自适应小世界粒子群优化算法

Adaptive small-world particle swarm optimization

提出一种基于小世界网络的自适应拓扑结构。每个粒子都与它的近邻粒子进行交互,其有一定概率通过小世界重置与远方的粒子进行沟通;为粒子群体的每个维度分配一个特定的小世界网络,不同维能够学习不同邻居的历史信息;粒子的邻域大小与小世界重置的概率将在种群收敛状态的基础上进行自适应调整。利用标准函数集对该算法进行测试,测试结果表明,通过该机制,粒子群体具有更好的搜索多样性,能够平衡全局探索与局部开发。

An adaptive small‐world topology was developed .Each particle interacted with its cohesive neighbors frequently and communicated to some distant particles via small‐world randomization with certain probability .Each dimension of the particle swarm was assigned with a specific small‐world network ,so that a particle learnt from the historical information from different neighbors on different dimensions .Moreover ,the neighborhood size and the probability of small‐world randomization were adap‐ted automatically according to the convergence stage of the swarm .Results of experiments performed on a benchmark test set show ,by adopting such topology ,the particle swarm not only gains better search diversity ,but also balances the global explora‐tion and local exploitation .