改进的粒子群算法及在数值函数优化中应用

为提高粒子群算法的优化能力,提出了一种改进的粒子群优化算法。在该算法中,采用Beta分布初始化种群,采用逆不完全伽马函数更新惯性权重,在速度更新式中,引入了基于差分进化的新算子,对于粒子的越界处理,采用了基于边界对称映射的新方法。以50个不同类型的数值函数作为优化实例,基于威尔柯克斯符号秩检验的测试结果表明,该算法明显优于普通粒子群优化算法、差分进化算法、人工蜂群优化算法和量子行为粒子群算法。

Application of improved particle swarm optimization to numerical function optimization

To enhance the optimization ability of the particle swarm optimization (PSO), an improved PSO algorithm was proposed in this paper. In the proposed approach, the Beta distribution function is used to initialize population, and the inverse incomplete gamma function is used to update the inertia weight. For adjustment of velocity, a new operator based on differential evolution is introduced. For cross-border processing of particles, a new method based on boundary symmetry mapping is designed. With taking 50 different types of benchmark functions as optimization examples, the experimental results based on the Wilcoxon-Signed rank test show that the proposed algorithm is obviously superior to the common PSO, differential evolution, attificial bee colony algorithm and quantum-behaved particle swarm optimization algorithm.