On the Time-Variant Step Size Posterior LMS Algorithm

Syed Ahmed Pasha; Muhammad Omer Bin Saeed; Azzedine Zerguine

doi:10.1109/IBCAST59916.2023.10713033

On the Time-Variant Step Size Posterior LMS Algorithm

Syed Ahmed Pasha^*, Muhammad Omer Bin Saeed, Azzedine Zerguine

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

The conventional least mean square (LMS) algorithm can be regarded as a numerical approximation of a set of first-order ordinary differential equations developed using the explicit method. The convergence and performance characteristics of the LMS algorithm are determined by the relationship between the fixed gain and the input correlation statistics. The problem is less severe in the normalized LMS algorithm which removes the dependence on the input correlation statistics but limits the parameter value to a finite fixed interval for stable behavior. In this paper, we show, perhaps for the first time, the stability condition for the vector case of the LMS-type algorithm developed using the implicit method. Based on this, we present a time-variant step size LMS algorithm. Empirical studies demonstrate the superior convergence and steady-state behavior of the proposed filter.

Original language	English
Title of host publication	Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	136-141
Number of pages	6
ISBN (Electronic)	9798350308259
DOIs	https://doi.org/10.1109/IBCAST59916.2023.10713033
State	Published - 2023
Event	20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023 - Bhurban, Mur ree, Pakistan Duration: 22 Aug 2023 → 25 Aug 2023

Publication series

Name	Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023

Conference

Conference	20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023
Country/Territory	Pakistan
City	Bhurban, Mur ree
Period	22/08/23 → 25/08/23

Bibliographical note

Publisher Copyright:
© 2023 IEEE.

Keywords

adaptive algorithm
implicit method
least mean square
steady-state performance
variable step size

ASJC Scopus subject areas

Aerospace Engineering
Civil and Structural Engineering
Ceramics and Composites
Electronic, Optical and Magnetic Materials
Metals and Alloys
Surfaces, Coatings and Films

Access to Document

10.1109/IBCAST59916.2023.10713033

Cite this

Pasha, S. A., Bin Saeed, M. O., & Zerguine, A. (2023). On the Time-Variant Step Size Posterior LMS Algorithm. In Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023 (pp. 136-141). (Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/IBCAST59916.2023.10713033

Pasha, Syed Ahmed ; Bin Saeed, Muhammad Omer ; Zerguine, Azzedine. / On the Time-Variant Step Size Posterior LMS Algorithm. Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023. Institute of Electrical and Electronics Engineers Inc., 2023. pp. 136-141 (Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023).

@inproceedings{a8f95b9369f2460c8b114996ccc3e1bd,

title = "On the Time-Variant Step Size Posterior LMS Algorithm",

abstract = "The conventional least mean square (LMS) algorithm can be regarded as a numerical approximation of a set of first-order ordinary differential equations developed using the explicit method. The convergence and performance characteristics of the LMS algorithm are determined by the relationship between the fixed gain and the input correlation statistics. The problem is less severe in the normalized LMS algorithm which removes the dependence on the input correlation statistics but limits the parameter value to a finite fixed interval for stable behavior. In this paper, we show, perhaps for the first time, the stability condition for the vector case of the LMS-type algorithm developed using the implicit method. Based on this, we present a time-variant step size LMS algorithm. Empirical studies demonstrate the superior convergence and steady-state behavior of the proposed filter.",

keywords = "adaptive algorithm, implicit method, least mean square, steady-state performance, variable step size",

author = "Pasha, {Syed Ahmed} and {Bin Saeed}, {Muhammad Omer} and Azzedine Zerguine",

note = "Publisher Copyright: {\textcopyright} 2023 IEEE.; 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023 ; Conference date: 22-08-2023 Through 25-08-2023",

year = "2023",

doi = "10.1109/IBCAST59916.2023.10713033",

language = "English",

series = "Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "136--141",

booktitle = "Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023",

address = "United States",

}

Pasha, SA, Bin Saeed, MO & Zerguine, A 2023, On the Time-Variant Step Size Posterior LMS Algorithm. in Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023. Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023, Institute of Electrical and Electronics Engineers Inc., pp. 136-141, 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023, Bhurban, Mur ree, Pakistan, 22/08/23. https://doi.org/10.1109/IBCAST59916.2023.10713033

On the Time-Variant Step Size Posterior LMS Algorithm. / Pasha, Syed Ahmed; Bin Saeed, Muhammad Omer; Zerguine, Azzedine.
Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023. Institute of Electrical and Electronics Engineers Inc., 2023. p. 136-141 (Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - On the Time-Variant Step Size Posterior LMS Algorithm

AU - Pasha, Syed Ahmed

AU - Bin Saeed, Muhammad Omer

AU - Zerguine, Azzedine

PY - 2023

Y1 - 2023

N2 - The conventional least mean square (LMS) algorithm can be regarded as a numerical approximation of a set of first-order ordinary differential equations developed using the explicit method. The convergence and performance characteristics of the LMS algorithm are determined by the relationship between the fixed gain and the input correlation statistics. The problem is less severe in the normalized LMS algorithm which removes the dependence on the input correlation statistics but limits the parameter value to a finite fixed interval for stable behavior. In this paper, we show, perhaps for the first time, the stability condition for the vector case of the LMS-type algorithm developed using the implicit method. Based on this, we present a time-variant step size LMS algorithm. Empirical studies demonstrate the superior convergence and steady-state behavior of the proposed filter.

AB - The conventional least mean square (LMS) algorithm can be regarded as a numerical approximation of a set of first-order ordinary differential equations developed using the explicit method. The convergence and performance characteristics of the LMS algorithm are determined by the relationship between the fixed gain and the input correlation statistics. The problem is less severe in the normalized LMS algorithm which removes the dependence on the input correlation statistics but limits the parameter value to a finite fixed interval for stable behavior. In this paper, we show, perhaps for the first time, the stability condition for the vector case of the LMS-type algorithm developed using the implicit method. Based on this, we present a time-variant step size LMS algorithm. Empirical studies demonstrate the superior convergence and steady-state behavior of the proposed filter.

KW - adaptive algorithm

KW - implicit method

KW - least mean square

KW - steady-state performance

KW - variable step size

UR - http://www.scopus.com/inward/record.url?scp=85208139463&partnerID=8YFLogxK

U2 - 10.1109/IBCAST59916.2023.10713033

DO - 10.1109/IBCAST59916.2023.10713033

M3 - Conference contribution

AN - SCOPUS:85208139463

T3 - Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023

SP - 136

EP - 141

BT - Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023

Y2 - 22 August 2023 through 25 August 2023

ER -

Pasha SA, Bin Saeed MO, Zerguine A. On the Time-Variant Step Size Posterior LMS Algorithm. In Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023. Institute of Electrical and Electronics Engineers Inc. 2023. p. 136-141. (Proceedings of 2023 20th International Bhurban Conference on Applied Sciences and Technology, IBCAST 2023). doi: 10.1109/IBCAST59916.2023.10713033

On the Time-Variant Step Size Posterior LMS Algorithm

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this