基于多子群的社会群体优化算法

Multiple subgroups based social group optimization algorithm

社会群体优化(social group optimization,SGO)算法是一种基于社会群体学习而提出的一种新型优化算法。针对社会群体优化算法易于陷入局部最优问题,提出了一种多子群社会群体学习算法(MPSGO)。本算法采用多子群学习方法,对算法两个阶段的个体学习方法进行改进,在维持群体收敛性能的前提下提高群体多样性,同时对部分个体中引入量子学习,使个体学习的有用信息得以增强;此外,每隔一定代数对子群进行随机重组,既能保证各子群个体充分进化,又维持了子群多样性。在设计算法的基础上,分析了其收敛性和多样性;通过与其他四种算法进行对比实验,验证了改进后算法性能更优。

Social group optimization (SGO) is a novel optimization algorithm which is based on social group learning. This paper proposed a multi-group social group learning algorithm (MPSGO) to solve the problem that the social group optimization algorithm was easy convergent to local optima. This algorithm adopted the multi-subgroup learning method, and the improvement of the individual learning method in the two stages of the algorithm. It improved the diversity of the population with maintaining the convergence of the population. At the same time, it introduced quantum-behaved learning method into MPSGO for parts of individuals to enhance the useful information of individual learning. In addition, it randomly regrouped the population to generate new subgroups after a certain generation. It fully evolved that the diversity of subgroups was maintained and the individuals in every subgroup. Based on the designing of algorithm, this paper analyzed the algorithm of convergence and diversity. Compared with the other 4 algorithms, it proves that the performance of the improved algorithm is better.