pith. sign in

arxiv: 2604.21271 · v2 · submitted 2026-04-23 · 💻 cs.IT · math.IT

Downlink Channel Matrix Estimation from PMI-Only Feedback in FDD Systems: Maximum Likelihood and Sharp Excess Risk Bound

Jinchi Chen , Mingxi Hu , Peigang Jiang , Xin Meng , Ke Wei , Xianyin Zhang This is my paper

Pith reviewed 2026-05-08 14:10 UTC · model grok-4.3

classification 💻 cs.IT math.IT
keywords FDD massive MIMOPMI feedbackchannel matrix estimationmaximum likelihood estimatorCramér-Rao boundexcess risk boundlimited feedback
0
0 comments X

The pith

A constrained maximum likelihood estimator can recover the downlink channel matrix from PMI-only feedback with statistical guarantees in FDD massive MIMO systems.

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

The paper proposes a constrained maximum likelihood estimator to recover the downlink channel matrix from PMI-only feedback under a probabilistic perturbation model in FDD massive MIMO systems. This estimator relaxes the hard empirical decision error and comes with a derived Cramér-Rao bound that accounts for phase ambiguity. A global excess-risk bound of order O(1/√T) is proven for the real-valued setting and refines to O(1/T) under identifiability. Numerical results confirm asymptotic attainment of the bound and superiority over baselines on both synthetic data and realistic FDD channels.

Core claim

Under a probabilistic perturbation model of PMI observations, the constrained maximum likelihood estimator for the channel matrix achieves consistency and an excess risk bound of O(1/√T) globally in the real-valued setting, sharpening to O(1/T) locally with identifiability, while attaining the Cramér-Rao bound asymptotically in the complex-valued case after gauge-fixing the phase ambiguity.

What carries the argument

Constrained maximum likelihood estimator based on the probabilistic perturbation model for PMI selection around the true reduced-dimensional channel.

If this is right

  • The MLE can be deployed in existing 5G NR limited-feedback frameworks to enhance channel estimation accuracy.
  • As the number of PMI reports T increases, the estimation error decreases at the predicted rates.
  • The approach outperforms existing methods on realistic channel models, suggesting broad applicability.
  • Under suitable conditions, the faster O(1/T) rate supports improved performance in time-varying scenarios.

Where Pith is reading between the lines

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

  • This method could be extended to joint estimation with other feedback types if the model is adapted.
  • The identifiability conditions might be characterized more explicitly to guide codebook design.
  • If validated in hardware, it could lower pilot overhead in FDD systems by relying more on feedback.
  • Connections to quantized estimation in other signal processing domains may yield similar risk bounds.

Load-bearing premise

The observations of PMI are generated according to a probabilistic perturbation model centered on the true reduced-dimensional channel matrix.

What would settle it

Generating synthetic PMI data from the assumed probabilistic model and verifying whether the empirical excess risk scales exactly as O(1/√T) with increasing T; significant deviation would falsify the bound.

Figures

Figures reproduced from arXiv: 2604.21271 by Jinchi Chen, Ke Wei, Mingxi Hu, Peigang Jiang, Xianyin Zhang, Xin Meng.

Figure 1
Figure 1. Figure 1: Comparison between the phase-aligned MSE of the MLE and the trace Cram´er–Rao bound versus view at source ↗
Figure 2
Figure 2. Figure 2: Beam precision versus communication rounds in the single-stream case ( view at source ↗
Figure 3
Figure 3. Figure 3: Beam precision versus communication rounds in the two-stream case ( view at source ↗
Figure 4
Figure 4. Figure 4: Ablation study on the temperature parameter view at source ↗
Figure 5
Figure 5. Figure 5: Ablation study on initialization under the structured dimensionality-reduction design (4.4). We view at source ↗
read the original abstract

We study downlink channel estimation in a frequency-division duplex (FDD) massive MIMO system from PMI-only feedback under a 5G NR-type limited-feedback architecture. In this architecture, the user selects a preferred codeword from a shared codebook based on the reduced-dimensional channel and only reports its index (known as the precoding matrix indicator, PMI) back to the base station. Therefore, the channel must be estimated from these highly quantized, nonlinear PMI observations. Based on a probabilistic perturbation model, a constrained maximum likelihood estimator (MLE) is proposed for this estimation problem, whose objective can also be interpreted as a relaxation of the hard empirical decision error. The Cram\'er--Rao bound is derived for the complex-valued model, with the global phase ambiguity handled via gauge-fixing. For the real-valued setting, a global excess-risk bound of order $O(1/\sqrt{T})$ is established, which is then refined to a sharp local rate of order $O(1/T)$ under suitable identifiability conditions. Numerical results show that the MLE asymptotically attains the Cram\'er--Rao bound and outperforms several baseline methods on both synthetic data and realistic FDD channels.

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 / 2 minor

Summary. The manuscript proposes a constrained maximum likelihood estimator (MLE) for downlink channel matrix estimation in frequency-division duplex (FDD) massive MIMO systems based on precoding matrix indicator (PMI) feedback only. Using a probabilistic perturbation model around the true reduced-dimensional channel, the paper derives the Cramér-Rao bound (CRB) for the complex-valued model with gauge-fixing for phase ambiguity, establishes a global excess-risk bound of order O(1/√T) for the real-valued setting which is refined to O(1/T) under identifiability conditions, and shows numerically that the MLE attains the CRB and outperforms baselines on synthetic data and realistic FDD channels.

Significance. This work offers a principled statistical approach to a practical problem in wireless communications where full channel state information is unavailable due to limited feedback in FDD systems. The theoretical contributions on excess risk bounds and CRB, combined with numerical evidence of asymptotic efficiency and superior performance on realistic channels, could inform the design of more efficient limited-feedback schemes. The explicit model and bounds provide falsifiable predictions that are partially validated numerically.

major comments (1)
  1. The refinement of the excess-risk bound to O(1/T) under suitable identifiability conditions is central to the sharp rate claim; however, the specific identifiability conditions and their verification for the codebook and channel distributions used should be stated more explicitly to ensure the local rate applies in the considered scenarios.
minor comments (2)
  1. The figures showing CRB attainment and performance comparison would benefit from error bars or multiple Monte Carlo runs to indicate variability.
  2. Clarify the exact form of the probabilistic perturbation model for PMI observations, including any assumptions on the noise distribution.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive comment and positive overall assessment. We address the point below and will incorporate the requested clarification in the revised manuscript.

read point-by-point responses

  1. Referee: The refinement of the excess-risk bound to O(1/T) under suitable identifiability conditions is central to the sharp rate claim; however, the specific identifiability conditions and their verification for the codebook and channel distributions used should be stated more explicitly to ensure the local rate applies in the considered scenarios.

    Authors: We agree that greater explicitness will strengthen the presentation. The identifiability conditions underlying the local O(1/T) excess-risk bound appear in Assumption 3 and are invoked in the statement and proof of Theorem 5 (Section IV). These conditions require that the codebook forms a tight frame for the reduced-dimensional channel subspace and that the minimum distance between distinct codewords exceeds twice the standard deviation of the perturbation model. In the numerical experiments of Sections V and VI we employ a DFT-based codebook together with both synthetic Gaussian channels and 3GPP-compliant FDD channels; under these choices the conditions hold, which is consistent with the observed asymptotic attainment of the CRB. To address the referee’s concern directly, we will add a short dedicated paragraph immediately following Theorem 5 that (i) restates the precise conditions in the notation of the paper and (ii) verifies their satisfaction for the specific codebook and channel distributions used in the simulations. This addition will make the applicability of the sharp local rate fully transparent without altering any technical claims. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained from explicit model

full rationale

The central claims rest on an explicitly stated probabilistic perturbation model for PMI observations, from which the constrained MLE objective, complex CRB (with gauge fixing), and real-valued excess-risk bounds (global O(1/√T), local O(1/T) under identifiability) are derived directly via standard statistical arguments. No step reduces a prediction or bound to a fitted quantity by construction, nor relies on a load-bearing self-citation chain or imported uniqueness theorem. Numerical validation on synthetic and realistic channels is separate from the analytic derivations and does not feed back into them. The model assumptions are granted as the starting point; the paper does not claim to derive the model itself.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claims rest on a probabilistic perturbation model for PMI selection and on identifiability conditions that are invoked but not fully detailed in the abstract.

axioms (2)
  • domain assumption PMI feedback is generated by a probabilistic perturbation model around the true reduced-dimensional channel
    Stated in the abstract as the basis for the MLE and all subsequent bounds.
  • domain assumption Suitable identifiability conditions hold for the local O(1/T) rate
    Invoked to refine the global excess-risk bound; location not specified in abstract.

pith-pipeline@v0.9.0 · 5530 in / 1335 out tokens · 31726 ms · 2026-05-08T14:10:40.671351+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

48 extracted references · 48 canonical work pages

  1. [1]

    Release 16

    3rd Generation Partnership Project (3GPP) , 5G; NR; Physical layer procedures for data(3GPP TS 38.214 version 16.2.0 Release 16) , Technical Specification V16.2.0 , 3GPP TS 38.214 , July 2020. Release 16

    work page 2020

  2. [2]

    height 2pt depth -1.6pt width 23pt, Study on Channel Model for Frequencies from 0.5 to 100 GHz (Release 19) , Technical Report v19.0.0, 3GPP, TR 38.901 , June 2025

    work page 2025

  3. [3]

    Almradi, M

    A. Almradi, M. Matthaiou, P. Xiao, and V. F. Fusco , Hybrid precoding for massive mimo with low rank channels: A two-stage user scheduling approach , IEEE Transactions on Communications, 68 (2020), pp. 4816--4831

    work page 2020

  4. [4]

    Bach , Learning Theory from First Principles , MIT Press, Cambridge, MA, 2024

    F. Bach , Learning Theory from First Principles , MIT Press, Cambridge, MA, 2024

    work page 2024

  5. [5]

    Ben-Haim and Y

    Z. Ben-Haim and Y. C. Eldar , On the constrained cram \'e r--rao bound with a singular fisher information matrix , IEEE Signal Processing Letters, 16 (2009), pp. 453--456

    work page 2009

  6. [6]

    C. R. Berger, Z. Wang, J. Huang, and S. Zhou , Application of compressive sensing to sparse channel estimation , IEEE Communications Magazine, 48 (2010), pp. 164--174

    work page 2010

  7. [7]

    Boloursaz Mashhadi and D

    M. Boloursaz Mashhadi and D. G \"u nd \"u z , Deep learning for massive mimo channel state acquisition and feedback , Journal of the Indian Institute of Science, 100 (2020), pp. 369--382

    work page 2020

  8. [8]

    E. J. Cand \`e s, X. Li, and M. Soltanolkotabi , Phase retrieval via wirtinger flow: Theory and algorithms , IEEE Transactions on Information Theory, 61 (2015), pp. 1985--2007

    work page 2015

  9. [9]

    T. Chen, J. Guo, S. Jin, C.-K. Wen, and G. Y. Li , A novel quantization method for deep learning-based massive mimo csi feedback , in 2019 IEEE Global Conference on Signal and Information Processing (GlobalSIP), IEEE, 2019, pp. 1--5

    work page 2019

  10. [10]

    J. Dai, A. Liu, and V. K. Lau , Joint channel estimation and user grouping for massive mimo systems , IEEE Transactions on Signal Processing, 67 (2018), pp. 622--637

    work page 2018

  11. [11]

    Do and A

    H. Do and A. Lozano , Cramer-rao bound for arbitrarily constrained sets , arXiv preprint arXiv:2601.19539, (2026)

    work page arXiv 2026

  12. [12]

    R. M. Dreifuerst and R. W. Heath , Neural codebook design for MIMO network beam management , IEEE Transactions on Wireless Communications, 24 (2025), pp. 3909--3922

    work page 2025

  13. [13]

    K.-T. Fang, S. Kotz, and K. W. Ng , Symmetric Multivariate and Related Distributions , vol. 36 of Monographs on Statistics and Applied Probability, Chapman and Hall, London, 1990

    work page 1990

  14. [14]

    Fannjiang and T

    A. Fannjiang and T. Strohmer , The numerics of phase retrieval , Acta Numerica, 29 (2020), pp. 125--228

    work page 2020

  15. [15]

    Z. Gao, L. Dai, Z. Wang, and S. Chen , Spatially common sparsity based adaptive channel estimation and feedback for FDD massive MIMO , IEEE Transactions on Signal Processing, 63 (2015), pp. 6169--6183

    work page 2015

  16. [16]

    Goldsmith , Wireless Communications , Cambridge University Press, 2005

    A. Goldsmith , Wireless Communications , Cambridge University Press, 2005

    work page 2005

  17. [17]

    E. J. Gumbel , Statistical theory of extreme values and some practical applications , NBS Applied Mathematics Series, 33 (1954), pp. 1--51

    work page 1954

  18. [18]

    Guo, C.-K

    J. Guo, C.-K. Wen, S. Jin, and G. Y. Li , Convolutional neural network-based multiple-rate compressive sensing for massive mimo csi feedback: Design, simulation, and analysis , IEEE Transactions on Wireless Communications, 19 (2020), pp. 2827--2840

    work page 2020

  19. [19]

    Huang, H.-M

    K.-W. Huang, H.-M. Wang, J. Hou, and S. Jin , Joint spatial division and diversity for massive mimo systems , IEEE Transactions on Communications, 67 (2018), pp. 258--272

    work page 2018

  20. [20]

    Jaeckel, L

    S. Jaeckel, L. Raschkowski, K. Borner, and L. Thiele , Quadriga: A 3-d multi-cell channel model with time evolution for enabling virtual field trials , IEEE Transactions on Antennas and Propagation, 62 (2014), pp. 3242--3256

    work page 2014

  21. [21]

    A. H. Joarder, W. S. Al-Sabah, and M. H. Omar , On the distributions of norms of spherical distributions , Journal of Probability and Statistical Science, 6 (2008), pp. 115--123

    work page 2008

  22. [22]

    M. B. Khalilsarai, S. Haghighatshoar, X. Yi, and G. Caire , FDD massive MIMO via UL/DL channel covariance extrapolation and active channel sparsification , IEEE Transactions on Wireless Communications, 18 (2019), pp. 121--135

    work page 2019

  23. [23]

    P.-H. Kuo, H. Kung, and P.-A. Ting , Compressive sensing based channel feedback protocols for spatially-correlated massive antenna arrays , in 2012 IEEE Wireless Communications and Networking Conference (WCNC), IEEE, 2012, pp. 492--497

    work page 2012

  24. [24]

    K. Li, Y. Li, L. Cheng, and Z.-Q. Luo , Enhancing multi-stream beamforming through cqis for 5g nr fdd massive mimo communications: A tuning-free scheme , IEEE Transactions on Wireless Communications, (2024)

    work page 2024

  25. [25]

    K. Li, Y. Li, L. Cheng, Q. Shi, and Z.-Q. Luo , Downlink channel covariance matrix reconstruction for FDD massive MIMO systems with limited feedback , IEEE Transactions on Signal Processing, 72 (2024), pp. 1032--1048

    work page 2024

  26. [26]

    K. Li, W. Pu, and Z.-Q. Luo , Efficient beamforming refinement for limited feedback FDD massive MIMO : An online alternating exploration-estimation approach , IEEE Transactions on Signal Processing, 73 (2025), pp. 4969--4985

    work page 2025

  27. [27]

    L. Li, X. Zeng, Y.-F. Liu, Y. Xu, and T.-H. Chang , Csi sensing from heterogeneous user feedbacks: A constrained phase retrieval approach , IEEE Transactions on Wireless Communications, 22 (2023), pp. 6930--6945

    work page 2023

  28. [28]

    Liu and V

    A. Liu and V. K. Lau , Two-stage subspace constrained precoding in massive mimo cellular systems , IEEE Transactions on Wireless Communications, 14 (2015), pp. 3271--3279

    work page 2015

  29. [29]

    Z. Liu, L. Zhang, and Z. Ding , An efficient deep learning framework for low rate massive mimo csi reporting , IEEE Transactions on Communications, 68 (2020), pp. 4761--4772

    work page 2020

  30. [30]

    Z. Lu, J. Wang, and J. Song , Multi-resolution csi feedback with deep learning in massive mimo system , in ICC 2020-2020 IEEE international conference on communications (ICC), IEEE, 2020, pp. 1--6

    work page 2020

  31. [31]

    M. B. Mashhadi, Q. Yang, and D. G \"u nd \"u z , Cnn-based analog csi feedback in fdd mimo-ofdm systems , in ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2020, pp. 8579--8583

    work page 2020

  32. [32]

    S. L. H. Nguyen and A. Ghrayeb , Compressive sensing-based channel estimation for massive multiuser mimo systems , in 2013 IEEE Wireless Communications and Networking Conference (WCNC), IEEE, 2013, pp. 2890--2895

    work page 2013

  33. [33]

    B. Ning, H. Yin, S. Liu, H. Deng, S. Yang, Y. Zhang, W. Mei, D. Gesbert, J. Park, R. W. Heath, et al. , Precoding matrix indicator in the 5g nr protocol: A tutorial on 3gpp beamforming codebooks , IEEE Communications Surveys & Tutorials, (2026)

    work page 2026

  34. [34]

    X. Ning, S. Zhang, Y. Xue, X. Zheng, Q. Shi, and T.-H. Chang , Learning beams adaptive to the environment: An RSRP -based codebook design , in 2023 IEEE 24th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), 2023, pp. 521--525

    work page 2023

  35. [35]

    Rao and V

    X. Rao and V. K. Lau , Distributed compressive csit estimation and feedback for fdd multi-user massive mimo systems , IEEE Transactions on Signal Processing, 62 (2014), pp. 3261--3271

    work page 2014

  36. [36]

    S. T. Smith , Covariance, subspace, and intrinsic cram \'e r--rao bounds , IEEE Transactions on Signal Processing, 53 (2005), pp. 1610--1630

    work page 2005

  37. [37]

    Stoica and B

    P. Stoica and B. C. Ng , On the cram \'e r-rao bound under parametric constraints , IEEE Signal Processing Letters, 5 (1998), pp. 177--179

    work page 1998

  38. [38]

    J. A. Tropp , An introduction to matrix concentration inequalities , Foundations and trends in machine learning, 8 (2015), pp. 1--230

    work page 2015

  39. [39]

    Tse and P

    D. Tse and P. Viswanath , Fundamentals of Wireless Communication , Cambridge University Press, 2005

    work page 2005

  40. [40]

    R. Vershynin , High-Dimensional Probability: An Introduction with Applications in Data Science , Cambridge Series in Statistical and Probabilistic Mathematics, Cambridge University Press, 2018

    work page 2018

  41. [41]

    G. Wang, G. B. Giannakis, and Y. C. Eldar , Solving systems of random quadratic equations via truncated amplitude flow , IEEE Transactions on Information Theory, 64 (2018), pp. 773--794

    work page 2018

  42. [42]

    Wen, W.-T

    C.-K. Wen, W.-T. Shih, and S. Jin , Deep learning for massive MIMO CSI feedback , IEEE Wireless Communications Letters, 7 (2018), pp. 748--751

    work page 2018

  43. [43]

    H. Xie, F. Gao, S. Jin, J. Fang, and Y.-C. Liang , Channel estimation for TDD/FDD massive MIMO systems with channel covariance computing , IEEE Transactions on Wireless Communications, 17 (2018), pp. 4206--4218

    work page 2018

  44. [44]

    Y. Xu, G. Yue, N. Prasad, S. Rangarajan, and S. Mao , User grouping and scheduling for large scale mimo systems with two-stage precoding , in 2014 IEEE International Conference on Communications (ICC), IEEE, 2014, pp. 5197--5202

    work page 2014

  45. [45]

    F. Yang, Q. Zhu, B. Xu, and C. Que , Wave beam forming method, signal emission equipment and signal reception equipment , Sept. 2016

    work page 2016

  46. [46]

    Yin and D

    H. Yin and D. Gesbert , A partial channel reciprocity-based codebook for wideband fdd massive mimo , IEEE Transactions on Wireless Communications, 21 (2022), pp. 7696--7710

    work page 2022

  47. [47]

    Zhong, L

    Z. Zhong, L. Fan, and S. Ge , Fdd massive mimo uplink and downlink channel reciprocity properties: Full or partial reciprocity? , in GLOBECOM 2020-2020 IEEE Global Communications Conference, IEEE, 2020, pp. 1--5

    work page 2020

  48. [48]

    Ziao and Y

    Q. Ziao and Y. Haifan , A review of codebooks for csi feedback in 5g new radio and beyond , China Communications, 22 (2025), pp. 112--127

    work page 2025