针对粒子群算法和混合蛙跳算法在复杂函数寻优上易于陷入局部最优值的缺点,提出一种新的粒子群与混合蛙跳融合算法.算法采用多种群粒子群方法,每次进化后,将各子群中的最优粒子组成新的群体,采用混合蛙跳模式进化.以提高种群的多样性.粒子群各子群的进化模式中,除考虑本子群最好的粒子外,还考虑整合群体最好的粒子.相对于其它一些改进的粒子群或混合蛙跳算法,融合算法概念简单,易于实现,具有良好的全局搜索能力和较快的收敛速度.基准测试函数的仿真结果表明,本文算法优于目前一些常见的改进粒子群算法.
We propose a new hybrid algorithm, countering the shortcoming of particle Swarm optimization and shuffled frog leaping algorithm being easy to fall into local optimum in high-dimensional complex function optimization. The algorithm uses multiswarm particle swarm optimization, and after each evolution, groups the best particles in the sub-swarms into a frog group and uses shuffled frog leaping algorithm to evolve it. Inthe evolutionary model of each sub-swarm, in addition to considering the best particle of the sub- swarm, the best particle of the whole swarm is also considered. Compared with some improved particle swarm optimization or shuf- fled frog leaping algorithm, the hybrid algorithm is simple in concept, easy to implement, has a good global search capability and fas- ter convergence speed. The MPSO-SFLA has comprehensively been evaluated on 8 unimodal and multimodal benchmark functions. Results show that MPSO-SFLA is better then some of the common improved particle swarm optimization.