Pith. sign in

REVIEW 4 major objections 4 minor 32 references

Online Adaptive Real-Time Beamforming Design for Dynamic Environments in Cell-Free Systems

T0 review · 4 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims a CNN can deliver millisecond beamforming for cell-free systems as channels, access points, and user counts change, and that an online update of about 3% of its parameters further improves the achieved sum rate.

desk verdict A well-motivated architecture with a likely fatal input-representation flaw that undermines the headline result. read the letter →

arxiv 2412.09629 v1 pith:OFRLP3X6 submitted 2024-11-27 eess.SP cs.ITcs.NImath.IT

Guanghui Chen , Zheng Wang , Hongxin Lin , Pengguang Du , Yongming Huang This is my paper · ORCID
classification eess.SPcs.ITcs.NImath.IT
keywords cell-freesystemsbeamformingdesignreal-timeprocessingdeeplearningdomaingeneralizationonlineadaptationconvolutionalneuralnetworksum-ratemaximization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to solve real-time beamforming design in cell-free wireless systems where channels, user counts, and access-point counts change between periods. It proposes a convolutional network, HGNet, that maps CSI directly to beamforming vectors, plus a module that discards channel-sensitive features to improve generalization across changing channel distributions. A theoretical bound is given showing the module reduces an upper bound on the generalization error. To track a changing environment online, the OAU algorithm updates only the batch-normalization affine parameters (about 3% of the weights) using an information-entropy loss, keeping computation in the millisecond range. Numerical comparisons report higher sum rates and lower compute than WMMSE, Edge-GNN, and SUNet.

What carries the argument

The load-bearing object is the high-generalization beamforming module embedded in the convolutional stack. It uses a gradient reversal layer to make the feature extractor adversarially ignore channel identity, a sensitive-feature discriminator that scores each feature channel by its contribution to predicting the channel distribution, and weighted random selection to build a binary mask that zeroes out the most sensitive channels during training. The masked features then flow on to produce the beamforming tensor. Online, OAU updates only the affine parameters $\gamma_l$ and $\beta_l$ of each batch-normalization layer, about 3% of all parameters, using an information-entropy loss in place of the expensive sum-rate loss.

What would settle it

Take a channel with a fixed set of magnitudes $|H_{ij}|$ and vary only the phases of the complex channel entries; feed the two phase-different versions through the trained HGNet and compare the achieved sum rates, since the network's stated input cannot distinguish them, any difference in output rate would reveal hidden phase use, while identical rates would show that phase-sensitive beamforming gains are not being achieved.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a fixed CNN can serve as a real-time beamformer for dynamic cell-free environments if (i) its convolution architecture is chosen so output dimensions automatically track the varying AP and user counts, (ii) an adversarial feature-discarding module removes the feature-map components most predictive of which channel distribution a sample came from, and (iii) at inference time only the scale and shift parameters of the batch-normalization layers are updated with a cheap entropy objective. The paper proves that the feature-discarding step lowers a G-MMD-based upper bound on the generalization error of the feature extractor, and it reports that the combined system achieves higher average sum rate than WMMSE, Edge-GNN, and SUNet while running in the $10^{-3}$ to $10^{-2}$ second range.

Load-bearing premise

Everything rests on the assumption that beamforming can be learned from channel magnitudes alone, because the input module converts complex CSI into a real tensor of modulus values and discards phase, yet optimal beamforming depends on phase alignment and interference structure.

Editorial extensions

If this is right

  • If the central claim is right, cell-free systems can re-optimize beamformers between periods without retraining, because the CNN architecture is designed so output dimensions follow the input dimensions as AP and user counts change.
  • Operators can keep one trained network deployed across environments and use the OAU update to recover sum rate after a channel-distribution shift, at a reported computational cost below $10^{-2}$ seconds.
  • The feature-discarding module directly targets distribution shift: by removing features that best predict which channel model produced the data, the network transfers across Rayleigh, Rician, and geometric multi-path channels better than fixed-configuration GNN or CNN baselines.
  • The generalization benefit is presented as structural rather than incidental, because the high-generalization module is proven to lower the G-MMD upper bound on the feature extractor's generalization error.

Reading between the lines

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

  • Editorial extension: a natural ablation the paper does not run is replacing the modulus-only input with a magnitude-plus-phase or complex two-channel input; such an ablation would show whether the reported gains come from the generalization module or from phase information the current input discards.
  • Editorial reading: the G-MMD bound in Proposition 3 is a bound on the feature extractor, not on end-to-end achievable sum rate; the sum-rate claim is carried by the numerical comparisons, with the theorem supporting the feature-transfer component.
  • Editorial extension: the OAU recipe of freezing the network and tuning only batch-normalization affine parameters with a cheap surrogate loss could transfer to other wireless optimization tasks with differentiable objectives and varying system sizes.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper considers downlink beamforming design in cell-free systems under dynamic wireless environments in which channel distributions and the numbers of APs and users change between periods. The authors formulate a sum-rate maximization problem and propose HGNet, a CNN-based architecture whose output dimensions adapt to variable AP/user counts, together with a high-generalization module that masks features deemed sensitive to period identity. They also propose an OAU algorithm that online updates about 3% of the network parameters, namely the affine parameters of batch-normalization layers, using a computationally cheap information-entropy-like loss. The paper claims a theoretical proof that the masking module decreases an upper bound on generalization error, and numerical results compare HGNet with OAU against WMMSE, Edge-GNN, and SUNet in terms of sum rate and computation time.

Significance. The problem is timely, and the paper contains useful engineering pieces: Propositions 1 and 2 give explicit architectural conditions that let a CNN output layer track variable AP/user counts; the partial-update strategy of updating only batch-normalization affine parameters is computationally plausible; and the experiments compare against WMMSE and two recent learning-based baselines rather than only toy settings. If the method worked as described, it would be a practical real-time beamforming solution with a generalization story. However, several load-bearing points are not supported: the input representation appears to discard channel phase, the claimed bound reduction from feature masking is not proven, and the surrogate loss used for online adaptation is not a valid information entropy nor is it linked to sum rate. These issues undermine the central claims of the paper.

major comments (4)
  1. [Section III-A] The input module converts the complex channel H_t into a real tensor H3D_t,mod in R^{Q_t x I_t x MN} 'by computing modulus values'. This keeps only the per-entry magnitudes of the complex channel and discards all phase information. For beamforming in a MIMO cell-free system, relative phases across the antennas of each AP and across APs determine coherent combining and interference suppression; two channels with identical magnitudes but different relative phases can have different optimal beamformers and different WMMSE sum rates. Since the unsupervised training loss in Section III-D is computed from the true complex CSI, a network fed only magnitudes can at best minimize an average over phase realizations and cannot track the per-channel optimum. The paper provides no theorem or ablation showing that phase-free CSI is sufficient, and no experiment uses a full-CSI input to HGNet. If the implementation actually feeds phase information, the description of Section III-A and the stated tensor dimension are inconsistent. Either way, the central empirical claim is not supported by the described method.
  2. [Section III-E and Appendix A] Proposition 3 and its proof derive a generic MMD-type upper bound on the generalization error for the output of G(·, θ_l). The proof does not compare the masking operation with a baseline without masking, and it does not bound the G-MMD distance of the masked output relative to the unmasked output. The sentences following Proposition 3 assert that discarding sensitive features 'promotes' learning non-sensitive features and 'would become more similar' to the target, but this is a heuristic statement, not a consequence of the proposition. The advertised claim that the high-generalization beamforming module 'theoretically' decreases the upper bound of the generalization error is therefore not established; the proposition as stated is also compatible with masking increasing the bound.
  3. [Section IV, Eq. (21)] The loss L_ie is called information entropy, but it is not an entropy. The entries of V3D_t,HGNet are complex-valued beamforming coefficients, not probabilities; the logarithm of a complex quantity is not the standard entropy operation, and the absolute-value weighting does not convert the expression into a meaningful entropy. No theoretical or experimental evidence is given that minimizing L_ie improves achievable sum rate. Since OAU is claimed to improve sum rate specifically through this surrogate loss, the mechanism is unvalidated. The empirical improvement may be an artifact of updating batch-normalization parameters with any gradient signal, but the paper does not test or explain this.
  4. [Section III-D, Eq. (15)] The residual identity mapping adds V_IM to C_L in AL(C_L + V_IM). The text states that V_IM is the output of the identity mapping with H3D_t,mod as input, whose third dimension is MN, while C_L has third dimension 2M by Remark 4. For arbitrary M and N, MN and 2M are not equal, so the addition is undefined unless N=2. The experiments use N=2, which hides this inconsistency, but the paper claims general MIMO cell-free systems. The architecture needs a projection for the residual path or an explicit restriction to N=2.
minor comments (4)
  1. [Section V-A] The sentence 'The learning rate and batch size are set to 64 and 0.1, respectively' appears to have the two values swapped; please clarify the learning rate and the batch size.
  2. [Section III-C] The weighted random selection mask is described as being applied during the training stage, but no description states how the mask is chosen at inference time. Specify whether a fixed mask, an averaged mask, or the expected value is used during inference; otherwise the network's output is stochastic.
  3. [Throughout] There are typographical errors, including 'inclueds' in Section III-C, 'traing' in the caption of Fig. 3, and 'Y ongming' in the author affiliation. These should be corrected.
  4. [Section V-B] The figures labeled as average generalization sum rate are plotted against 'Data stream' with 2000 points; clarify whether these are averages over 2000 test samples and over how many random channel realizations, and provide standard deviations or confidence intervals for the main comparisons.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the network is trained against external benchmarks and a standard domain-adaptation bound; the main weaknesses are an unsupported phase-free input and an unproved generalization-bound reduction, not circular reasoning.

full rationale

The derivation chain is not circular. HGNet is trained by minimizing the negative sum rate evaluated with the true complex CSI (Section III-D, 'following the commonly utilized unsupervised training method [26], we also take the negative of the sum rate as the loss function'), and its reported gains are measured against external algorithms WMMSE [7] and Edge-GNN [10], plus the SUNet baseline [11]; no fitted parameter is later renamed as a prediction. The OAU algorithm updates only BN affine parameters by minimizing the information-entropy loss (Eq. 21), which is not the sum-rate metric, so the reported sum-rate improvement is an empirical outcome rather than a construction-level identity. Proposition 3 is a standard domain-adaptation upper bound assembled from external results ([28], [32]); the text's assertion that discarding 'sensitive' features reduces the G-MMD terms is an unproved heuristic, not a self-referential definition, because the mask is chosen by discriminator scores (Eqs. 10-13) and no equation connects those scores to d_G-MMD. Self-citations [3], [11], [26] are minor and non-load-bearing: [11] is a baseline, [26] is a standard training reference, and [3] is background. Two non-circular weaknesses deserve separate weight: Section III-A's 'computing modulus values' input, if literal, discards the phase needed for coherent beamforming and threatens the central empirical claim; and Section III-E does not actually prove the claimed bound reduction. Both are correctness/rigor gaps, not circular reductions. Score 2 reflects only the minor non-load-bearing self-citations, with no circular step exhibited.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The central method depends on several unquantified or unproven choices: C_dis is never specified, the G-MMD reduction is assumed rather than shown, the entropy loss has no proven link to sum rate, and the magnitude-only input is unjustified. The paper contributes a combination of known components rather than a first-principles derivation.

free parameters (1)
  • C_dis, number of discarded sensitive features = not reported
    Eq (13) in Section III-C.2 selects C_dis features with the largest keys to mask, but the paper never provides the value or a selection rule.
assumptions (5)
  • standard math MMD-based domain adaptation bounds from [28] and [32] apply to the proposed G-MMD and the beamforming problem.
    Appendix A invokes Lemma 5.3 from [28] and Theorem 29 from [32] without proof; these are accepted results in domain adaptation theory.
  • ad hoc to paper Discarding features that are sensitive to period identity reduces distribution shift without removing sum-rate-relevant information.
    Section III-E asserts this after Proposition 3; no theorem or experiment measures the MMD terms before and after masking, so the assumed benefit is not established.
  • ad hoc to paper The information entropy of the beamforming output is a suitable surrogate for sum rate in the online update.
    Section IV introduces Eq (21) as a cheap loss, but no derivation links entropy minimization or maximization to the sum-rate objective or to beamforming quality.
  • ad hoc to paper A CNN can produce near-optimal complex beamforming from magnitude-only CSI.
    Section III-A converts CSI to modulus values, and the paper implicitly assumes this phase-free input is sufficient, without theoretical or experimental justification.
  • domain assumption Channels are constant within each period, vary between periods, and perfect global CSI is available at the CPU.
    Section II models the wireless environment this way and assumes the CPU has global CSI to design beamforming.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Online Adaptive Real-Time Beamforming Design for Dynamic Environments in Cell-Free Systems." pith.science (2026). https://pith.science/paper/OFRLP3X6

@misc{pith2026241209629,
  author       = {Pith},
  title        = {Pith review of: Online Adaptive Real-Time Beamforming Design for Dynamic Environments in Cell-Free Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OFRLP3X6}},
  note         = {Machine review of arXiv:2412.09629}
}
read the original abstract

In this paper, we consider real-time beamforming design for dynamic wireless environments with varying channels and different numbers of access points (APs) and users in cell-free systems. Specifically, a sum-rate maximization optimization problem is formulated for the beamforming design in dynamic wireless environments of cell-free systems. To efficiently solve it, a high-generalization network (HGNet) is proposed to adapt to the changing numbers of APs and users. Then, a high-generalization beamforming module is also designed in HGNet to extract the valuable features for the varying channels, and we theoretically prove that such a high-generalization beamforming module is able to reduce the upper bound of the generalization error. Subsequently, by online adaptively updating about 3% of the parameters of HGNet, an online adaptive updating (OAU) algorithm is proposed to enable the online adaptive real-time beamforming design for improving the sum rate. Numerical results demonstrate that the proposed HGNet with OAU algorithm achieves a higher sum rate with a lower computational cost on the order of milliseconds, thus realizing the real-time beamforming design for dynamic wireless environments in cell-free systems.

Figures

Figures reproduced from arXiv: 2412.09629 by the authors.

Figure 1
Figure 1. An illustration for dynamic wireless environments [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Proposed HGNet. A. Input Since deep learning methods such as CNNs normally deal with three-dimensional (3D) real numbers, the input module converts Ht ∈ C ItN×QtM into a 3D real-valued CSI tensor H3D t,mod ∈ R Qt×It×MN by computing modulus values and dimension transformations. 1 B. Convolution Unit C (·, θl) C (·, θl) aims to achieve the mapping from H3D t,mod ∈ R Qt×It×MN to beamforming with fulfilling (Qt, It) 6= … view at source ↗
Figure 3
Figure 3. Plot of training loss with the number of iterations. [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (5 more)
Figure 5
Figure 5. Figure 5: Comparative results of average computation time [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Comparative results of average generalization sum [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 9
Figure 9. Figure 9: Average computation time of OAU algorithm across [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: Average generalization sum rate of OAU algorithm [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]
Figure 11
Figure 11. Figure 11: Average computation time of OAU algorithm across [PITH_FULL_IMAGE:figures/full_fig_p011_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 31 canonical work pages

  1. [1]

    Toward 6g TK µ extreme connec- tivity: Architecture, key technologies and experiments,

    X. Y ou, Y . Huang, S. Liu et al. , “Toward 6g TK µ extreme connec- tivity: Architecture, key technologies and experiments,” IEEE Wireless Communications, vol. 30, no. 3, pp. 86–95, Jun. 2023

    work page 2023

  2. [2]

    Ris-aided cell-free massive mimo systems for 6g: F undamen- tals, system design, and applications,

    E. Shi, J. Zhang, H. Du, B. Ai, C. Y uen, D. Niyato, K. B. Leta ief, and X. Shen, “Ris-aided cell-free massive mimo systems for 6g: F undamen- tals, system design, and applications,” Proceedings of the IEEE , vol. 112, no. 4, pp. 331–364, 2024

    work page 2024

  3. [3]

    A deep learnin g method: Qos-aware joint ap clustering and beamforming desi gn for cell- free networks,

    G. Chen, S. He, Z. An, Y . Huang, and L. Y ang, “A deep learnin g method: Qos-aware joint ap clustering and beamforming desi gn for cell- free networks,” IEEE Transactions on Communications , vol. 71, no. 12, pp. 7023–7038, Dec. 2023

    work page 2023

  4. [4]

    St acked intelligent metasurfaces for holographic mimo aided cell- free networks,

    Q. Li, M. El-Hajjar, C. Xu, J. An, C. Y uen, and L. Hanzo, “St acked intelligent metasurfaces for holographic mimo aided cell- free networks,” IEEE Transactions on Communications , 2024, Early Access

    work page 2024

  5. [5]

    Uplink p recoding design for cell-free massive mimo with iteratively weighte d mmse,

    Z. Wang, J. Zhang, H. Q. Ngo, B. Ai, and M. Debbah, “Uplink p recoding design for cell-free massive mimo with iteratively weighte d mmse,” IEEE Transactions on Communications , vol. 71, no. 3, pp. 1646–1664, Mar. 2023

    work page 2023

  6. [6]

    Ener gy- efficient hybrid beamforming with dynamic on-off control fo r integrated sensing, communications, and powering,

    Z. Hao, Y . Fang, X. Y u, J. Xu, L. Qiu, L. Xu, and S. Cui, “Ener gy- efficient hybrid beamforming with dynamic on-off control fo r integrated sensing, communications, and powering,” IEEE Transactions on Com- munications, 2024, Early Access

    work page 2024

  7. [7]

    An iterative ly weighted mmse approach to distributed sum-utility maximization for a mimo interfering broadcast channel,

    Q. Shi, M. Razaviyayn, Z.-Q. Luo, and C. He, “An iterative ly weighted mmse approach to distributed sum-utility maximization for a mimo interfering broadcast channel,” IEEE Transactions on Signal Processing , vol. 59, no. 9, pp. 4331–4340, Sep. 2011

    work page 2011

  8. [8]

    Cell-free xl-mimo meets multi-agent reinforcement learning: Architectures, challenges, and future directions,

    Z. Liu, J. Zhang, Z. Liu, H. Du, Z. Wang, D. Niyato, M. Guiza ni, and B. Ai, “Cell-free xl-mimo meets multi-agent reinforcement learning: Architectures, challenges, and future directions,” IEEE Wireless Com- munications, vol. 31, no. 4, pp. 155–162, 2024

    work page 2024

Show all 32 references
  1. [9]

    Mobile cell-free massive mimo: Challenges, solu tions, and future directions,

    J. Zheng, J. Zhang, H. Du, D. Niyato, B. Ai, M. Debbah, and K . B. Letaief, “Mobile cell-free massive mimo: Challenges, solu tions, and future directions,” IEEE Wireless Communications , vol. 31, no. 3, pp. 140–147, 2024

    2024

  2. [10]

    Learning cooperati ve beam- forming with edge-update empowered graph neural networks,

    Y . Wang, Y . Li, Q. Shi, and Y .-C. Wu, “Learning cooperati ve beam- forming with edge-update empowered graph neural networks, ” in ICC 2023-IEEE International Conference on Communications . IEEE, 2023, pp. 5111–5116

    2023

  3. [11]

    An efficie nt architec- ture search for scalable beamforming design in cell-free sy stems,

    G. Chen, Z. Wang, Y . Jia, Y . Huang, and L. Y ang, “An efficie nt architec- ture search for scalable beamforming design in cell-free sy stems,” IEEE Transactions on V ehicular Technology, vol. 73, no. 7, pp. 10 241–10 253, 2024

    2024

  4. [12]

    An adaptive and robust deep learning framework for thz ultr a-massive mimo channel estimation,

    W. Y u, Y . Shen, H. He, X. Y u, S. Song, J. Zhang, and K. B. Let aief, “An adaptive and robust deep learning framework for thz ultr a-massive mimo channel estimation,” IEEE Journal of Selected Topics in Signal Processing, vol. 17, no. 4, pp. 761–776, Jul. 2023

    2023

  5. [13]

    Modeling cellular networks in fading environm ents with dominant specular components,

    A. AlAmmouri, H. ElSawy, A. Sultan-Salem, M. Di Renzo, a nd M.- S. Alouini, “Modeling cellular networks in fading environm ents with dominant specular components,” in 2016 IEEE International Conference on Communications (ICC) , 2016, pp. 1–7

    2016

  6. [14]

    Generalizing to unseen domains: A surv ey on domain generalization,

    J. Wang, C. Lan, C. Liu, Y . Ouyang, T. Qin, W. Lu, Y . Chen, W. Zeng, and P . S. Y u, “Generalizing to unseen domains: A surv ey on domain generalization,” IEEE Transactions on Knowledge and Data Engineering, vol. 35, no. 8, pp. 8052–8072, Aug. 2023

    2023

  7. [15]

    Learning t o continuously optimize wireless resource in a dynamic envir onment: A bilevel optimization perspective,

    H. Sun, W. Pu, X. Fu, T.-H. Chang, and M. Hong, “Learning t o continuously optimize wireless resource in a dynamic envir onment: A bilevel optimization perspective,” IEEE Transactions on Signal Process- ing, vol. 70, pp. 1900–1917, Jan. 2022

    1900

  8. [16]

    Meta-gating framework for fast and continuous resource optimization in dynamic wireless envi ronments,

    Q. Hou, M. Lee, G. Y u, and Y . Cai, “Meta-gating framework for fast and continuous resource optimization in dynamic wireless envi ronments,” IEEE Transactions on Communications , vol. 71, no. 9, pp. 5259–5273, Sep. 2023

    2023

  9. [17]

    Meta-critic reinforcement le arning for intelligent omnidirectional surface assisted multi-user communications,

    Q. Luo, Z. Han, and B. Di, “Meta-critic reinforcement le arning for intelligent omnidirectional surface assisted multi-user communications,” IEEE Transactions on Wireless Communications , vol. 23, no. 8, pp. 9085–9098, 2024

    2024

  10. [18]

    Beamforming des ign in short-packet transmission for urllc in cell-free massive m imo system,

    J. Fu, P . Zhu, J. Li, Y . Wang, and X. Y ou, “Beamforming des ign in short-packet transmission for urllc in cell-free massive m imo system,” IEEE Systems Journal , vol. 17, no. 3, pp. 4715–4724, Sep. 2023

    2023

  11. [19]

    Estimating traffic density on roads using convolutional neural network with batch norm alization,

    M. Hasan, S. Das, and M. N. T. Akhand, “Estimating traffic density on roads using convolutional neural network with batch norm alization,” in 2021 5th International Conference on Electrical Engineeri ng and Information Communication Technology . IEEE, 2021, pp. 1–6

    2021

  12. [20]

    V oltage o ver-scaling- based lightweight authentication for iot security,

    J. Zhang, C. Shen, H. Su, M. T. Arafin, and G. Qu, “V oltage o ver-scaling- based lightweight authentication for iot security,” IEEE Transactions on Computers, vol. 71, no. 2, pp. 323–336, Feb. 2022

    2022

  13. [21]

    Unsupervised domain adapta tion by back- propagation,

    Y . Ganin and V . Lempitsky, “Unsupervised domain adapta tion by back- propagation,” in International conference on machine learning . PMLR, 2015, pp. 1180–1189

    2015

  14. [22]

    Learning i mportant features through propagating activation differences,

    A. Shrikumar, P . Greenside, and A. Kundaje, “Learning i mportant features through propagating activation differences,” in International conference on machine learning . PMLR, 2017, pp. 3145–3153

    2017

  15. [23]

    Weighted random sa mpling with a reservoir,

    P . S. Efraimidis and P . G. Spirakis, “Weighted random sa mpling with a reservoir,” Information processing letters , vol. 97, no. 5, pp. 181–185, Mar. 2006

    2006

  16. [24]

    Deep residual learni ng for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learni ng for image recognition,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 770–778

    2016

  17. [25]

    Deep weighte d mmse downlink beamforming,

    L. Pellaco, M. Bengtsson, and J. Jald´ en, “Deep weighte d mmse downlink beamforming,” in ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2021, pp. 4915–4919

    2021

  18. [26]

    A n unsuper- vised deep unrolling framework for constrained optimizati on problems in wireless networks,

    S. He, S. Xiong, Z. An, W. Zhang, Y . Huang, and Y . Zhang, “A n unsuper- vised deep unrolling framework for constrained optimizati on problems in wireless networks,” IEEE Transactions on Wireless Communications , vol. 21, no. 10, pp. 8552–8564, 2022

    2022

  19. [27]

    Weighted quantil e discrepancy- based deep domain adaptation network for intelligent fault diagnosis,

    Z. Fan, Q. Xu, C. Jiang, and S. X. Ding, “Weighted quantil e discrepancy- based deep domain adaptation network for intelligent fault diagnosis,” Knowledge-Based Systems , vol. 240, p. 108149, Mar. 2022

    2022

  20. [28]

    Redko, E

    I. Redko, E. Morvant, A. Habrard, M. Sebban, and Y . Benna ni, Advances in domain adaptation theory . Elsevier, 2019

    2019

  21. [29]

    Adversarial en tropy optimization for unsupervised domain adaptation,

    A. Ma, J. Li, K. Lu, L. Zhu, and H. T. Shen, “Adversarial en tropy optimization for unsupervised domain adaptation,” IEEE Transactions on Neural Networks and Learning Systems , vol. 33, no. 11, pp. 6263– 6274, Nov. 2022

    2022

  22. [30]

    Toward ma ssive mimo 2.0: Understanding spatial correlation, interference suppres sion, and pilot 13 contamination,

    L. Sanguinetti, E. Bj¨ ornson, and J. Hoydis, “Toward ma ssive mimo 2.0: Understanding spatial correlation, interference suppres sion, and pilot 13 contamination,” IEEE Transactions on Communications , vol. 68, no. 1, pp. 232–257, Jan. 2019

    2019

  23. [31]

    Ris-assisted communica tion radar coexistence: Joint beamforming design and analysis,

    Y . He, Y . Cai, H. Mao, and G. Y u, “Ris-assisted communica tion radar coexistence: Joint beamforming design and analysis,” IEEE Journal on Selected Areas in Communications , vol. 40, no. 7, pp. 2131–2145, Jul. 2022

    2022

  24. [32]

    Learning via hilbert space embedding of distr ibutions,

    L. Song, “Learning via hilbert space embedding of distr ibutions,” University of Sydney (2008) , vol. 17, 2008

    2008

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.