摘要
该文提出一种基于非均匀变异和多阶段扰动的粒子群优化算法,并对算法的搜索性能进行了一般性分析.首先,在算法执行的不同阶段利用对当前最优解施加大小不同的邻域扰动操作,很好地增加了群体多样性,提高了跳出局部陷阱的概率,同时加强了对当前最优解邻域内的精细搜索;其次,在粒子群优化算法中引入非均匀变异运算,并依据非均匀变异运算规律适应性地调整解向量的搜索步长.算法性能分析表明,本算法较好地兼顾了群体优化算法的多样性和精英学习强度之间的平衡问题.数值实验上,首先用12个经典测试函数,验证该文提出的几种新措施的有效性与互助性;其次,针对30维和50维的CEC2005测试函数集,所提算法NmP3PSO与经典算法wFIPS、CLPSO和OLPSO做了大量的仿真实验,结果表明该文提出的算法表现出富有竞争力的性能和稳定性.
A new Particle Swarm Optimization (PSO) algorithm is proposed based on non- uniform mutation and multiple stages perturbation. Its search mechanism is also analyzed. Firstly, multiple stages perturbation operation with different radii is executed at different stages of algorithm. It diversifies the particle population and increases the probability of escaping from local trap. It also enhances the fine search at the neighborhood of the current best solution. Secondly, non-uniform mutation operator is introduced into PSO and the proposed algorithm adaptively adjusts the step size of solution vectors with non-uniform mutation operation. The per- formance analysis indicates that the proposed algorithm deals well with the balance between popu- lation diversity and learning intension from elitists. Twelve classical benchmarks are firstly used to verify the validity and the cooperation of the proposed strategies. Then the comparisons with the state-of-the-art evolutionary algorithms (wFIPS, CLPSO and OLPSO) are made on the shifted and rotated benchmarks from CEC2005 with cases of 30 and 50 dimensions. Experimental results indicate the competitive performance and stability of the proposed algorithm.
出处
《计算机学报》
EI
CSCD
北大核心
2014年第9期2058-2070,共13页
Chinese Journal of Computers
基金
国家自然科学基金(61105127
61375066)资助
关键词
粒子群优化
非均匀变异
多阶段扰动
群体多样性
particle swarm algorithm
non-uniform mutation
multiple stages perturbation
population diversity