Stochastic analysis of diffusion dual sign algorithm for cyclostationary white Gaussian inputs and nonstationary system

This brief proposes a novel diffusion dual sign algorithm (DDSA) over networks to improve the convergence performance of the existing diffusion signed error algorithm (DSEA). The stochastic analysis of DDSA with the cyclostationary white Gaussian inputs is derived for the nonstationary system with random walk varying parameters in the presence of non-Gaussian impulsive noises. Simulation results illustrate the excellent agreement between the simulated learning curves and the predictions of derived analytical models, as well as the faster convergence speed compared with DSEA.