Search

This paper presents a new algorithm, PCLPSO, based on particle swarm optimization, which uses comprehensive learning particle swarm optimizer. Our algorithm executes C parallel CLPSO algorithms. We adopted as a criterion of completion a maximum value of evaluations of the objective function. During the execution of the CLPSO algorithms, when a certain evaluation value of the functions is reached, the best k are selected, and different initialization criteria are applied to continue the execution of the CLPSO algorithms: restarting the worst ones for the best solution or restores the worst ones to a random solution. For this restart, we use the Boltzmann criterion in a similar way as Simulating Annealing (SA) does. In this work, the experimental results obtained for the search of the minimum of 16 multimodal test functions such as Rosenbrock, Griewank, Rastrigin, Brannin, Schwefel, and others. Our algorithm proved to be more efficient than the traditional CLPSO in its experimental results, and the nonparametric Wilcoxon test confirmed this. [ABSTRACT FROM AUTHOR]