Steady-State Analysis of the Deficient Length Incremental LMS Adaptive Networks.

The recently proposed distributed incremental least mean-square (DILMS) adaptive networks assume that the length of the adaptive filter at each node is equal to that of the unknown parameter; in other words, a sufficient length adaptive filter is assumed for each node. However, in many practical situations, the length of the employed adaptive filter at each node is less than that of the unknown parameter. In other words, at each node, a deficient length adaptive filter is employed. Since the analysis results for the sufficient length DILMS algorithm are not necessarily applicable to the deficient length case, so in this paper, we extend existing analysis to study the performance of the DILMS algorithm in this realistic case. More precisely, we derive a closed-form expression for the mean-square deviation (MSD) to explain the steady-state performance at each individual node. Simulation results support the theoretical analysis. [ABSTRACT FROM AUTHOR]