Non‐parametric variable step‐size diffusion LMS algorithm over adaptive networks

Abstract

To address the conflicting requirement of fast convergence rate and low misadjustment, a new non‐parametric variable step‐size diffusion least mean squares (LMS) (NPVSSDifLMS) algorithm is presented. The non‐parametric variable step‐size scheme is obtained to cancel the influence of noise signals in the intermediate posteriori errors at every node. Simulation results show that this NPVSSDifLMS algorithm outperforms the DifLMS and VSSDifLMS algorithms.