摘要
针对传统量子进化算法采用精英个体作为吸引子,存在种群学习范围窄、优秀基因易丢失的缺陷,提出了一种采用群体统计学习的量子进化算法.该算法抛弃了传统量子进化算法中的精英保留策略,通过截断、比例、竞赛选择等方式对进化过程中优秀群体统计分析后构建整个种群的吸引子,避免了以单一个体为单位的学习方式,能较为全面地从整个优秀种群学习知识,并保留群体的优秀基因信息.同时,吸引子每代更新,避免了采用精英保留策略易陷入局部极值的问题.通过测试实验表明,提出的算法搜索精度和效率提高,收敛速度更快,算法综合性能提高.
A statistical learning quantum-inspired evolutionary algorithm(SLQEA) is proposed to overcome the problem that the traditional quantum evolutionary algorithm has some inherent shortcomings such as the limited scope of learning and the easy-omission of genes during evolution.The SLQEA abandons the elite-retention strategies used in traditional algorithms.The attractor in the proposed algorithm is constituted of elite individuals who are selected from the population through methods such as proportion,truncation and tournament.Since the attractor covers the information of superior individuals of whole population,it can prevent the population from one individual and avoid premature convergence.Experiments show that SLQEA effectively improves search speed and accuracy,and that it is a highly scalable algorithm as well.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2012年第2期51-58,共8页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金资助项目(70971020)
关键词
量子进化
统计学习
基因信息
背包问题
组合优化
quantum evolutionary algorithm
statistical learning
genetic information
knapsack problem
combinatorial optimization