Active target particle swarm optimization

Abstract

SUMMARYWe propose an active target particle swarm optimization (APSO). APSO uses a new three-target velocity updating formula, i.e. the best previous position, the global best position and a new target position (called active target). In this study, we distinguish APSO from EPSO (extended PSO)/PSOPC (PSO with passive congregation) by the different methods of getting the active target. Note that here EPSO and PSOPC are the two existing methods for using three-target velocity updating formula, and getting the third (active) target from the obtained positions by the swarm. The proposed APSO gets the active (third) target using complex method, where the active target does not belong to the existing positions. We find that the APSO has the advantages of jumping out of the local optimum and keeping diversity; however, it also has the disadvantages of adding some extra computation expenses. The experimental results show the competitive performance of APSO when compared with PSO, EPSO, and PSOPC.

Particle Swarm Optimization Methods, Taxonomy and Applications

“…Calculating the third term is complicated and it does not belong to the existing positions. This method maintains the diversity of the PSO as well as not trapping in the local optimum [103].…”

Neural Networks Applied in Chemistry. II. Neuro-Evolutionary Techniques in Process Modeling and Optimization

“…PSO offers several potential advantages in science: it is simple to understand and implement, is well-suited to parallel computation, does not require access to derivatives, scales well, and has few adjustable parameters, though this final characteristic does reduce flexibility. Variants of PSO have appeared, such as active target PSO, adaptive mutation PSO, adaptive PSO guided by acceleration information, angle modulated PSO, best rotation PSO, cooperatively coevolving particle swarms, modified genetic PSO, and others …”

“…PSO offers several potential advantages in science: it is simple to understand and implement, is well-suited to parallel computation, does not require access to derivatives, scales well, and has few adjustable parameters, though this final characteristic does reduce flexibility. Variants of PSO have appeared, such as active target PSO, 168 adaptive mutation PSO, 169 adaptive PSO guided by acceleration information, 170 angle modulated PSO, 171 best rotation PSO, 172 cooperatively coevolving particle swarms, 173 modified genetic PSO, 174 and others. 175 Ahmadi and Shadizadeh 49 used PSO methods to determine the network weights in a feed-forward network whose role was to predict the degree of asphaltene precipitation in oil samples; they obtained good agreement with experimental data, and argued that their results were superior to alternative methods of prediction.…”