Scholarlink

<p>Based on the generalized maximum correntropy (GMC) criterion, the recursive GMC (RGMC) algorithm can achieve better robustness against impulsive noise. However, there is a trade-off between the tracking and the misadjustment in the recursive least squares framework with a fixed forgetting factor. To this end, the variable forgetting factor strategy can be an alternative at the expense of computational burden. In this paper, motivated by the data-reuse method, we can realize a computationally efficient data-reuse RGMC (DR-RGMC) algorithm to remedy the trade-off. Moreover, to improve the ability of DR-RGMC in sparse system identification, based on the different proportional update strategies, we also obtain two kinds of data-reuse proportional RGMC (DR-PRGMC) algorithms. In addition, we theoretically conduct the mean convergence analysis of DR-RGMC and the steady-state excess mean square error (EMSE) analysis of DR-PRGMC. Several simulation results for system identification demonstrate the effectiveness of our proposed algorithms.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Family of Data-Reuse-Based Recursive Generalized Maximum Correntropy Algorithm

  • 期刊论文

  • 发表日期：

    2025年6月9日

查看全文
  • Ji Zhao,
  • Biao Xie,
  • Qiang Li,
  • Hongbin Zhang

摘要

Based on the generalized maximum correntropy (GMC) criterion, the recursive GMC (RGMC) algorithm can achieve better robustness against impulsive noise. However, there is a trade-off between the tracking and the misadjustment in the recursive least squares framework with a fixed forgetting factor. To this end, the variable forgetting factor strategy can be an alternative at the expense of computational burden. In this paper, motivated by the data-reuse method, we can realize a computationally efficient data-reuse RGMC (DR-RGMC) algorithm to remedy the trade-off. Moreover, to improve the ability of DR-RGMC in sparse system identification, based on the different proportional update strategies, we also obtain two kinds of data-reuse proportional RGMC (DR-PRGMC) algorithms. In addition, we theoretically conduct the mean convergence analysis of DR-RGMC and the steady-state excess mean square error (EMSE) analysis of DR-PRGMC. Several simulation results for system identification demonstrate the effectiveness of our proposed algorithms.