一种催化粒子群算法及其性能分析

针对粒子群算法（PSO）在解决高维、多模复杂问题时容易陷入局部最优的问题，提出了一种新颖的混合算法—催化粒子群算法（CPSO）。在CPSO优化过程中，种群中的粒子始终保持其个体历史最优值pbests。CPSO种群更新由改造PSO、横向交叉以及垂直交叉三个搜索算子交替进行，其中，每个算子产生的中庸解均通过贪婪思想产生占优解pbests，并作为下一个算子的父代种群。在CPSO中，纵横交叉算法（CSO）作为PSO的加速催化剂，一方面通过横向交叉改善PSO的全局收敛性能，另一方面通过纵向交叉维持种群的多样性。对6个典型benchmark函数的仿真结果表明，相比其它主流PSO变体，CPSO在全局收敛能力和收敛速率方面具有明显优势。

A Catlytic Particle Swarm Optimization and Its Performance Analysis

To address the problems that the particle swarm optimization (PSO) algorithm is likely to be trapped into the local optima when solving the high-dimensional and multimodal optimization problems, this paper proposed a novel hybrid optimization algorithm called catalytic particle swarm optimization (CPSO). In the optimization process of CPSO, pbests represent each particle in the population, directly. And modified PSO, horizontal crossover and vertical crossover update the population of particles in CPSO, alternatively. Each operator reproduces the moderation solutions, and the moderation solutions will generate the dominant solution pbests through greed thoughts, then the pbests will act as the father population of next operator. As an evolutionary catalytic of PSO, on one hand, CSO enhances the global search ability of PSO by horizontal crossover, and on the other, maintaining the diversity through vertical crossover. Simulation results for six benchmark functions show that the proposed algorithm demonstrates obvious advantage over other state-of-art PSO variants in terms of global convergence capacity and convergence rate.