Diffusion minimum Wilcoxon affine projection algorithm over distributed networks

Abstract The least squares based affine projection algorithm (APA) is sensitive to outlier/impulsive noise in the desired data. A novel robust APA over distributed networks scenario is proposed in this manuscript, which is based on the rank based robust estimator named Wilcoxon norm. The proposed diffusion Wilcoxon affine projection algorithm (DWx-APA) based on pseudo least squares formulation is robust against outliers in the desired data and converges faster than the diffusion minimum Wilcoxon norm algorithm (DWx). The QR based diffusion minimum Wilcoxon norm (QR-DWx) algorithm is also robust against outliers in the desired data and converges faster than the DWx and proposed DWx-APA, but the computational complexity is very high in comparison to its counterparts. The proposed DWx-APA is a compromise between DWx and QR-DWx in terms of computational complexity and convergence speed. The mean stability, tracking capability and computational complexity of the proposed algorithm are investigated. The simulation based experiments and analysis validate that the proposed algorithm performs better than the state-of-the-art algorithms in the presence of color data.