Diffusion adaptive networks with changing topologies

Adaptive networks (AN) have been recently proposed to address distributed estimation problems [1]-[4]. Here we extend prior work to changing topologies and data-normalized algorithms. The resulting framework may also treat signals with general distributions, rather than Gaussian, provided that certain data statistical moments are known. A byproduct of this formulation is a probabilistic diffusion adaptive network: a simpler yet robust variant of the standard diffusion algorithm [2].