Pith. sign in

REVIEW 4 major objections 5 minor 32 references

Measuring Robustness of Speech Recognition from MEG Signals Under Distribution Shift

T0 review · 4 major / 5 minor · reviewed 2026-07-13 · grok-4.5

Challenge this review Download PDF Read on arXiv ↗

Pith's one-line read Joint market-and-wireless design lets prosumers share energy with differential privacy while still converging near the Nash equilibrium.

desk verdict Solid co-design of DP energy-sharing GNE with OTA MIMO; the attack model and channel-noise privacy credit are the real additions, under the usual static-CSI idealizations. read the letter →

arxiv 2604.04129 v1 submitted 2026-04-05 cs.SD cs.LG

Sheng-You Chien , Bo-Yi Mao , Yi-Ning Chang , Po-Chih Kuo This is my paper · ORCID
classification cs.SDcs.LG
keywords energysharingdifferentialprivacyover-the-aircomputationNashgameprosumerMIMOmechanism-communicationco-design
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

Prosumers who trade surplus electricity form a game that needs iterative bid exchange, but that exchange runs over wireless channels that can leak private demand and cost parameters. This paper co-designs the market clearing rule with over-the-air MIMO aggregation so that inherent channel noise and multi-user interference do part of the privacy work. Prosumers add only the remaining calibrated artificial noise needed for (ε, δ)-differential privacy, and the platform still recovers a price that converges in expectation to a neighborhood of the generalized Nash equilibrium. Simulations show the co-design can cut the artificial-noise burden by roughly half relative to ideal-channel methods, and over-the-air aggregation can be stronger for privacy than orthogonal access when many users share the same antennas.

What carries the argument

Algorithm 1: differentially private equilibrium seeking over OTA MIMO, with privacy bound (33) that uses both artificial noise ratio α and the adversary’s maximum SINR, and Theorem 1’s contraction of expected squared price error under a market-sensitivity condition on a.

What would settle it

Run the algorithm on time-varying channels with imperfect CSI or with bids that violate the stated bound L and check whether the measured privacy budget still meets the claimed (ε, δ) while the price still converges as predicted by Theorem 1.

Watch

Extended reading notes

Core claim

When energy-sharing bids are sent by over-the-air MIMO aggregation with calibrated artificial noise, the protocol can be made (ε_i, δ_i)-differentially private per prosumer while the price iterates satisfy a mean-square bound that contracts toward equilibrium plus a noise floor set by privacy and channel noise; under quadratic costs and utilities the price is asymptotically unbiased in mean.

Load-bearing premise

Uplink channels stay fixed for the whole run with perfect channel knowledge at the base station, downlink prices arrive error-free, and every bid stays inside a known bound used both for power scaling and privacy sensitivity.

Editorial extensions

If this is right

  • Artificial noise can be set lower than ideal-channel DP designs because channel noise and multi-user interference already mask individual bids.
  • OTA aggregation yields stronger privacy than TDMA/FDMA when the number of prosumers is large relative to receive antennas.
  • Mean trajectories of production, demand, and clearing price still approach the same equilibrium under quadratic models even as privacy noise grows.
  • Realistic 3GPP-style channels increase variance but do not destroy average convergence to the equilibrium price.

Reading between the lines

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

  • The same co-design pattern could apply to other iterative market or consensus algorithms that currently inject noise as if the channel were perfect.
  • If channels vary fast, adaptive recalibration of α would be needed; that extension is left open and would test how fragile the fixed-CSI privacy bound is.
  • Colluding prosumers or active jammers would change the adversary model and might erase the OTA privacy gain the paper relies on.
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 / 5 minor

Summary. The paper studies prosumer energy sharing formulated as a generalized Nash game, implemented over wireless OTA MIMO uplink aggregation with an assumed error-free downlink price broadcast. It introduces an honest-but-curious base-station adversary that uses MMSE/SINR-optimal spatial extractors and gain normalization to recover individual bids and private net demand q_i = d_i − p_i from the superimposed observation. To mitigate leakage, prosumers inject calibrated complex Gaussian artificial noise under a common noise-to-signal ratio α, yielding a per-prosumer (ε_i, δ_i)-DP guarantee that exploits both artificial and channel noise (bound (33)). Algorithm 1 iterates local best responses and OTA price updates; Theorem 1 gives an MSE contraction to a noise-dependent neighborhood of the equilibrium price under strong convexity/concavity, Lipschitz gradients, and a market-sensitivity condition on a, with asymptotic mean-unbiasedness in the quadratic case. Simulations (privacy dispersion vs α, α savings vs ideal channels, convergence CIs, OTA vs orthogonal privacy budgets, Sionna TR 38.901 check) support the analysis for small I.

Significance. The joint treatment of strategic energy-sharing equilibrium seeking, OTA MIMO aggregation, and differential privacy is a genuine and timely co-design contribution. Prior energy-sharing GNE work largely assumes ideal communication; prior OTA-DP work largely addresses federated learning rather than coupled strategic games. Exploiting inherent channel noise and multi-user interference to reduce artificial noise (Fig. 4, Table II) is practically meaningful and cleanly quantified. Appendices A–B give standard but carefully specialized MMSE-SINR and contraction arguments; the MSE recursion and quadratic unbiasedness case are coherent under the stated assumptions. Strengths include an explicit adversarial model, closed-form privacy calibration via bisection on α, and a realistic-channel sanity check with Sionna. Limitations of scale (primarily I = 3) and idealized CSI do not erase the conceptual advance if those assumptions are clearly scoped.

major comments (4)
  1. Section III (paragraph after (11)) and the privacy/convergence analyses assume uplink channels constant over the entire run with perfect CSI at the BS, plus error-free downlink. Privacy bound (33), the adversary extractors (18)/(A.3), sensitivity (30), and the e_align/e_noise split (35)–(37) all depend on this. The paper defers dynamics and imperfect CSI to future work, but the central claims of calibrated (ε,δ)-DP and the Theorem 1 MSE bound are load-bearing on this idealization. At minimum, the manuscript needs a sensitivity study (or first-order analysis) of privacy budget and MSE under CSI error / block fading, or a clearly stated scope restriction that the guarantees hold only under static perfect CSI.
  2. Encoding (8)–(10), power split (27), and the claim that e_align is “design-removable via proper pre-equalization” (after (37)) are not jointly reconciled under heterogeneous channels. Perfect alignment requires f_0^H h_i s_{i,1}/(√η L) = 1 for all i simultaneously while |s_{i,1}|^2 + |s_{i,2}|^2 ≤ P and |s_{i,2}|^2 = α|s_{i,1}|^2. Weak-channel prosumers may be unable to meet both alignment and the common-α DP allocation. Algorithm 1 and Appendix A treat α as the sole free knob after “perfect OTA pre-equalization,” but feasibility of that pre-equalization under (27) is not proven. Either give an explicit feasible pre-equalization construction (or a residual e_align bound when it is imperfect) or weaken the convergence statement to include a non-vanishing alignment bias.
  3. Sensitivity Δ_i ≤ 2|f_0^H h_i||s_{i,1}| in (30) and the encoding scale L in (7) assume a known uniform bid bound that is never violated. If L is chosen conservatively large, artificial noise is over-provisioned and convergence degrades; if bids can exceed L, both the power constraint and the Gaussian-mechanism sensitivity are invalid. The manuscript should state how L is set in practice (e.g., from known demand/cost bounds) and quantify privacy/MSE degradation under bound misspecification, or replace the hard bound with a high-probability sensitivity argument.
  4. Empirical support is thin for the scalability narrative. Core privacy and convergence figures use I = 3; the a-condition check uses I = 10; Table II’s “overloaded” case reaches I = 12 with N_r = 8. Claims that OTA’s privacy gain grows with loading ratio I/N_r and that the scheme is suitable for “scalable prosumer coordination” need either larger-I Monte Carlo evidence or a clearer caveat that results are demonstrated only at small community size.
minor comments (5)
  1. Notation: λ is used both for clearing price and (in places) as a dual variable ζ; the switch λ★ = ζ★ is correct but easy to miss on first read. A short notation table would help.
  2. Fig. 1 and Fig. 2 are useful; axis labels and α legends in Fig. 3–5 are dense. Consider larger fonts and explicit units on inferred (d−p).
  3. Related-work coverage of DP Nash seeking and OTA federated learning is adequate; a one-sentence contrast with secure aggregation / homomorphic schemes already mentioned in the introduction would round out the privacy landscape.
  4. Typos/style: “overlook the constraints” (I-A), occasional missing spaces before citations, and “prosumeriself-produces” spacing artifacts in the compiled text should be cleaned.
  5. Reproducibility: coefficients in Table I and SNR/N_r grids are given, but random seeds, exact combiner η choice, and code availability are not stated. A short reproducibility note would strengthen the experimental claims.

Circularity Check

1 steps flagged · score 1.0 of 10

No significant circularity: privacy and MSE claims follow from standard Gaussian-mechanism and contraction arguments applied to an explicitly modeled OTA game; self-citations supply background GNE setup, not the target results.

  1. self citation load bearing [Section II, after Definition 1; Eq. (4)–(5)]
    "It has been shown in [14, Proposition 1] that a GNE of G=(I,S,U) exists. Moreover, for any GNE (p⋆,d⋆,b⋆) with clearing price λ⋆, the pair (p⋆,d⋆) is the unique optimal solution to [problem (4)] ... and the GNE price satisfies λ⋆=ζ⋆ ... Consequently, the production and demand (p⋆_i,d⋆_i) of each prosumer i maximize the surrogate payoff U_i(p_i,d_i;λ) ... [Eq. (5)]."

    Existence of the GNE and the reduction to the surrogate local problem (5) rest on a prior paper by an overlapping author rather than being re-derived here. This is only background for the game formulation; the paper’s central privacy bound (33) and Theorem 1 do not reduce to [14] by construction and remain independent once the game is fixed. Hence minor, non-load-bearing self-citation.

full rationale

The load-bearing claims are the per-prosumer (ε_i, δ_i)-DP bound (33) and the MSE recursion of Theorem 1. Privacy is obtained from the classical Gaussian mechanism [22] with sensitivity upper-bounded by the known bid range L via (30); the only novelty is folding inherent channel noise and multi-user interference into the denominator of (32)–(33), which is a direct algebraic consequence of the OTA observation model (28) and is not defined to equal any target privacy level. Convergence is proved in Appendix B by showing that the ideal price map T is a contraction under Assumptions 1–2 and condition (38), then adding zero-mean noise e_noise; the recursion (B.7)–(B.8) is a standard noisy fixed-point argument and does not force E[|λ^K−λ★|^2] to any pre-chosen value. Equilibrium existence and the surrogate payoff (5) are imported from [14] (overlapping author Yue Chen), but that citation only sets the game; it is not used to derive the privacy budget or the MSE bound, and those bounds remain falsifiable by Monte-Carlo under the stated channel and power assumptions. No fitted parameter is re-labeled a prediction, no uniqueness theorem is smuggled, and no quantity equals its input by construction. Residual self-citation is background only, hence score 1.

Assumptions & free parameters 6 free parameters · 5 assumptions · 2 invented entities

The central privacy and convergence claims rest on standard convex-game and Gaussian-DP machinery plus several communication idealizations and design choices (uniform α, bid bound L, perfect CSI, fixed channels, error-free downlink). Free parameters are mostly experimental knobs and the privacy/power allocation ratio, not hidden fits that define the theorem statements. No new physical entity is postulated; the 'invented' pieces are modeling constructs (adversary pipeline, DP-OTA price estimator).

free parameters (6)
  • noise-to-signal ratio α (uniform across prosumers)
    Primary privacy control knob; α_min chosen by bisection to meet target ε; also sweeps 0–0.8 in experiments.
  • market sensitivity a
    Fixed a=100 in main sims; must satisfy inequality (38) for contraction; treated as design parameter, not learned from data.
  • bid magnitude bound L
    Used for encoding, power constraint, and sensitivity upper bound Δ_i ≤ 2|f^H h_i||s_{i,1}|; assumed known and tight enough.
  • privacy targets (ε_i, δ) and horizon K
    δ fixed at 1e-5; ε targets and K enter the Gaussian-mechanism noise requirement and thus α calibration.
  • transmit power P, SNR, N_r, combiner/normalization (f_0, η)
    Communication design parameters that change SINR★, privacy bound, and e_noise variance; set by experiment rather than derived.
  • quadratic cost/utility coefficients {c1,i,c2,i,v1,i,v2,i}
    Table I hand-chosen instance for case study; theorems are more general but unbiasedness claim specializes to quadratic.
assumptions (5)
  • standard math Gaussian mechanism / standard (ε,δ)-DP composition over K rounds for Gaussian noise on bounded-sensitivity queries
    Invoked for inequality (29)–(33) via [22]-style sensitivity arguments.
  • domain assumption Utility strongly concave, cost strongly convex, gradients Lipschitz (Assumptions 1–2); GNE exists as in prior energy-sharing analysis
    Needed for unique local maximizers, γ bounds, and contraction of T(λ) in Theorem 1 / Appendix B.
  • domain assumption Perfect CSI, time-invariant uplink channels during the algorithm, synchronized OTA symbols, error-free downlink broadcast
    Section III communication model; underpins both aggregation formula and adversary extractor.
  • ad hoc to paper Honest-but-curious BS adversary using MMSE/SINR-optimal spatial extractors then gain normalization to recover q_i
    Attack model in III-D motivates DP; privacy is measured against this pipeline, not arbitrary side information.
  • ad hoc to paper Uniform noise-to-signal ratio α across prosumers to decouple multi-user privacy constraints
    Section IV-C simplification so a single α controls all ε_i bounds via SINR★_i(α).
invented entities (2)
  • DP-OTA clearing-price estimator λ̂^k_DP with alignment and noise error split e_align + e_noise
    purpose: Links wireless receive model to the noisy price map used in convergence analysis
    Modeling construct, not a new physical object; independent evidence is only simulation consistency.
  • Per-prosumer MIMO isolation adversary (f_adv_i / f★_i) mapping OTA y^k to private net demand q_i
    purpose: Makes privacy risk concrete for energy sharing over MIMO OTA
    Composes known MMSE beamforming with market relation (3b); falsifiable only as an attack success metric in their sims.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Measuring Robustness of Speech Recognition from MEG Signals Under Distribution Shift." pith.science (2026). https://pith.science/paper/2604.04129

@misc{pith2026260404129,
  author       = {Pith},
  title        = {Pith review of: Measuring Robustness of Speech Recognition from MEG Signals Under Distribution Shift},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2604.04129}},
  note         = {Machine review of arXiv:2604.04129}
}
read the original abstract

This study investigates robust speech-related decoding from non-invasive MEG signals using the LibriBrain phoneme-classification benchmark from the 2025 PNPL competition. We compare residual convolutional neural networks (CNNs), an STFT-based CNN, and a CNN--Transformer hybrid, while also examining the effects of group averaging, label balancing, repeated grouping, normalization strategies, and data augmentation. Across our in-house implementations, preprocessing and data-configuration choices matter more than additional architectural complexity, among which instance normalization emerges as the most influential modification for generalization. The strongest of our own models, a CNN with group averaging, label balancing, repeated grouping, and instance normalization, achieves 60.95% F1-macro on the test split, compared with 39.53% for the plain CNN baseline. However, most of our models, without instance normalization, show substantial validation-to-test degradation, indicating that distribution shift induced by different normalization statistics is a major obstacle to generalization in our experiments. By contrast, MEGConformer maintains 64.09% F1-macro on both validation and test, and saliency-map analysis is qualitatively consistent with this contrast: weaker models exhibit more concentrated or repetitive phoneme-sensitive patterns across splits, whereas MEGConformer appears more distributed. Overall, the results suggest that improving the reliability of non-invasive phoneme decoding will likely require better handling of normalization-related distribution shift while also addressing the challenge of single-trial decoding.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 1 linked inside Pith

  1. [1]

    A literature review on integration of distributed energy resources in the perspective of control, protection and stability of microgrid,

    P. Basak, S. Chowdhury, S. H. nee Dey, and S. Chowdhury, “A literature review on integration of distributed energy resources in the perspective of control, protection and stability of microgrid,”Renewable and Sustainable Energy Reviews, vol. 16, no. 8, pp. 5545–5556, 2012

    2012

  2. [2]

    Goldsmith,Wireless communications

    A. Goldsmith,Wireless communications. Cambridge university press, 2005

    2005

  3. [3]

    Fully homomorphic encryption using ideal lattices,

    C. Gentry, “Fully homomorphic encryption using ideal lattices,” in Proceedings of the forty-first annual ACM symposium on Theory of computing, 2009, pp. 169–178

    2009

  4. [4]

    Protocols for secure computations,

    A. C. Yao, “Protocols for secure computations,” in23rd annual sympo- sium on foundations of computer science (sfcs 1982). IEEE, 1982, pp. 160–164

    1982

  5. [5]

    Calibrating noise to sensitivity in private data analysis,

    C. Dwork, F. McSherry, K. Nissim, and A. Smith, “Calibrating noise to sensitivity in private data analysis,” inTheory of cryptography conference. Springer, 2006, pp. 265–284

    2006

  6. [6]

    Peer-to- peer trading in electricity networks: An overview,

    W. Tushar, T. K. Saha, C. Yuen, D. Smith, and H. V . Poor, “Peer-to- peer trading in electricity networks: An overview,”IEEE transactions on smart grid, vol. 11, no. 4, pp. 3185–3200, 2020

    2020

  7. [7]

    Review of energy sharing: Business models, mechanisms, and prospects,

    Y . Chen and C. Zhao, “Review of energy sharing: Business models, mechanisms, and prospects,”IET Renewable Power Generation, vol. 16, no. 12, pp. 2468–2480, 2022

    2022

  8. [8]

    A priority- based approach for peer-to-peer energy trading using cooperative game theory in local energy community,

    S. Malik, M. Duffy, S. Thakur, B. Hayes, and J. Breslin, “A priority- based approach for peer-to-peer energy trading using cooperative game theory in local energy community,”International Journal of Electrical Power & Energy Systems, vol. 137, p. 107865, 2022

    2022

Show all 32 references
  1. [9]

    A motivational game-theoretic approach for peer-to-peer energy trading in the smart grid,

    W. Tushar, T. K. Saha, C. Yuen, T. Morstyn, M. D. McCulloch, H. V . Poor, and K. L. Wood, “A motivational game-theoretic approach for peer-to-peer energy trading in the smart grid,”Applied energy, vol. 243, pp. 10–20, 2019

    2019

  2. [10]

    Managing price uncertainty in prosumer-centric energy trading: A prospect-theoretic stackelberg game approach,

    G. El Rahi, S. R. Etesami, W. Saad, N. B. Mandayam, and H. V . Poor, “Managing price uncertainty in prosumer-centric energy trading: A prospect-theoretic stackelberg game approach,”IEEE Transactions on Smart Grid, vol. 10, no. 1, pp. 702–713, 2017

    2017

  3. [11]

    Energy sharing management for microgrids with pv prosumers: A Stackelberg game approach,

    N. Liu, X. Yu, C. Wang, and J. Wang, “Energy sharing management for microgrids with pv prosumers: A Stackelberg game approach,”IEEE Transactions on Industrial Informatics, vol. 13, no. 3, pp. 1088–1098, 2017

    2017

  4. [12]

    Game-based optimal aggregation of energy prosumer community with mixed-pricing scheme in two-settlement electricity market,

    J. Wang and G. Hu, “Game-based optimal aggregation of energy prosumer community with mixed-pricing scheme in two-settlement electricity market,”IEEE Transactions on Automation Science and Engineering, vol. 22, pp. 3433–3444, 2024

    2024

  5. [13]

    An energy sharing mechanism considering network constraints and market power limita- tion,

    Y . Chen, C. Zhao, S. H. Low, and A. Wierman, “An energy sharing mechanism considering network constraints and market power limita- tion,”IEEE transactions on smart grid, vol. 14, no. 2, pp. 1027–1041, 2022

    2022

  6. [14]

    Approaching prosumer social optimum via energy sharing with proof of convergence,

    Y . Chen, C. Zhao, S. H. Low, and S. Mei, “Approaching prosumer social optimum via energy sharing with proof of convergence,”IEEE Transactions on Smart Grid, vol. 12, no. 3, pp. 2484–2495, 2020

    2020

  7. [15]

    Online distributed generalized nash equilibrium seeking of energy sharing markets in distribution networks,

    Z. Wang, W. Hao, W. Wei, and Z. Sun, “Online distributed generalized nash equilibrium seeking of energy sharing markets in distribution networks,”IEEE Transactions on Automation Science and Engineering, 2025

    2025

  8. [16]

    Distributed energy trading in smart grid over directed communication network,

    M. H. Ullah and J.-D. Park, “Distributed energy trading in smart grid over directed communication network,”IEEE Transactions on Smart Grid, vol. 12, no. 4, pp. 3669–3672, 2021

    2021

  9. [17]

    Communication reliability-restricted energy sharing strategy in active distribution networks,

    L. Chen, J. Wang, Z. Wu, G. Li, M. Zhou, P. Li, and Y . Zhang, “Communication reliability-restricted energy sharing strategy in active distribution networks,”Applied Energy, vol. 282, p. 116238, 2021

    2021

  10. [18]

    Accelerating distributed optimization via over-the- air computing,

    N. A. Mitsiou, P. S. Bouzinis, P. D. Diamantoulakis, R. Schober, and G. K. Karagiannidis, “Accelerating distributed optimization via over-the- air computing,”IEEE Transactions on Communications, vol. 71, no. 9, pp. 5565–5579, 2023

    2023

  11. [19]

    Distributed over-the-air computing for fast distributed optimization: Beamforming design and convergence analysis,

    Z. Lin, Y . Gong, and K. Huang, “Distributed over-the-air computing for fast distributed optimization: Beamforming design and convergence analysis,”IEEE Journal on Selected Areas in Communications, vol. 41, no. 1, pp. 274–287, 2022

    2022

  12. [20]

    Reduced-dimension design of mimo over-the-air computing for data aggregation in clustered iot networks,

    D. Wen, G. Zhu, and K. Huang, “Reduced-dimension design of mimo over-the-air computing for data aggregation in clustered iot networks,” IEEE Transactions on Wireless Communications, vol. 18, no. 11, pp. 5255–5268, 2019

    2019

  13. [21]

    Federated learning over wireless device-to-device networks: Algorithms and convergence analysis,

    H. Xing, O. Simeone, and S. Bi, “Federated learning over wireless device-to-device networks: Algorithms and convergence analysis,”IEEE Journal on Selected Areas in Communications, vol. 39, no. 12, pp. 3723– 3741, 2021

    2021

  14. [22]

    User- level privacy-preserving federated learning: Analysis and performance optimization,

    K. Wei, J. Li, M. Ding, C. Ma, H. Su, B. Zhang, and H. V . Poor, “User- level privacy-preserving federated learning: Analysis and performance optimization,”IEEE Transactions on Mobile Computing, vol. 21, no. 9, pp. 3388–3401, 2021

    2021

  15. [23]

    Differential privacy in distributed optimization with gradient tracking,

    L. Huang, J. Wu, D. Shi, S. Dey, and L. Shi, “Differential privacy in distributed optimization with gradient tracking,”IEEE Transactions on Automatic Control, vol. 69, no. 9, pp. 5727–5742, 2024

    2024

  16. [24]

    A differentially private energy trading mechanism approaching social optimum,

    Y . Cao and Y . Chen, “A differentially private energy trading mechanism approaching social optimum,”IEEE Transactions on Smart Grid, 2025

    2025

  17. [25]

    Dif- ferentially private distributed algorithm for energy sharing game with generalized demand bidding,

    H. Liu, S. Lei, L. Zhang, Y . Huang, H. Zhang, and C. Peng, “Dif- ferentially private distributed algorithm for energy sharing game with generalized demand bidding,” in2024 IEEE Power & Energy Society General Meeting (PESGM). IEEE, 2024, pp. 1–5

    2024

  18. [26]

    Peer-to-peer energy trading of carbon-aware prosumers: An online accelerated distributed approach with differential privacy,

    X. Wei, Y . Xu, H. Sun, and W. K. Chan, “Peer-to-peer energy trading of carbon-aware prosumers: An online accelerated distributed approach with differential privacy,”IEEE Transactions on Smart Grid, vol. 15, no. 6, pp. 5595–5609, 2024

    2024

  19. [27]

    Differentially private nash equilibrium seeking in quadratic network games,

    L. Wang, K. Ding, Y . Leng, X. Ren, and G. Shi, “Differentially private nash equilibrium seeking in quadratic network games,”IEEE Transactions on Control of Network Systems, vol. 12, no. 1, pp. 673– 686, 2024

    2024

  20. [28]

    On the differential privacy in federated learning based on over-the-air computation,

    S. Park and W. Choi, “On the differential privacy in federated learning based on over-the-air computation,”IEEE Transactions on Wireless Communications, vol. 23, no. 5, pp. 4269–4283, 2023

    2023

  21. [29]

    Differentially private over-the-air federated learning over mimo fading channels,

    H. Liu, J. Yan, and Y .-J. A. Zhang, “Differentially private over-the-air federated learning over mimo fading channels,”IEEE Transactions on Wireless Communications, vol. 23, no. 8, pp. 8232–8247, 2024

    2024

  22. [30]

    Strategic gaming analysis for electric power systems: An MPEC approach,

    B. F. Hobbs, C. B. Metzler, and J.-S. Pang, “Strategic gaming analysis for electric power systems: An MPEC approach,”IEEE transactions on power systems, vol. 15, no. 2, pp. 638–645, 2000

    2000

  23. [31]

    S. M. Kay,Fundamentals of statistical signal processing: estimation theory. Prentice-Hall, Inc., 1993

    1993

  24. [32]

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

    J. Hoydis, S. Cammerer, F. Ait Aoudia, A. Vem, N. Bber, G. Marcus, and A. Keller, “Sionna: An Open-Source Library for Next-Generation Physical Layer Research,”arXiv preprint arXiv:2203.11854, 2022. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 11 APPENDIXA DERIVATI...

    2022 arXiv

Pith tools

Reviewed July 13, 2026 · model on record in the stance chip above.