Pith. sign in

REVIEW 4 major objections 5 minor 45 references

Nonlinear Bias-Compensated Adaptive Filter and Its Application for Time-Series Prediction

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read A random-Fourier adaptive filter with a bias-compensation term and a tunable robust loss handles input and output noise simultaneously at fixed computational cost.

desk verdict Useful incremental combination of RFF, bias compensation, and Barron's robust loss, but the bias-compensation term is never derived for the robust loss, so the paper's central theoretical claim doesn't hold as written. read the letter →

arxiv 2607.19902 v1 pith:TVPMYW5X submitted 2026-07-22 cs.LG eess.AS

Yi Peng , Haiquan Zhao , Jinhui Hu This is my paper
classification cs.LGeess.AS
keywords randomFourierfeatureserrors-in-variablesmodelbiascompensationadaptivefilteringgenerallossnon-Gaussiannoisetime-seriespredictionkernelmethods
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 claims that a nonlinear adaptive filter can simultaneously handle noise in both its input and its output without letting the network grow. The proposed RFFBCGA algorithm combines random Fourier features, a bias-compensation term derived for the errors-in-variables model, and a general adaptive loss function whose shape parameter lets it interpolate between LMS, LMLS, and MCC behavior. The authors argue that this fixes the two weaknesses of the existing bias-compensated kernel LMS (BCKLMS) algorithm: BCKLMS's fixed dictionary can miss input-signal structure, and its quadratic loss degrades under non-Gaussian output noise. They support the claim with mean and mean-square stability analysis, an identity showing the random-feature autocorrelation matrix is the identity scaled by 1/D even for noisy inputs, and simulations including noisy sunspot and Chua's circuit time-series prediction.

What carries the argument

The central object is the bias-compensated GA cost function: a general adaptive loss in the error minus a penalty built from the second derivative of the random-Fourier network output. The GA loss is the flexible component that recovers LMS, LMLS, and MCC for different shape-parameter values. The bias-compensation penalty and the resulting weighted outer-product term are what compensate for input noise; the random Fourier map keeps the network at a fixed D-dimensional size; and the two theorems showing that the autocorrelation matrix of the random features is (1/D)I, even for noisy inputs, are the simplifying identities that make the mean and mean-square analysis tractable.

What would settle it

Compute or estimate the true expected gradient of the GA cost under noisy input with errors-in-variables noise. If the expected gradient evaluated at the optimal weight vector differs from the clean-input gradient by something other than the compensation term, the bias compensation is not unbiased; a direct numerical check with a strongly non-quadratic shape parameter and input SNR 10 dB would settle it. Alternatively, run RFFBCGA with the input-noise variance deliberately misspecified, say set to twice or half the true value, and compare steady-state MSE to the same algorithm with the compens

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is a derivation. Starting from a general adaptive cost with a bias-compensation penalty, the update rule becomes: the GA-loss gradient term that gives robustness to outliers, plus a compensation term that counters the bias that input noise would otherwise introduce. The paper then proves that for cosine random Fourier features, E[G(u)G(u)^T] = (1/D)I, and that the same identity holds when u is replaced by the noisy input u + eta, because the zero off-diagonal terms come from averaging over the random phase theta in [0, 2π) rather than from the input statistics. That identity lets the stability analysis go through, yielding explicit mean and mean-square

Load-bearing premise

The bias-compensation term is borrowed from quadratic-loss analysis and is assumed to remain unbiased for the non-quadratic GA loss; the derivation also assumes that the input-noise variance is known exactly, and no proof is given that a key cross-expectation vanishes for the GA update.

Editorial extensions

If this is right

  • Users get a nonlinear adaptive filter whose computational cost per iteration stays fixed rather than growing with the number of samples.
  • In the tested nonlinear errors-in-variables settings, the algorithm converges faster than BCKLMS under Gaussian output noise and maintains that advantage under Bernoulli-Gaussian and alpha-stable output noise.
  • The derived step-size bounds give a practical convergence guarantee: mean stability holds below one bound and mean-square stability below half of it.
  • Because the network is dictionary-free, prediction quality does not depend on how the dictionary is chosen, removing a source of variance visible in the BCKLMS sunspot results.
  • Setting the shape parameter near negative infinity makes the update MCC-like, providing a built-in switch from quadratic to correntropy behavior by changing one hyperparameter.

Reading between the lines

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

  • The compensation term is written in terms of the second derivative of the network output, not of the loss, so the same correction may carry over to other random-Fourier cost functions; the paper demonstrates it only inside the GA family.
  • An online tuning rule for the shape and scale parameters, treating them as adaptive rather than hand-set constants, is a natural next step that the paper does not explore.
  • The identity E[G(u+eta)G(u+eta)^T] = (1/D)I suggests that other random-Fourier algorithms with random phase inherit a simplified steady-state analysis regardless of input noise statistics.
  • The fixed-structure property could be useful for distributed or embedded implementations where dictionary sharing is costly, though the paper does not test distributed settings.
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. This paper proposes RFFBCGA, a random-Fourier-feature adaptive filter that combines a bias-compensation (BC) term for input noise (in the style of BCKLMS) with the general adaptive (GA) loss for robustness against non-Gaussian output noise. The algorithm maintains a fixed network dimension. The paper provides mean and mean-square stability analyses and reports simulations on synthetic systems and two real-world time series (sunspot and Chua’s circuit). The claimed contributions are a new bias-compensated robust nonlinear filter and two theoretical results about the autocorrelation of cos-RFF features for clean and noisy inputs.

Significance. If the central claim were established, the paper would fill a genuine gap: existing nonlinear adaptive filters either lack input-noise compensation or are not robust to heavy-tailed output noise. The RFF framework gives a fixed-size filter, and the GA loss offers a flexible way to interpolate among LMS, LMLS, and MCC. The paper’s transparency is a plus: Algorithm 1 is clearly specified, and Theorems 1 and 2 are simple and verified in simulation. However, the theoretical foundation—the derivation of the bias-compensation penalty for the GA loss—is absent, and the EIV model is stated inconsistently, so the significance cannot be assessed from the current manuscript.

major comments (4)
  1. [Section II.B, Eq. (9)] The EIV model is mis-specified. Eq. (1) defines d_i = ϑ(u_i)+v_i with u_i the clean input; Eq. (9) rewrites it as d_i = ϑ(u_i+η_i)+v_i = ϑ(¯u_i)+v_i. These are different models. If d_i is generated from the noisy input, the filter is solving standard regression with noisy regressors and no bias compensation is needed. If d_i is generated from the clean input, then e_i in Eq. (32) must include the term [G(u_i)-G(¯u_i)]^T Ω_o, which is omitted. The update (18) and all subsequent analysis are therefore not for the EIV problem described in the introduction.
  2. [Section III.B, Eq. (15)] The bias-compensation penalty −γσ_in^4/4 (Σ_m ∂²ϑ̂/∂u_m²)² is introduced as “inspired by [39]” but no derivation is given for the GA loss. For a non-quadratic loss L(e), a second-order expansion of E_η[L(d−G^T(¯u)Ω)] yields bias terms of the form L′(e0)σ²ΔG and L″(e0)σ²‖∇G‖², which do not reduce to the loss-independent squared-Laplacian penalty used here. Since the algorithm’s EIV robustness claim (Contribution 1, Abstract) is carried by this term, the central theoretical claim is unsupported.
  3. [Section IV.A, Eq. (32)–(34)] The mean-stability recursion assumes exact RFF representation ϑ(¯u_i)=G^T(¯u_i)Ω_o and sets e_i = G^T(¯u_i)ΔΩ_{i-1}+v_i. This neglects the representation error and the input-noise mismatch. The step from Eq. (32) to Eq. (34) also uses independence assumptions that are only approximate. Consequently, the bound (40) may not characterize the implemented algorithm. The term “−0” in Eq. (34) is justified by E[G(¯u)]=0, but this should be stated explicitly.
  4. [Section IV.B, Eq. (41)] The mean-square stability result is presented as a lengthy expression without a derivation. The text refers to “Theorems 1-3” but Theorem 3 is never stated. Several terms (e.g., the factor (D+1/2) and the fourth-moment of W_D in Eq. (45)) appear without derivation. This prevents verification of the stability condition (52).
minor comments (5)
  1. [Algorithm 1 and Section III.B] The algorithm requires the input-noise variance σ_in² as an input (Step 8) but the paper never discusses how it is obtained or what happens when it is misspecified. The assumption that σ_in² is known exactly should be stated as a limitation.
  2. [Eq. (34)] The “−0” term in Eq. (34) is confusing. If it is meant as E[v_i Q(e_i)]=0, that is generally false; if it is E[G(¯u_i)]=0, that is true but should be stated.
  3. [Figures] The simulation figures do not show error bars or confidence intervals, and several parameter values (e.g., the RFF dimension D) are not given in the captions. This makes it hard to assess the statistical significance of the reported improvements.
  4. [Throughout] There are many typos and inconsistencies: “RFF-BCGA” vs “RFFBCGA,” “sufficient,” “efficient,” “A WGN,” and the caption of Fig. 2 omits the legend values. The reference [39] is incomplete (no volume/pages/DOI).
  5. [Section III.A, Eq. (14)] The text says “δ → +∞” but the preceding line reads “δ → −∞”; the formula itself appears correct for +∞.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: RFFBCGA combines an external GA loss with an imported BC penalty; no prediction reduces to its input by construction.

full rationale

The derivation chain is not circular. The proposed update (Eq. 18) is obtained by gradient descent on the cost function in Eq. 15, which adds a bias-compensation penalty borrowed from the external BCKLMS work [39] to the GA loss from Barron [41]. Neither component is defined in terms of the paper's own outputs or fitted values: the GA loss is an externally published function, and the BC penalty is imported from an external quadratic-loss analysis. The stability analysis (Eqs. 32-52) rests on Theorems 1-2, which are proven from the RFF construction and the uniformity of theta, and Section V validates those theorems; no fitted parameter is re-presented as a prediction. The self-citation [3] is not load-bearing and in fact does not appear to be cited in the text. The claim that BCKLMS is the only available nonlinear-EIV solution is attributed to external [39], not to the authors. The genuine weaknesses - that Eq. 15 adopts the quadratic-loss BC penalty for the non-quadratic GA loss without re-deriving it, and that Eq. 34 drops E[v_i Q(e_i)] = 0 without justification - are correctness or support gaps, not reductions of an output to an input by construction. Similarly, the stray '[?]' after 'bias-compensated (BC) algorithm' is a missing reference, not a circular step. The paper is self-contained against external baselines and no claimed prediction is forced by its own definitions, so the circularity score is 0.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard RFF approximation, the unproven applicability of a quadratic-loss bias-compensation penalty to the GA loss, known/estimated input-noise variance, and four conventional independence assumptions that are not derived. No new particles or physical entities are introduced.

free parameters (6)
  • γ (regularization parameter) = 0.45, 2, 0.15, 0.25, 0.3, 0.6, 0.8 across examples
    Chosen per experiment to make RFFBCGA converge; no selection rule or sensitivity analysis provided.
  • δ (GA shape parameter) = -100 in all experiments
    Fixed to approximate MCC; the algorithm is named 'general adaptive' but δ is not adapted online.
  • l (GA scale parameter) = 1
    Chosen by hand; no sensitivity analysis.
  • μGA (step size) = 0.01
    Chosen ad hoc; no automatic scheduling or rule beyond stability bounds.
  • σ_in² (input noise variance) = assumed known from experimental SNR
    Algorithm 1 requires σ_in² as an input; the paper provides no estimation method for real applications.
  • RFF dimension D and kernel width ξ = D=100; ξ=0.35, 0.6, 1.8
    Standard hyperparameters; tuned per experiment, not part of the central derivation.
assumptions (4)
  • standard math Bochner's theorem and the unbiased RFF approximation κ(u_i,u_t) ≈ G(u_i)^T G(u_t) (Eqs. 5–7)
    Standard random-feature theory; used without proof.
  • domain assumption Independence assumptions A1–A4 (weight-error uncorrelated with input, Q(e) uncorrelated with G, zero-mean Gaussian noises, small step-size)
    Section IV states these as 'fundamental assumptions'; A3 restricts output noise to Gaussian while the central claim includes non-Gaussian robustness.
  • ad hoc to paper The quadratic-loss bias-compensation penalty (γσ_in^4/4 Σ(∂²ϑ̂/∂u_m²)²) cancels input-noise bias when the loss is the non-quadratic GA function
    Eq. (15) imports the correction from BCKLMS/BC-LMS theory without deriving it for robust losses; this is the load-bearing premise of the algorithm.
  • domain assumption Exact knowledge of input-noise variance σ_in² in Algorithm 1
    Step 8 of Algorithm 1 requires σ_in²; the paper assumes it is known rather than estimating it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Nonlinear Bias-Compensated Adaptive Filter and Its Application for Time-Series Prediction." pith.science (2026). https://pith.science/paper/TVPMYW5X

@misc{pith2026260719902,
  author       = {Pith},
  title        = {Pith review of: Nonlinear Bias-Compensated Adaptive Filter and Its Application for Time-Series Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TVPMYW5X}},
  note         = {Machine review of arXiv:2607.19902}
}
read the original abstract

Most existing nonlinear adaptive filtering algorithms only account for output noise, neglecting the fact that input noise is also prevalent in practice. Although the recently proposed bias-compensated kernel least mean square (BCKLMS) algorithm addresses input noise in the nonlinear errors-in-variables (EIV) model, it still suffers from two major limitations. First, the use of a fixed-size dictionary restricts network growth but also prevents it from fully capturing the characteristics of the input signal. Second, as an least mean square (LMS) based algorithm, it exhibits poor robustness in the presence of non-Gaussian noise in the output signal. To overcome these issues, this paper proposes the random Fourier bias-compensated filter under general adaptive function (RFFBCGA) algorithm. Within the random Fourier feature based bias-compensated (RFFBC) framework, the proposed algorithm not only maintains a fixed network structure and effectively mitigates input noise interference through the BC term, but also achieves improved characterization of the input signal. Moreover, by leveraging the flexible form of the general adaptive (GA) function, the algorithm's robustness across various noise scenarios is further enhanced. Extensive simulations, including real-world time series prediction tasks, demonstrate the superiority of the proposed method.

Figures

Figures reproduced from arXiv: 2607.19902 by the authors.

Figure 1
Figure 1. Curves for different cost functions and gradients [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Example 1. Testing MSE with the kernel bandwidth [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 4
Figure 4. Testing MSE for prediction of noisy sunspots time series [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figures from the paper (4 more)
Figure 3
Figure 3. Figure 3: Example 2. Testing MSE with the kernel bandwidth [PITH_FULL_IMAGE:figures/full_fig_p008_3.png]
Figure 7
Figure 7. Figure 7: Testing MSE for prediction of noisy Chua’s circut time series [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 6
Figure 6. Figure 6: Chua’s circuit system schematic diagram 2) Chua’s circuit Time Series Prediction: This subsec￾tion utilizes a physical Chua’s circuit to generate chaotic time series, thereby facilitating the evaluation of the RFFBCGA algorithm’s nonlinear learning capability. The circ…
Figure 8
Figure 8. Figure 8: Simulation verification of Theorem 1 and 2 [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

45 extracted references · 1 linked inside Pith

  1. [39]

    Bias compen- sation for kernel least-mean-square algorithms,

    Y.-R. Chien, J.-L. Liu, E.-T. Lin, and G. Qian, “Bias compen- sation for kernel least-mean-square algorithms,” IEEE Sensors Letters, 2025

    2025

  2. [1]

    W. Liu, J. C. Principe, and S. Haykin, Kernel adaptive filtering: a comprehensive introduction. John Wiley & Sons, 2011

    2011

  3. [2]

    Zhao and B

    H. Zhao and B. Chen, Efficient Nonlinear Adaptive Filters: Design, Analysis and Applications. Springer Nature, 2023

    2023

  4. [3]

    A family of robust generalized adaptive filters and application for time-series prediction,

    Y. Peng, H. Zhao, and J. Hu, “A family of robust generalized adaptive filters and application for time-series prediction,” IEEE Transactions on Circuits and Systems I: Regular Papers, 2025

    2025

  5. [4]

    The kernel least-mean-square algorithm,

    W. Liu, P. P. Pokharel, and J. C. Principe, “The kernel least-mean-square algorithm,” IEEE Transactions on signal processing, vol. 56, no. 2, pp. 543–554, 2008

    2008

  6. [5]

    Mean square convergence analysis for kernel least mean square algorithm,

    B. Chen, S. Zhao, P. Zhu, and J. C. Príncipe, “Mean square convergence analysis for kernel least mean square algorithm,” Signal processing, vol. 92, no. 11, pp. 2624–2632, 2012

    2012

  7. [6]

    Mixture kernel least mean square,

    R. Pokharel, S. Seth, and J. C. Principe, “Mixture kernel least mean square,” in The 2013 International Joint Conference on Neural Networks (IJCNN). IEEE, 2013, pp. 1–7

    2013

  8. [7]

    The kernel recursive least- squares algorithm,

    Y. Engel, S. Mannor, and R. Meir, “The kernel recursive least- squares algorithm,” IEEE Transactions on signal processing, vol. 52, no. 8, pp. 2275–2285, 2004

    2004

Show all 45 references
  1. [8]

    Research progress on intelligent system’s learning, optimization, and control—part ii: online sparse kernel adaptive algorithm,

    F. Tan and X. Guan, “Research progress on intelligent system’s learning, optimization, and control—part ii: online sparse kernel adaptive algorithm,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 50, no. 12, pp. 5369–5385, 2018

    2018

  2. [9]

    Sparse online greedy support vector regression,

    Y. Engel, S. Mannor, and R. Meir, “Sparse online greedy support vector regression,” in European Conference on Machine Learning. Springer, 2002, pp. 84–96

    2002

  3. [10]

    Quantized kernel least mean square algorithm,

    B. Chen, S. Zhao, P. Zhu, and J. C. Príncipe, “Quantized kernel least mean square algorithm,” IEEE Transactions on Neural Networks and Learning Systems, vol. 23, no. 1, pp. 22–32, 2011

    2011

  4. [11]

    Quantized kernel lleast lncosh algorithm,

    Q. Wu, Y. Li, Y. V. Zakharov, and W. Xue, “Quantized kernel lleast lncosh algorithm,” Signal Processing, vol. 189, p. 108255, 2021

    2021

  5. [12]

    Online learning with kernels: Overcoming the growing sum problem,

    A. Singh, N. Ahuja, and P. Moulin, “Online learning with kernels: Overcoming the growing sum problem,” in 2012 IEEE international workshop on machine learning for signal process- ing. IEEE, 2012, pp. 1–6

    2012

  6. [13]

    On the error of random fourier features,

    D. J. Sutherland and J. Schneider, “On the error of random fourier features,” arXiv preprint arXiv:1506.02785, 2015

    2015 arXiv

  7. [14]

    Random features for large-scale kernel machines,

    A. Rahimi and B. Recht, “Random features for large-scale kernel machines,” Advances in neural information processing systems, vol. 20, 2007

    2007

  8. [15]

    Online optimization with costly and noisy measurements using random fourier expansions,

    L. Bliek, H. R. Verstraete, M. Verhaegen, and S. Wahls, “Online optimization with costly and noisy measurements using random fourier expansions,” IEEE transactions on neural networks and learning systems, vol. 29, no. 1, pp. 167–182, 2016

    2016

  9. [16]

    An improved robust kernel adaptive filtering method for time- series prediction,

    L. Shi, R. Lu, Z. Liu, J. Yin, Y. Chen, J. Wang, and L. Lu, “An improved robust kernel adaptive filtering method for time- series prediction,” IEEE Sensors Journal, vol. 23, no. 18, pp. 21 463–21 473, 2023

    2023

  10. [17]

    Robust kernel recursive adaptive filtering algorithms based on m-estimate,

    X. Yang, Y. Mu, K. Cao, M. Lv, B. Peng, Y. Zhang, and G. Wang, “Robust kernel recursive adaptive filtering algorithms based on m-estimate,” Signal Processing, vol. 207, p. 108952, 2023. 11

    2023

  11. [18]

    Kernel robust mixed-norm adaptive filtering,

    J. Liu, H. Qu, B. Chen, and W. Ma, “Kernel robust mixed-norm adaptive filtering,” in 2014 International Joint Conference on Neural Networks (IJCNN). IEEE, 2014, pp. 3021–3024

    2014

  12. [19]

    Kernel adaptive filtering with maximum correntropy criterion,

    S. Zhao, B. Chen, and J. C. Principe, “Kernel adaptive filtering with maximum correntropy criterion,” in The 2011 International Joint Conference on Neural Networks. IEEE, 2011, pp. 2012– 2017

    2011

  13. [20]

    Kernel least mean p-power algorithm,

    W. Gao and J. Chen, “Kernel least mean p-power algorithm,” IEEE Signal Processing Letters, vol. 24, no. 7, pp. 996–1000, 2017

    2017

  14. [21]

    Random fourier filters under maximum correntropy criterion,

    S. Wang, L. Dang, B. Chen, S. Duan, L. Wang, and C. K. Tse, “Random fourier filters under maximum correntropy criterion,” IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 65, no. 10, pp. 3390–3403, 2018

    2018

  15. [22]

    Mixture random features spaces for robust adaptive filtering,

    M. Cui, D. Lin, Y. Zheng, Y. Jiang, and S. Wang, “Mixture random features spaces for robust adaptive filtering,” in 2024 International Conference on Neuromorphic Computing (ICNC). IEEE, 2024, pp. 1–4

    2024

  16. [23]

    Laguerre kernel adaptive filter with arctangent criterion for nonlinear system identification,

    Y. Zhu, H. Zhao, and M. G. Christensen, “Laguerre kernel adaptive filter with arctangent criterion for nonlinear system identification,” IEEE Signal Processing Letters, 2024

    2024

  17. [24]

    Robust generalized maximum blake– zisserman total correntropy adaptive filter for generalized gaus- sian noise and noisy input,

    H. Zhao and Z. Cao, “Robust generalized maximum blake– zisserman total correntropy adaptive filter for generalized gaus- sian noise and noisy input,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 53, no. 11, pp. 6757–6765, 2023

    2023

  18. [25]

    Proportionate total adaptive filtering algorithms for sparse system identification,

    J. Ni, Y. Xing, Z. Zhu, and J. Chen, “Proportionate total adaptive filtering algorithms for sparse system identification,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 54, no. 3, pp. 1840–1852, 2023

    2023

  19. [26]

    Maximum total complex correntropy for adaptive filter,

    G. Qian, S. Wang, and H. H. Iu, “Maximum total complex correntropy for adaptive filter,” IEEE Transactions on Signal Processing, vol. 68, pp. 978–989, 2020

    2020

  20. [27]

    Total least squares normalized subband adaptive filter algorithm for noisy input,

    H. Zhao, Y. Chen, J. Liu, and Y. Zhu, “Total least squares normalized subband adaptive filter algorithm for noisy input,” IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 69, no. 3, pp. 1977–1981, 2021

    1977

  21. [28]

    A fractional-order gradient-descent total least mean p-norm adaptive filtering algorithm in impulsive noise environments,

    J. Yang, Q. Zhang, Y. Luo, and S. Yan, “A fractional-order gradient-descent total least mean p-norm adaptive filtering algorithm in impulsive noise environments,” IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 70, no. 3, pp. 1204–1208, 2022

    2022

  22. [29]

    Robust bias-compensated cr-nsaf algorithm: Design and per- formance analysis,

    P. Wen, B. Wang, B. Qu, S. Zhang, H. Zhao, and J. Liang, “Robust bias-compensated cr-nsaf algorithm: Design and per- formance analysis,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2024

    2024

  23. [30]

    Bias-compensated sign subband adaptive filter algorithm with individual weighting factors for input noise,

    H. Zhao, D. Liu, and X. He, “Bias-compensated sign subband adaptive filter algorithm with individual weighting factors for input noise,” IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 69, no. 3, pp. 1872–1876, 2021

    2021

  24. [31]

    A bias-compensated nmmcc algorithm against noisy input and non-gaussian interference,

    X. Long, H. Zhao, and X. Hou, “A bias-compensated nmmcc algorithm against noisy input and non-gaussian interference,” IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 70, no. 9, pp. 3689–3693, 2023

    2023

  25. [32]

    Stability analysis of the bias compensated lms algorithm,

    R. M. Pimenta, M. R. Petraglia, and D. B. Haddad, “Stability analysis of the bias compensated lms algorithm,” Digital Signal Processing, vol. 147, p. 104395, 2024

    2024

  26. [33]

    Robust bias-compensated subband adaptive filtering,

    F. Song, F. Huang, S. Zhang, H. Chen, and P. Wen, “Robust bias-compensated subband adaptive filtering,” IEEE Transac- tions on Circuits and Systems II: Express Briefs, vol. 70, no. 9, pp. 3679–3683, 2023

    2023

  27. [34]

    Bias-compensated mccc algorithm for widely linear adaptive filtering with noisy data,

    F. Dong, G. Qian, and S. Wang, “Bias-compensated mccc algorithm for widely linear adaptive filtering with noisy data,” IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 67, no. 12, pp. 3587–3591, 2020

    2020

  28. [35]

    Enhanced bias- compensated nlms for adaptive doa estimation,

    S. Joel, S. K. Yadav, and N. V. George, “Enhanced bias- compensated nlms for adaptive doa estimation,” IEEE Sensors Letters, vol. 8, no. 5, pp. 1–4, 2024

    2024

  29. [36]

    Bias- compensated normalized maximum correntropy criterion al- gorithm for system identification with noisy input,

    W. Ma, D. Zheng, Y. Li, Z. Zhang, and B. Chen, “Bias- compensated normalized maximum correntropy criterion al- gorithm for system identification with noisy input,” Signal Processing, vol. 152, pp. 160–164, 2018

    2018

  30. [37]

    Bias-compensated sign algorithm for noisy inputs and its step-size optimization,

    J. Ni, Y. Gao, X. Chen, and J. Chen, “Bias-compensated sign algorithm for noisy inputs and its step-size optimization,” IEEE Transactions on Signal Processing, vol. 69, pp. 2330–2342, 2021

    2021

  31. [38]

    An efficient recursive total least squares algo- rithm for fir adaptive filtering,

    C. E. Davila, “An efficient recursive total least squares algo- rithm for fir adaptive filtering,” IEEE Transactions on Signal Processing, vol. 42, no. 2, pp. 268–280, 2002

    2002

  32. [40]

    Algorithm and architecture design of random fourier features- based kernel adaptive filters,

    V. C. Gogineni, R. Sambangi, D. Alex, S. Mula, and S. Werner, “Algorithm and architecture design of random fourier features- based kernel adaptive filters,” IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 70, no. 2, pp. 833–845, 2022

    2022

  33. [41]

    A general and adaptive robust loss function,

    J. T. Barron, “A general and adaptive robust loss function,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 4331–4339

    2019

  34. [42]

    S. S. Haykin, Adaptive filter theory. Pearson Education India, 2002

    2002

  35. [43]

    Robust least mean logarithmic square adaptive filtering algorithms,

    K. Xiong and S. Wang, “Robust least mean logarithmic square adaptive filtering algorithms,” Journal of the Franklin Institute, vol. 356, no. 1, pp. 654–674, 2019

    2019

  36. [44]

    Correntropy: Properties and applications in non-gaussian signal processing,

    W. Liu, P. P. Pokharel, and J. C. Principe, “Correntropy: Properties and applications in non-gaussian signal processing,” IEEE Transactions on signal processing, vol. 55, no. 11, pp. 5286–5298, 2007

    2007

  37. [45]

    Predicting sun spots using a layered perceptron neural network,

    Y. R. Park, T. J. Murray, and C. Chen, “Predicting sun spots using a layered perceptron neural network,” IEEE Transactions on Neural Networks, vol. 7, no. 2, pp. 501–505, 1996

    1996

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.