that denoised the channel labels with separable time-frequency smoothing removed most of the high-OTA degradation, confirming that input-correlated supervision noise, not limited data diversity, is responsible.

Load-bearing premise

The load-bearing premise, stated in Sec. III-D, is that CRC-validated decoded bits are effectively error-free supervision: residual undetected decoding errors are assumed negligible, so the end-to-end bit targets do not inherit input-correlated noise.

Editorial extensions

If this is right

End-to-end neural receivers trained on measured data should be trained with the largest available OTA fraction; the paper observed monotonic improvement through $p_m=100\%$.
Neural channel estimators should be trained with moderate OTA fractions (about 40% in this study) and denoised labels; fully measured raw labels can underperform simulation-only training.
Adding more OTA measurements cannot cure bias caused by input-correlated supervision noise; the fix is to change the target, not grow the dataset.
When supervision targets must be derived from noisy observations, label denoising should be applied before training, for example the same smoothing used by decision-directed estimation.
The same principle constrains any learned receiver stage whose target is a function of the received signal, not just channel estimators in OFDM.

Reading between the lines

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

The same mechanism should bind any learned intermediate estimator whose target is constructed from the received signal, such as MIMO channel estimators using pilot-division labels; the paper notes MIMO is out of scope, but the argument does not rely on SISO-specific structure beyond the label construction.
A testable extension: replacing noisy OTA channel labels with estimates from a separate clean reference pass, such as repeated pilot averaging or a second receiver chain, should push the optimal OTA fraction toward 100%, because the bias term would then be uncorrelated with the input.
The paper's asymmetry suggests a design heuristic for future receiver pipelines: prefer bit-level or semantic supervision over intermediate physical quantities whenever the intermediate quantity is observed through the same noisy channel; alternatively, learn the intermediate quantity with a self-supervised objective that avoids noisy explicit targets.
The optimal OTA fraction for channel estimators is likely deployment-dependent, but the qualitative shape of improvement followed by decline should transfer across architectures and frequencies; practitioners can probe it with a small denoising sweep rather than assuming more measurements are always better.

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

The short version: this paper reports a genuine, useful empirical finding. On a 5.88 GHz measurement campaign, an end-to-end learned receiver keeps improving as you add OTA data, while a learned channel estimator peaks around 40% OTA and degrades at 100%. That asymmetry is new, and the proposed explanation — OTA channel labels are built from the same noisy received signal the network sees as input, so the label noise is correlated with the input and cannot be averaged away — is correct as a theoretical statement. The authors also run a label-denoising experiment intended as a causal check.

What the paper does well: the dataset is substantial (~44k OFDM slots across urban, suburban, rural, and industrial environments, pedestrian and vehicular mobility), the training protocol is careful (five seeds, standard deviations on the headline metrics, controlled OTA fractions), and the write-up is honest about limitations. The MSE conditional-expectation argument in Eq. (6) is textbook but applied cleanly to a practical problem. The practical guidance — use fully measured data for end-to-end receivers, but improve label quality for intermediate estimators — is actionable and clearly stated.

The soft spots. The denoising experiment in Sec. V-C does not isolate the correlation component. Smoothing the labels changes three things at once: it lowers label-noise variance, it biases the target relative to the genie channel, and it acts as a regularizer against overfitting to the finite ~40k-slot OTA training set. The improvement after denoising is consistent with the correlation story, but also with a finite-data/overfitting story. Because pm=100 means roughly 80 passes over the same OTA slots (with input augmentation), high-pm degradation could be partly memorization. The paper's claim that more measurements cannot remove the bias is true in the infinite-data limit, but the experiments never approach that limit; they don't show that more OTA data at pm=100 fails to close the gap. That is the main gap.

A secondary caveat: the capacity matching is approximate. CM DeepRx has about 15% more parameters and 50% more FLOPs, and is trained four times longer, so the end-to-end versus modular comparison is not purely about supervision. The authors acknowledge this, so it is a minor issue, not a fatal one.

Overall, the empirical finding is solid and the theoretical framing is sensible, but the causal claim — that correlation, not data diversity, causes the degradation — is overstrong relative to the evidence. A serious referee should ask for a control with independent label noise (e.g., corrupting targets with synthetic noise uncorrelated with the input) or a finite-data analysis that separates correlation from overfitting. That is a major-revision request, not a rejection; the core contribution survives.

Recommendation: send it to peer review. Anyone working on learned receivers or OTA training will want this paper, and the authors have the measurements and careful methodology to address the remaining questions.

Referee Report

3 major / 4 minor

Summary. The paper reports a 5.88 GHz over-the-air (OTA) OFDM measurement campaign and uses it to train two learned receivers with controlled mixtures of measured and synthetic data: ACM-E, a neural channel estimator, and CM DeepRx, an end-to-end LLR receiver. The central empirical finding is an asymmetry: CM DeepRx improves monotonically as the OTA fraction increases, while ACM-E peaks at an intermediate OTA fraction and degrades at p_m = 100%. The paper attributes this to input-correlated supervision noise: OTA channel labels are constructed as y_ij/x_ij, so the label error shares the noise realization present in the network input, and the MSE objective converges to a conditional expectation that retains a bias (Eq. 6). A denoising experiment that smooths the OTA labels with the DDCE kernel is presented as confirming the mechanism. The paper concludes with practical guidance: end-to-end receivers benefit from fully measured training, whereas channel estimators require improved supervision quality rather than simply more OTA data.

Significance. If the causal claim holds, the paper gives the first systematic OTA evidence that the value of measured training data depends on the supervision target, and it identifies a mechanism—input-correlated supervision noise—that is distinct from the usual independent-label-noise setting. The measurement campaign is substantial, the experiments use five seeds with standard deviations on the headline metric, and the evaluation is on held-out OTA data, which is a clear strength. The theoretical argument in Sec. IV is a straightforward and correct application of the conditional-expectation property of MSE training, and the within-ACM-E denoising intervention changes only the target construction. These strengths make the paper a useful contribution to the practical literature on training learned receivers with real-world data, provided the causal attribution is supported or appropriately qualified.

major comments (3)

[Sec. V-C] The denoising experiment does not uniquely identify input-correlated supervision noise as the cause of the high-OTA degradation. Smoothing the label y_ij/x_ij with the DDCE kernel simultaneously (i) reduces the variance of the label noise regardless of whether that noise is correlated with the input, (ii) introduces a bias relative to the true channel, and (iii) acts as a regularizer against overfitting to the fixed pool of roughly 44,000 measured slots, which is increasingly re-used as p_m grows. Equation (6) proves that in the infinite-data limit correlated label noise leaves a residual bias, but the experiments do not approach that limit and do not show that more unique OTA data at p_m = 100% fails to close the gap. A control that corrupts synthetic labels with independent noise of the same variance, or one that compares smoothed correlated versus smoothed independent noise at matched variance, is needed to support the abstract's causal claim that the correlation, rather than limited data diversity or finite-data overfitting, caused the degradation.
[Sec. III-D] The claim that CRC-validated decoded bits provide 'effectively error-free' supervision rests on an unquantified assumption: the paper discards the fewer than 1% of slots failing CRC, but it does not bound the rate of undetected decoding errors in accepted frames. If residual undetected errors are non-negligible, CM DeepRx's bit targets would also carry input-correlated noise, and the asymmetry attributed to the supervision objective would be confounded. Please provide an estimate of the residual post-CRC frame error probability, for example by re-encoding and re-modulating decoded bits and comparing with the received signal on held-out data, or explain quantitatively why code rate 0.25 combined with CRC makes this assumption safe.
[Sec. III-C] The statement that comparing architectures at similar parameter counts 'isolates the benefit of end-to-end optimization from differences in model size' is stronger than the evidence supports. CM DeepRx has 475K parameters versus 412K for ACM-E (about 15% more), roughly 1.9 times the MFLOPs, and a different architecture family and inductive bias. The cross-architecture asymmetry could therefore be influenced by architecture rather than purely by the supervision objective. The within-ACM-E denoising experiment does control architecture, but the headline comparison of 'capacity-matched' models should either be softened or supplemented by a sensitivity analysis over parameter count or latent dimension.

minor comments (4)

[Sec. V-B / Table IV] At the headline operating point, the raw-OTA degradation of ACM-E is small: the required SINR at BER = 10^-2 is 16.07 dB at p_m = 100% versus 15.86 dB at p_m = 0%, a 0.21 dB gap. This is significant relative to the standard deviations reported, but the practical magnitude should be stated more explicitly so that readers can calibrate the 'degradation' claim.
[Sec. IV] The notation 'E[H_ij | Y] + E[n_ij/x_ij | Y]' in Eq. (6) is correct, but the sentence 'the label error shares the same realization of measurement noise as the input' could be misread as saying the error is exactly n_ij; it is n_ij/x_ij, which is known only up to the transmitted symbol. A brief clarifying remark that x_ij is known at training time would improve readability.
[Sec. V-C] The phrase 'controlled denoising experiment' is potentially misleading because the intervention changes the variance and bias of the target in addition to the correlation structure; consider renaming it a 'target-smoothing experiment' or explicitly listing the three confounded effects.
[References] Reference [12] is typeset as 'AI - Based' with an unintended space; this should be corrected.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: Eq. (6) is a direct conditional-expectation identity, and the denoising experiment is an independent manipulation rather than a fitted re-statement.

full rationale

The paper's central claim is not circular. Section IV derives the residual bias from the definition of the OTA label H^gt_ij = y_ij/x_ij = H_ij + n_ij/x_ij and the standard MSE property that the minimizer is the conditional expectation E[H^gt|Y]. This is a mathematical identity with no fitted parameters, and the statement that additional data cannot remove the conditional-mean bias follows from the identity rather than from the experimental results. The observed asymmetry between CM DeepRx and ACM-E is measured, and the Sec. V-C denoising experiment changes only the target construction while keeping input, architecture, and optimization fixed, so the mechanism claim is not equivalent to its inputs. The only self-citation, [12], is used to motivate the topic and is not load-bearing; no uniqueness theorem or ansatz is imported from prior author work. The skeptic's concern that smoothing also reduces label-noise variance and acts as regularization is a possible threat to causal uniqueness, but that is a validity confound, not a circularity, and does not affect the self-contained derivation. No circular step is identified.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The theory introduces no new entities or fitted constants. The only hand-picked numbers are kernel sizes and architecture widths, which influence quantitative results but not the qualitative mechanism. The core derivation is a textbook application of conditional expectation.

free parameters (3)

DDCE/denoising smoothing kernel sizes (kf, kt) = 5, 3
Hand-chosen separable uniform smoothing kernels used for the DDCE reference and for the denoised OTA labels in Sec. V-C; the quantitative effect of denoising depends on this choice but the qualitative trend does not.
ACM-E encoder width d and depthwise kernel sizes = d=128, Kt=3, Kf=5
Architecture hyperparameters chosen by the authors; they affect the capacity comparison with CM DeepRx but are not fitted to the central claim.
CM DeepRx hidden dimension d and number of blocks L = d=256, L=6
Chosen to roughly match ACM-E capacity (within ~15%); the approximate match underpins the attribution of the asymmetry to supervision rather than architecture.

assumptions (4)

standard math The MSE-optimal predictor equals the conditional expectation of the target given the input, E[H^gt|Y].
Invoked in Sec. IV Eq. (6); standard result for mean squared error loss.
domain assumption OTA channel labels H^gt = y/x are unbiased for the true channel, with label error correlated with the input Y through the shared noise realization.
Stated in Sec. III-B and Sec. IV Eq. (5); this correlation is the basis of the claimed mechanism.
domain assumption CRC-validated decoded bits are effectively error-free, making bit supervision independent of the input noise.
Stated in Sec. III-D; underpins the positive result for CM DeepRx and the asymmetry claim.
domain assumption The conditional expectation of the label-error term n/x given Y is non-zero under OTA measurement.
Sec. IV; requires the label and input to share the same noise realization, which holds under the measurement model.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Input-Correlated Supervision Noise Limits the Benefits of OTA Training for Learned Receivers." pith.science (2026). https://pith.science/paper/L4ZO6JB4

@misc{pith2026260812918,
  author       = {Pith},
  title        = {Pith review of: Input-Correlated Supervision Noise Limits the Benefits of OTA Training for Learned Receivers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L4ZO6JB4}},
  note         = {Machine review of arXiv:2608.12918}
}

read the original abstract

While learned wireless receivers are typically studied using synthetic data, the impact of over-the-air (OTA) measurements for training remains unclear. We conducted a 5.88 GHz measurement campaign with a 5G/6G-like orthogonal frequency-division multiplexing (OFDM) system across diverse environments and mobility conditions, and trained a neural channel estimator and a capacity-matched end-to-end neural receiver using mixtures of measured and synthetic data. Increasing the OTA fraction revealed a fundamental asymmetry: measured data consistently improved the end-to-end receiver, whereas the channel estimator peaked at an intermediate fraction and degraded with fully measured training. We showed that this difference arises from the supervision target: OTA channel labels are derived from noisy received signals and therefore contain supervision errors correlated with the receiver input, whereas decoded bits validated by a cyclic redundancy check (CRC) provide effectively error-free supervision. A controlled denoising experiment confirmed that this correlation, rather than limited data diversity, caused the degradation. These results provide practical guidance for training learned receivers with OTA data: end-to-end receivers benefit from fully measured training, whereas channel estimators benefit from moderate OTA fractions but require improved label quality, e.g. via denoising, to unlock further gains.

Figures

Figures reproduced from arXiv: 2608.12918 by the authors.

**Figure 3.** Equivalent SNR gain over the simulation-only baseline ( [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

**Figure 2.** Equivalent SNR gain over the simulation-only baseline ( [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Equivalent SNR gain over the simulation-only baseline ( [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]

**Figure 5.** Figure 5: OTA BER versus SINR at 16-QAM for LS, DDCE, ACM-E [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

19 extracted references · 16 canonical work pages

[1]

A neural network based channel estimation scheme for OFDM system,

K. Hiray and K. V . Babu, “A neural network based channel estimation scheme for OFDM system,” in2016 International Conference on Communication and Signal Processing (ICCSP). Melmaruvathur, Tamilnadu, India: IEEE, Apr. 2016, pp. 0438–0441

work page 2016
[2]

A Low Complexity Learning-Based Channel Estimation for OFDM Systems With Online Training,

K. Meiet al., “A Low Complexity Learning-Based Channel Estimation for OFDM Systems With Online Training,”IEEE Transactions on Communications, vol. 69, no. 10, pp. 6722–6733, Oct. 2021

work page 2021
[3]

DeepRx: Fully Convolutional Deep Learning Receiver,

M. Honkala, D. Korpi, and J. M. J. Huttunen, “DeepRx: Fully Convolutional Deep Learning Receiver,”IEEE Transactions on Wireless Communications, vol. 20, no. 6, pp. 3925–3940, Jun. 2021

work page 2021
[4]

HybridDeepRx: Deep Learning Receiver for High-EVM Signals,

J. Pihlajasaloet al., “HybridDeepRx: Deep Learning Receiver for High-EVM Signals,” 2021, version Number: 1

work page 2021
[5]

Study on 6G Scenarios and Requirements,

3GPP, “Study on 6G Scenarios and Requirements,” 3rd Generation Partnership Project, Technical Contribution TR 38.914, Version 20.0.0, 2026

work page 2026
[6]

Feature Lead summary on 6G waveform,

3GPP, “Feature Lead summary on 6G waveform,” inTSG RAN WG1 Meeting #122bis, vol. TDoc R1-2508043, 2025

work page 2025
[7]

Channel estimation techniques based on pilot arrangement in OFDM systems,

S. Coleriet al., “Channel estimation techniques based on pilot arrangement in OFDM systems,”IEEE Transactions on Broadcasting, vol. 48, no. 3, pp. 223–229, Sep. 2002

work page 2002
[8]

A comparison of pilot-aided channel estimation methods for OFDM systems,

M. Morelli and U. Mengali, “A comparison of pilot-aided channel estimation methods for OFDM systems,”IEEE Transactions on Signal Processing, vol. 49, no. 12, pp. 3065–3073, Dec. 2001

work page 2001

Show all 19 references

[9]

CD3-OFDM: a novel demodulation scheme for fixed and mobile receivers,

V . Mignone and A. Morello, “CD3-OFDM: a novel demodulation scheme for fixed and mobile receivers,”IEEE Transactions on Communications, vol. 44, no. 9, pp. 1144–1151, Sep. 1996

1996
[10]

A comparative study of iterative channel estimators for mobile ofdm systems,

F. Sanzi, S. Jelting, and J. Speidel, “A comparative study of iterative channel estimators for mobile ofdm systems,”IEEE Transactions on Wireless Communications, vol. 2, no. 5, pp. 849–859, Sep. 2003

2003
[11]

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

C. Komninakiset al., “Multi-input multi-output fading channel tracking and equalization using Kalman estimation,”IEEE Transactions on Signal Processing, vol. 50, no. 5, pp. 1065–1076, May 2002

2002
[12]

Adapting to Reality: Over-the-Air Validation of AI - Based Receivers Trained with Simulated Channels,

R. Luostariet al., “Adapting to Reality: Over-the-Air Validation of AI - Based Receivers Trained with Simulated Channels,” in2025 IEEE Wireless Communications and Networking Conference (WCNC), Milan, Italy, Mar. 2025, pp. 01–06

2025
[13]

T. F. Collinset al.,Software-defined radio for engineers, ser. Artech House mobile communications series. Boston London: Artech House, 2018

2018
[14]

Axial Attention in Multidimensional Transformers,

J. Hoet al., “Axial Attention in Multidimensional Transformers,” Dec. 2019, arXiv:1912.12180 [cs]

2019 arXiv
[15]

Patches Are All You Need?

A. Trockman and J. Z. Kolter, “Patches Are All You Need?” Jan. 2022, arXiv:2201.09792 [cs]

2022 arXiv
[16]

Deep Residual Learning for Image Recognition,

K. Heet al., “Deep Residual Learning for Image Recognition,” in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV , USA, Jun. 2016, pp. 770–778

2016
[17]

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

3GPP, “Study on channel model for frequencies from 0.5 to 100 GHz,” Technical Specification (TS) TS 38.901, Sep. 2020

2020
[18]

Sionna: An Open-Source Library for Next-Generation Physical Layer Research,

J. Hoydiset al., “Sionna: An Open-Source Library for Next-Generation Physical Layer Research,” Mar. 2023, arXiv:2203.11854 [cs]

2023 arXiv
[19]

Classification in the Presence of Label Noise: A Survey,

B. Frenay and M. Verleysen, “Classification in the Presence of Label Noise: A Survey,”IEEE Transactions on Neural Networks and Learning Systems, vol. 25, no. 5, pp. 845–869, May 2014. 6

2014

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

A neural network based channel estimation scheme for OFDM system,

K. Hiray and K. V . Babu, “A neural network based channel estimation scheme for OFDM system,” in2016 International Conference on Communication and Signal Processing (ICCSP). Melmaruvathur, Tamilnadu, India: IEEE, Apr. 2016, pp. 0438–0441

work page 2016

[2] [2]

A Low Complexity Learning-Based Channel Estimation for OFDM Systems With Online Training,

K. Meiet al., “A Low Complexity Learning-Based Channel Estimation for OFDM Systems With Online Training,”IEEE Transactions on Communications, vol. 69, no. 10, pp. 6722–6733, Oct. 2021

work page 2021

[3] [3]

DeepRx: Fully Convolutional Deep Learning Receiver,

M. Honkala, D. Korpi, and J. M. J. Huttunen, “DeepRx: Fully Convolutional Deep Learning Receiver,”IEEE Transactions on Wireless Communications, vol. 20, no. 6, pp. 3925–3940, Jun. 2021

work page 2021

[4] [4]

HybridDeepRx: Deep Learning Receiver for High-EVM Signals,

J. Pihlajasaloet al., “HybridDeepRx: Deep Learning Receiver for High-EVM Signals,” 2021, version Number: 1

work page 2021

[5] [5]

Study on 6G Scenarios and Requirements,

3GPP, “Study on 6G Scenarios and Requirements,” 3rd Generation Partnership Project, Technical Contribution TR 38.914, Version 20.0.0, 2026

work page 2026

[6] [6]

Feature Lead summary on 6G waveform,

3GPP, “Feature Lead summary on 6G waveform,” inTSG RAN WG1 Meeting #122bis, vol. TDoc R1-2508043, 2025

work page 2025

[7] [7]

Channel estimation techniques based on pilot arrangement in OFDM systems,

S. Coleriet al., “Channel estimation techniques based on pilot arrangement in OFDM systems,”IEEE Transactions on Broadcasting, vol. 48, no. 3, pp. 223–229, Sep. 2002

work page 2002

[8] [8]

A comparison of pilot-aided channel estimation methods for OFDM systems,

M. Morelli and U. Mengali, “A comparison of pilot-aided channel estimation methods for OFDM systems,”IEEE Transactions on Signal Processing, vol. 49, no. 12, pp. 3065–3073, Dec. 2001

work page 2001