Diffusion PSO-LMS Adaptation over Networks

A hybrid diffusion scheme, PSO-LMS, combining a particle swarm optimization (PSO) strategy and the least-mean square (LMS) algorithm, is developed here to optimize cost functions in a cooperative manner over networks, with a specific focus on channel estimation. The PSO component, through its built-in diffusion process, enhances the proposed scheme’s ability to quickly locate the region containing the global minimum, thus enabling the LMS to take over and exploit both the temporal and spatial diversity of the data so as to quickly hone in on the global minimum. Simulation results illustrate the improved performance of this hybrid scheme over both non-cooperative and other hybrid strategies.