Asynchronous particle swarm optimizer with relearning strategy

Relearning strategy is a commonly used method to improve human memory or skills. In this work, relearning strategy is adopted in asynchronous particle swarm optimizer (PSO) to enhance its convergence. Although asynchronous PSO converges faster than synchronous PSO in most cases, it cannot guarantee a high successful rate of reproduction of better offspring in each generation. When a particle cannot search a better personal best position, the relearning strategy is utilized to enforce the particle learn again according to the original PSO formula. Moreover, a new mutation operator called Gaussian hypermutation is proposed to maintain the population diversity. Simulation results based on nine benchmark functions show that relearning strategy significantly improves the performance of asynchronous PSO.