دانلود رایگان مقاله بهینه سازی ازدحام ذرات با توپولوژی مسابقه پویا

abstract

Particle swarm optimization (PSO) is a nature-inspired global optimization method that uses interaction between particles to find the optimal solution in a complex search space. The swarm’s evolving solution is represented by the best solution found by any particle. However, using this best solution often limits the search area. In this paper, we propose a dynamic tournament topology strategy to improve PSO. In our method, each particle is guided by several better solutions, chosen from the entire population. The selection of the better particles is stochastic, but still favors particles with better solutions. Experimental results on benchmark functions indicate that the proposed method is promising. Furthermore, the application of our dynamic tournament topology strategy in optimization of artificial neural networks indicates that this method has favorable performance.

6. Conclusions

A dynamic tournament topology strategy to improve PSO has been proposed in this study. The strategy generalizes the definition of the best position and ensures a wide exchange of information. Instead of focussing on the historical best solution, the proposed strategy chooses a number of guides selected randomly, but with preference going to particles with higher fitness. Moreover, each particle can be guided by a different set of better particles, encouraging the sharing of information across a wide network. Compared with traditional PSO, this strategy preserves the diversity of the population, and enables the swarm to find a better solution, avoiding falling into a local optimum. Its effectiveness is especially pronounced when solving complex problems. Sixteen well-known benchmark functions are used to evaluate the performance of the proposed approach. We also compared its performance against a host of different algorithms. Experimental results illustrate that the proposed dynamic tournament topology strategy demonstrates favorable performance, especially in terms of accuracy and convergence speed. Furthermore, as far as the time resources are concerned, although the introduction of this strategy increases time complexity of the optimization, the relative influence of dynamic tournament topology on computation time decreases gradually while its accuracy increases. Furthermore, the DTT-PSO is applied to the optimization of a neural network for a simple robot foraging task. DTT-PSO demonstrates a better performance among all compared methods on this task and shows that DTT-PSO is able to solve real problems. For future work, more real-world applications from other fields will help to further investigate the effectiveness of DTT-PSO. In order to improve the performance of robot foraging task, the self organization neural network will be used instead of feedforward neural network. Furthermore, we plan to investigate an adaptive method for selecting and changing user defined parameters to improve the method’s stability.