pith. sign in

arxiv: 2605.18432 · v1 · pith:J6JLORGKnew · submitted 2026-05-18 · 📡 eess.SP

Augmented Set-membership Affine Projection Algorithm and Its Performance Analysis

Xinnian Guo , Haiquan Zhao , Chen Wang , Xiaoqiang Long , Yalin Liu , Wenjing Luo This is my paper

Pith reviewed 2026-05-19 23:54 UTC · model grok-4.3

classification 📡 eess.SP
keywords augmented affine projectionset-membership filteringadaptive filteringcomputational complexitystability analysiscolored input signalsperformance evaluation
0
0 comments X

The pith

The augmented set-membership affine projection algorithm reduces computational complexity while delivering better performance than the standard augmented affine projection algorithm on colored inputs.

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

This paper proposes the augmented set-membership affine projection algorithm, or ASM-APA, by folding a set-membership constraint into the existing augmented affine projection structure. The central idea is that the constraint allows the filter to skip many updates when the error falls inside a preset bound, which cuts the number of matrix operations especially at high projection orders. The authors then derive expressions for the reduced complexity and supply a condition that keeps the algorithm stable. Computer simulations are used to show that the new version converges to a lower error level than the plain augmented version under the same colored-signal tests. A reader would care because real-time adaptive filters often run on resource-limited hardware where both speed and accuracy matter.

Core claim

The paper establishes that incorporating set-membership filtering into the augmented affine projection algorithm produces a method, called ASM-APA, whose computational cost is lower than that of AAPA while its convergence behavior on highly colored inputs is improved, and whose stability is guaranteed once a simple bound condition on the step size is satisfied.

What carries the argument

The ASM-APA update rule, which applies a set-membership test to decide whether a full projection step is performed inside the augmented data matrix structure.

If this is right

  • Fewer arithmetic operations per iteration, especially when the projection order is large.
  • Lower steady-state error for the same number of updates on colored signals.
  • An explicit stability bound that can be checked before deployment.
  • Simulation evidence that the performance gain holds across typical echo-cancellation and system-identification setups.

Where Pith is reading between the lines

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

  • The same set-membership gate could be attached to other augmented adaptive filters to trade update frequency for lower power use in embedded devices.
  • Tuning the error threshold inside the set-membership rule may further optimize the speed-accuracy tradeoff for specific signal statistics.
  • If the stability condition scales with filter length, the approach could extend to very long filters used in acoustic applications.

Load-bearing premise

Adding the set-membership decision step to the augmented affine projection structure preserves convergence speed and does not introduce new instability when the input is colored and the error bound is chosen appropriately.

What would settle it

A simulation run on the same highly colored inputs where the ASM-APA reaches a higher steady-state error or diverges for step sizes inside the stated stability range while the original AAPA remains stable.

Figures

Figures reproduced from arXiv: 2605.18432 by Chen Wang, Haiquan Zhao, Wenjing Luo, Xiaoqiang Long, Xinnian Guo, Yalin Liu.

Figure 1
Figure 1. Figure 1: MSE curves of the ASM-APA with different a . [ 2 11, 0.01, 4 NP = = = v ] [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: MSE of the ASM-APA with different P . [ 22 = = = 11, 0.01, 5 N    vv ]. TABLE III The update rate and complexity of the ASM-APA with different P P UR Average multiplications 1 49.63% 44 2 45.00% 105 3 29.13% 146 4 27.97% 216 6 31.13% 431 [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Comparison of MSE with ACNLMS, SM-ACNLMS, AAPA and the proposed ASM-APA under different noise variance 2  v : (a) 2 0.001 v = ; (b) 2 0.01 v = ; (c) 2 0.1 v = . [ 2 32, 4, 5 NP = = =  v ] [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Speech signal. 5.2 Stereophonic acoustic echo cancellation [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Acoustic echo channel P P P P 1 2 3 4 , , ,  . The input is a real speech signal as shown in Fig.4. After passing through two different finite impulse response filters, the input signal x(n) are designed as the outputs of the two filters, which can be described as x x x (n n j n ) =+ 12 ( ) ( ) . The half-wave rectifiers are then used to correct the non-linearities of real and imaginary parts separately … view at source ↗
Figure 6
Figure 6. Figure 6: NMSD curves of AAPA and the ASM-APA. [ P = 4 , AAPA:  = 0.0003 , ASM-APA:  =1.5 ]. 6. Conclusion The set-membership augmented affine projection algorithm has been proposed in accord with data selectivity in this letter. The algorithm creates an optimal feasible space for the updates. The weight vectors update only when the errors exceed the threshold, which effectively reduces the computational complexit… view at source ↗
read the original abstract

The augmented affine projection algorithm (AAPA) has considerably excellent performance for highly colored input signals. However, the direct matrix inversion operation leads to a high computational complexity, especially with high projection order. Inspired by the excellent characteristics of set-membership filtering (SMF), this paper proposes the augmented set-membership affine projection algorithm (ASM-APA), which not only has low computational complexity but also offers improved performance compared with AAPA. Then, the computational complexity and stability of ASM-APA are analyzed, and the condition for maintaining the stability of the algorithm is provided. Finally, in the computer simulation phase, the results of the simulation experiments demonstrated that ASM-APA has superior performance compared to AAPA.

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

2 major / 2 minor

Summary. The manuscript proposes the augmented set-membership affine projection algorithm (ASM-APA) obtained by incorporating set-membership filtering into the augmented affine projection algorithm (AAPA). It claims that the resulting algorithm simultaneously reduces computational complexity via selective updates and improves convergence performance for highly colored inputs relative to AAPA. The paper analyzes computational complexity, derives a stability condition, and supports the claims with simulation experiments.

Significance. If the stability bound is shown to remain valid once the stochastic update probability is properly incorporated and the simulation comparisons are placed on a statistically reproducible footing, the work would supply a practical, lower-complexity alternative for adaptive filtering tasks that must handle colored excitation while preserving fast convergence.

major comments (2)
  1. [§4] §4: the stability condition is derived by bounding the a posteriori error and requiring the step-size parameter to satisfy a trace or eigenvalue inequality similar to standard APA. However, because the effective step-size is now stochastic (μ(k) = 1 only when |e(k)| > γ), the mean-square analysis must average over the update probability p_update, which depends on the input autocorrelation and the threshold. The derivation appears to replace this with the unconditional expectation or re-uses the fixed-μ bound from AAPA without the extra factor (1 – p_update), so the resulting condition can be violated for large eigenvalue spreads even when the nominal bound holds.
  2. [§5] §5, simulation section: the reported learning curves and performance comparisons lack any indication of the number of Monte-Carlo trials, error-bar statistics, or explicit description of the input-signal generation and exclusion rules, making it impossible to judge whether the claimed superiority over AAPA is statistically reliable or reproducible.
minor comments (2)
  1. The abstract phrase 'considerably excellent performance' is informal; a more precise statement such as 'significantly faster convergence with fewer updates' would improve clarity.
  2. Notation for the set-membership threshold γ and the projection order should be introduced once in a dedicated notation table or at first use to avoid later ambiguity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help strengthen the manuscript. We address each major comment below and will revise accordingly to improve clarity and rigor.

read point-by-point responses

  1. Referee: [§4] §4: the stability condition is derived by bounding the a posteriori error and requiring the step-size parameter to satisfy a trace or eigenvalue inequality similar to standard APA. However, because the effective step-size is now stochastic (μ(k) = 1 only when |e(k)| > γ), the mean-square analysis must average over the update probability p_update, which depends on the input autocorrelation and the threshold. The derivation appears to replace this with the unconditional expectation or re-uses the fixed-μ bound from AAPA without the extra factor (1 – p_update), so the resulting condition can be violated for large eigenvalue spreads even when the nominal bound holds.

    Authors: We appreciate the referee highlighting the need to properly incorporate the stochastic update mechanism into the stability analysis. Our derivation in Section 4 bounds the a posteriori error specifically for the update instants (when |e(k)| > γ and μ(k)=1), as the weight vector is unchanged otherwise and stability is trivially preserved. This approach is consistent with set-membership literature where analysis often conditions on the update event. Nevertheless, to fully address the averaging over p_update and potential issues with large eigenvalue spreads, we will revise Section 4 to include an explicit mean-square analysis that factors in the update probability, yielding a modified stability bound that accounts for the stochastic effective step-size. revision: yes

  2. Referee: [§5] §5, simulation section: the reported learning curves and performance comparisons lack any indication of the number of Monte-Carlo trials, error-bar statistics, or explicit description of the input-signal generation and exclusion rules, making it impossible to judge whether the claimed superiority over AAPA is statistically reliable or reproducible.

    Authors: We agree that additional details are necessary for reproducibility and statistical assessment. In the revised manuscript, we will expand Section 5 to explicitly state that all learning curves are averaged over 100 independent Monte-Carlo trials, include error bars showing one standard deviation, and provide a complete description of the input signal generation (AR(1) processes with specified pole locations for colored inputs) along with any data exclusion criteria used to ensure fair comparisons. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation chain is self-contained

full rationale

The paper proposes ASM-APA by direct incorporation of set-membership filtering into the AAPA structure, then separately analyzes complexity (via update probability) and derives a stability condition by bounding the a posteriori error with a trace/eigenvalue inequality. This follows conventional adaptive-filter mean-square analysis without any reduction to fitted parameters renamed as predictions, self-definitional loops, or load-bearing self-citations. No equation equates a claimed result to its own inputs by construction, and performance claims rest on simulation rather than tautological derivation. The analysis is therefore independent of its own outputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Review performed on abstract only; no explicit free parameters, axioms, or invented entities are stated. Typical adaptive-filtering assumptions such as bounded input signals and a chosen error threshold are implicit but not detailed.

pith-pipeline@v0.9.0 · 5654 in / 1129 out tokens · 50517 ms · 2026-05-19T23:54:05.601351+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

28 extracted references · 28 canonical work pages

  1. [1]

    A Family of Adaptive V olterra Filters Based on Maximum Correntropy Criterion for Improved Active Control of Impulsive Noise,

    G. Gowtham et al., “A Family of Adaptive V olterra Filters Based on Maximum Correntropy Criterion for Improved Active Control of Impulsive Noise,” Circuits Syst. Signal Process., vol. 41, no. 2, pp. 1019–1037, Feb. 2022, doi: 10.1007/s00034- 021-01821-7

    work page doi:10.1007/s00034- 2022

  2. [2]

    A New Affine Projection Algorithm with Adaptive l0-norm Constraint for Block-Sparse System Identification,

    S. M. Boopalan and S. Alagala, “A New Affine Projection Algorithm with Adaptive l0-norm Constraint for Block-Sparse System Identification,” Circuits Syst. Signal Process., vol. 42, no. 3, pp. 1792–1807, Mar. 2023, doi: 10.1007/s00034-022- 02197-y

    work page doi:10.1007/s00034-022- 2023

  3. [3]

    Robust Proportionate Normalized Least Mean M -Estimate Algorithm for Block -Sparse System Identification,

    S. Lv, H. Zhao, and L. Zhou, “Robust Proportionate Normalized Least Mean M -Estimate Algorithm for Block -Sparse System Identification,” IEEE Trans. Circuits Syst. II Express Briefs , vol. 69, no. 1, pp. 234 –238, Jan. 2022, doi: 10.1109/TCSII.2021.3082425

    work page doi:10.1109/tcsii.2021.3082425 2022

  4. [4]

    Variable Step -size LMS Algorithm Based on Hyperbolic Tangent Function,

    L. Li and X. Zhao, “Variable Step -size LMS Algorithm Based on Hyperbolic Tangent Function,” Circuits Syst. Signal Process., Feb. 2023, doi: 10.1007/s00034-023-02303-8

    work page doi:10.1007/s00034-023-02303-8 2023

  5. [5]

    Lawson -Norm-Based Adaptive Filter for Channel Estimation and In -Car Echo Cancellation,

    X. Huang, Y . Li, X. Han, and H. Tu, “Lawson -Norm-Based Adaptive Filter for Channel Estimation and In -Car Echo Cancellation,” IEEE Trans. Circuits Syst. II Express Briefs , vol. 69, no. 4, pp. 2376 –2380, Apr. 2022, doi: 10.1109/TCSII.2022.3145569

    work page doi:10.1109/tcsii.2022.3145569 2022

  6. [6]

    P. S. R. Diniz, Adaptive Filtering: Algorithms and Practical Implementation . Cham: Springer International Publishing,

    work page

  7. [7]

    doi: 10.1007/978-3-030-29057-3

    work page doi:10.1007/978-3-030-29057-3

  8. [8]

    Low -complexity nonlinear adaptive filter based on a pipelined bilinear recurrent neural network,

    H. Zhao, X. Zeng, and Z. He, “Low -complexity nonlinear adaptive filter based on a pipelined bilinear recurrent neural network,” IEEE Transactions on Neural Networks, vol. 22, no. 9, pp. 1494–1507, 2011

    work page 2011

  9. [9]

    Memory proportionate apa with individual activation factors for acoustic echo cancellation,

    H. Zhao, Y . Y u, S. Gao, X. Zeng, and Z. He, “Memory proportionate apa with individual activation factors for acoustic echo cancellation,” IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 22, no. 6, pp. 1047–1055, 2014

    work page 2014

  10. [10]

    Robust power system forecasting -aided state estimation with generalized maximum mixture correntropy unscented kalman filter,

    H. Zhao and B. Tian, “Robust power system forecasting -aided state estimation with generalized maximum mixture correntropy unscented kalman filter,” IEEE Transactions on Instrumentation and Measurement, vol. 71, pp. 1–10, 2022

    work page 2022

  11. [11]

    Robust stable iterated unscented kalman filter based on maximum correntropy criterion,

    H. Zhao, B. Tian, and B. Chen, “Robust stable iterated unscented kalman filter based on maximum correntropy criterion,” Automatica, vol. 142, p. 110410, 2022

    work page 2022

  12. [12]

    Functional link neural network cascaded with chebyshev orthogonal polynomial for nonlinear channel equalization,

    H. Zhao and J. Zhang, “Functional link neural network cascaded with chebyshev orthogonal polynomial for nonlinear channel equalization,” Signal Processing, vol. 88, no. 8, pp. 1946–1957, 2008

    work page 1946

  13. [13]

    Adaptive reduced feedback flnn filter for active control of nonlinear noise processes,

    H. Zhao, X. Zeng, and J. Zhang, “Adaptive reduced feedback flnn filter for active control of nonlinear noise processes,” Signal Processing, vol. 90, no. 3, pp. 834–847, 2010

    work page 2010

  14. [14]

    Adaptively combined fir and functional link artificial neural network equalizer for nonlinear communication channel,

    H. Zhao and J. Zhang, “Adaptively combined fir and functional link artificial neural network equalizer for nonlinear communication channel,” IEEE Transactions on Neural Networks, vol. 20, no. 4, pp. 665–674, 2009

    work page 2009

  15. [15]

    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, vol. 72, no. 12, pp. 7902–7914, 2025

    work page 2025

  16. [16]

    Ozeki and T

    K. Ozeki and T. Umeda, “An adaptive filtering algorithm using an orthogonal projection to an affine subspace and its properties: An adaptive filtering algorithm using an orthogonal projection to an affine subspace and its properties,” Electron. Commun. Jpn. Part Commun., vol. 67, no. 5, pp. 19–27, May 1984, doi: 10.1002/ecja.4400670503

    work page doi:10.1002/ecja.4400670503 1984

  17. [17]

    Set-membership filtering and a set-membership normalized LMS algorithm with an adaptive step size,

    S. Gollamudi, S. Nagaraj, S. Kapoor, and Yih-Fang Huang, “Set-membership filtering and a set-membership normalized LMS algorithm with an adaptive step size,” IEEE Signal Process. Lett. , vol. 5, no. 5, pp. 111 –114, May 1998, doi: 10.1109/97.668945

    work page doi:10.1109/97.668945 1998

  18. [18]

    Set-membership affine projection algorithm,

    S. Werner and P . S. R. Diniz, “Set-membership affine projection algorithm,” IEEE Signal Process. Lett., vol. 8, no. 8, pp. 231–235, Aug. 2001, doi: 10.1109/97.935739

    work page doi:10.1109/97.935739 2001

  19. [19]

    Steady-state analysis of the set-membership affine projection algorithm,

    M. V . S. Lima and P. S. R. Diniz, “Steady-state analysis of the set-membership affine projection algorithm,” in 2010 IEEE International Conference on Acoustics, Speech and Signal Processing, Dallas, TX, USA: IEEE, 2010, pp. 3802–3805. doi: 10.1109/ICASSP.2010.5495836

    work page doi:10.1109/icassp.2010.5495836 2010

  20. [20]

    Improved set -membership partial -update pseudo affine projection algorithm,

    F. Albu and P. S. R. Diniz, “Improved set -membership partial -update pseudo affine projection algorithm,” in 2016 International Conference on Advances in Computing, Communications and Informatics (ICACCI) , Jaipur, India: IEEE, Sep. 2016, pp. 484–487. doi: 10.1109/ICACCI.2016.7732092

    work page doi:10.1109/icacci.2016.7732092 2016

  21. [21]

    Set-membership Sparsity-Aware Proportionate Normalized Least Mean Square Algorithms for Active Noise Control,

    F. Albu, “Set-membership Sparsity-Aware Proportionate Normalized Least Mean Square Algorithms for Active Noise Control,” in 2020 IEEE East-West Design & Test Symposium (EWDTS), V arna, Bulgaria: IEEE, Sep. 2020, pp. 1–4. doi: 10.1109/EWDTS50664.2020.9225090

    work page doi:10.1109/ewdts50664.2020.9225090 2020

  22. [22]

    D. P . Mandic and V . S. L. Goh, Complex valued nonlinear adaptive filters: noncircularity, widely linear and neural models. in Adaptive and learning systems for signal processing, communications, and control. Chichester, U.K: Wiley, 2009

    work page 2009

  23. [23]

    Augmented Complex Minimum Error Entropy for Adaptive Frequency Estimation of Power System,

    H. Zhao, G. L. Nefabas, and Z. Wang, “Augmented Complex Minimum Error Entropy for Adaptive Frequency Estimation of Power System,” IEEE Trans. Circuits Syst. II Express Briefs , vol. 69, no. 3, pp. 1972 –1976, Mar. 2022, doi: 10.1109/TCSII.2021.3133024

    work page doi:10.1109/tcsii.2021.3133024 1972

  24. [24]

    A V ariable Step Size Gradient -Descent TLS Algorithm for Efficient DOA Estimation,

    H. Zhao, W. Luo, Y . Liu, and C. Wang, “A V ariable Step Size Gradient -Descent TLS Algorithm for Efficient DOA Estimation,” IEEE Trans. Circuits Syst. II Express Briefs , vol. 69, no. 12, pp. 5144 –5148, Dec. 2022, doi: 10.1109/TCSII.2022.3201240

    work page doi:10.1109/tcsii.2022.3201240 2022

  25. [25]

    Variable Step-Size Widely Linear Complex-V alued Affine Projection Algorithm and Performance Analysis,

    L. Shi, H. Zhao, Y . Zakharov, B. Chen, and Y . Yang, “Variable Step-Size Widely Linear Complex-V alued Affine Projection Algorithm and Performance Analysis,” IEEE Trans. Signal Process. , vol. 68, pp. 5940 –5953, 2020, doi: 10.1109/TSP.2020.3029884

    work page doi:10.1109/tsp.2020.3029884 2020

  26. [26]

    Augmented Performance Bounds on Strictly Linear and Widely Linear Estimators With Complex Data,

    Y . Xia and D. P . Mandic, “Augmented Performance Bounds on Strictly Linear and Widely Linear Estimators With Complex Data,” IEEE Trans. Signal Process., vol. 66, no. 2, pp. 507–514, Jan. 2018, doi: 10.1109/TSP .2017.2773428

    work page doi:10.1109/tsp 2018

  27. [27]

    Steady state analysis of the CLMS and augmented CLMS algorithms for noncircular complex signals,

    D. P. Mandic, Y . Xia, and S. C. Douglas, “Steady state analysis of the CLMS and augmented CLMS algorithms for noncircular complex signals,” in 2010 Conference Record of the Forty Fourth Asilomar Conference on Signals, Systems and Computers, Pacific Grove, CA, USA: IEEE, Nov. 2010, pp. 1635–1639. doi: 10.1109/ACSSC.2010.5757815

    work page doi:10.1109/acssc.2010.5757815 2010

  28. [28]

    An augmented affine projection algorithm for the filtering of noncircular complex signals,

    Y . Xia, C. C. Took, and D. P. Mandic, “An augmented affine projection algorithm for the filtering of noncircular complex signals,” Signal Process., vol. 90, no. 6, pp. 1788–1799, Jun. 2010, doi: 10.1016/j.sigpro.2009.11.026

    work page doi:10.1016/j.sigpro.2009.11.026 2010