A particle swarm optimization-least mean squares algorithm for adaptive filtering

A particle swarm optimization-least mean squares (PSO-LMS) algorithm is presented for adapting various classes of filter structures. The LMS algorithm is widely accepted as the preeminent adaptive filtering algorithm because of its speed, efficiency and provably convergent local search capabilities. However, for multimodal error surfaces, a global search algorithm, such as PSO or the genetic algorithm (GA), is required. The proposed PSO-LMS hybrid algorithm combines the advantageous properties of the two conventional algorithms to provide enhanced performance characteristics.