An estimation of distribution improved particle swarm optimization algorithm

PSO is a powerful evolutionary algorithm used for finding global solution to a multidimensional problem. Particles in PSO tend to re-explore already visited bad solution regions of search space because they do not learn as a whole. This is avoided by restricting particles into promising regions through probabilistic modeling of the archive of best solutions. This paper presents hybrids of estimation of distribution algorithm and two PSO variants. These algorithms are tested on benchmark functions having high dimensionalities. Results indicate that the methods strengthen the global optimization abilities of PSO and therefore, serve as attractive choices to determine solutions to optimization problems in areas including sensor networks.