A new modeling algorithm: normalized Kernel least mean square

Authors:
Hamed Modaghegh;R. Hossein Khosravi;Saeed Ahoon Manesh;Hadi Sadoghi Yazdi
Affiliations:
Engineering Department, Ferdowsi University of Mashhad, Iran;Engineering Department, Ferdowsi University of Mashhad, Iran;Engineering Department, Ferdowsi University of Mashhad, Iran;Engineering Department, Ferdowsi University of Mashhad, Iran
Venue:
IIT'09 Proceedings of the 6th international conference on Innovations in information technology
Year:
2009

Citing 6
Cited 1

Adaptive filter theory (3rd ed.)

Adaptive filter theory (3rd ed.)
Kernel Methods for Pattern Analysis

Kernel Methods for Pattern Analysis
Support Vector Machines for Pattern Classification (Advances in Pattern Recognition)

Support Vector Machines for Pattern Classification (Advances in Pattern Recognition)
Kernel Based Algorithms for Mining Huge Data Sets: Supervised, Semi-supervised, and Unsupervised Learning (Studies in Computational Intelligence)

Kernel Based Algorithms for Mining Huge Data Sets: Supervised, Semi-supervised, and Unsupervised Learning (Studies in Computational Intelligence)
A New Robust Variable Step-Size NLMS Algorithm

IEEE Transactions on Signal Processing
Optimal Variable Step-Size NLMS Algorithms With Auxiliary Noise Power Scheduling for Feedforward Active Noise Control

IEEE Transactions on Audio, Speech, and Language Processing

Unsupervised kernel least mean square algorithm for solving ordinary differential equations

Neurocomputing

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper Normalized Kernel Least Mean Square (NKLMS) algorithm is presented which has applications in system modeling and pattern recognition. In 2007 a similar algorithm was proposed Named Kernel Least Mean Square (KLMS), and a modified version of KLMS was introduced in 2008. Although KLMS has good results in prediction of some time series, high sensitivity to step-size and signal amplitude stability, still remain as problems. In this paper NKLMS and its ability in prediction and identification of time series is presented and is compared to KLMS method. A variable named stepsize that was used in the algorithm has made NKLMS more efficient in prediction of time-series which have inconsistency in amplitude. Thus, convergence speed and system tracking are improved. Furthermore the proposed algorithm is applied to channel modeling.