Diffusion inverse square-root recursive least square-based algorithm with adaptive network

Abstract

In this paper, a Combine-Then-Adapt diffusion strategy on adaptive recursive least square algorithm (RLS) is proposed with adaptive network. We apply the inverse autocorrelation matrix based on the QR-decomposition method in order to reduce the complexity. The Combine-Then-Adapt diffusion strategy is modified on the inverse square-root recursive least square (iQR-RLS) algorithm including with the combination and adaptation steps. By applying an idea of diffusion strategy and mixed cost function, we derive a diffused version of an inverse square-root recursive least square algorithm, named as DiQR-RLS. Experimental results demonstrate that the proposed DiQR-RLS can achieve clearly the better performance than the conventional recursive least square algorithm.