Kernel Recursive Maximum Versoria Criterion Algorithm Using Random Fourier Features

Reproducing Hilbert space (RKHS)-based adaptive algorithms have attracted increased attention in machine learning and nonlinear signal processing with applications in visible light communications, radar, radio frequency communications and others. However, performance of RKHS-based algorithms is high...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on circuits and systems. II, Express briefs Jg. 68; H. 7; S. 2725 - 2729
Hauptverfasser: Jain, Sandesh, Mitra, Rangeet, Bhatia, Vimal
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York IEEE 01.07.2021
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Schlagworte:
Adaptive algorithms
Adaptive filters
Algorithms
Computational complexity
Convergence
correntropy
Criteria
Hilbert space
Kernel
Kernels
Machine learning
minimum mean square error
Prediction algorithms
RFF
RKHS
Signal processing
Signal processing algorithms
Steady-state
Versoria criterion
Visible light communication
ISSN:1549-7747, 1558-3791
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Reproducing Hilbert space (RKHS)-based adaptive algorithms have attracted increased attention in machine learning and nonlinear signal processing with applications in visible light communications, radar, radio frequency communications and others. However, performance of RKHS-based algorithms is highly sensitive to a suitable learning criterion. In this regard, the Versoria criterion-based adaptive filtering has gained interest in recent works due to its superior convergence characteristics as compared to the popular criterion such as minimum mean square error, and maximum correntropy criterion. Therefore, in this brief, a novel random Fourier feature (RFF)-based kernel recursive maximum Versoria criterion (KRMVC) algorithm is proposed. Convergence analysis is presented next for the proposed RFF-KRMVC algorithm as guarantees of the promised performance benefits. Lastly, the analytical results are validated by corresponding computer-simulations over practical application-scenarios considered in this brief.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1549-7747
1558-3791
DOI:10.1109/TCSII.2021.3056729