Sparsity-Constrained Kernel Recursive Generalized Maximum Versoria Criterion Algorithm

Sandesh Jain

doi:10.1109/LSP.2023.3265604

2023-01-01

PRE

Sandesh Jain

Sudhan Majhi *

DOI：10.1109/LSP.2023.3265604

原文链接

原文求助

摘要

En 中文

This letter proposes a novel random Fourier features (RFF) based sparsity-aware Kernel recursive adaptive filtering algorithm which employs generalized maximum Versoria criterion (GMVC) as an adaptation cost and generalized Versoria zero attraction (GVZA) as a sparsifying-norm. The proposed RFF-based GVZA kernel recursive GMVC (RFF-GVZA-KRGMVC) algorithm is robust for nonlinear sparse channel estimation under non-Gaussian noise over both stationary and non-stationary environments. Simulations indicate that the proposed RFF-GVZA-KRGMVC approach delivers better convergence and bit error rate performance over the existing stochastic gradient based RFF-ZA-kernel-MVC (RFF-ZA-KMVC) algorithm at the cost of increased computational complexity. Furthermore, detailed convergence analysis is also performed for the proposed algorithm and corroborated by Monte-Carlo simulations.

Keyword:

Signal processing algorithms

Kernel

Convergence

Machine learning algorithms

Channel estimation

Computational complexity

Covariance matrices

Sparsity-constrained

RKHS

zero-attracting

maximum Versoria criterion