摘要
模式聚类问题可视为组合优化问题。针对传统遗传算法(TGA)在解决组合优化问题方面存在的不足,提出一种单亲遗传算法(PGA)。PGA不使用TGA常用的交叉算子,而是通过基因换位等遗传算子隐含交叉算子的功能来实现进化操作,简化了遗传过程。并且不要求初始群体具有广泛多样性,不存在"早熟收敛"问题,用PGA求解模式聚类问题可使聚类结果完全不依赖于初始聚类中心。仿真结果表明了这种算法的有效性。
Considering the deficiency of Traditional Genetic Algorithms (TGA) in solving combinatorial optimization, a Partheno-Genetic Algorithm (PGA) is proposed. PGA does not use crossover operators of TGA, while use gene exchange operators that have the same function as crossover operators. In PGA genetic operation is simpler and initial population need not be varied and there is not immature convergence. The problem of sensitivity with original clustering center is thoroughly solved by PGA. The perfect performance of PGA is demonstrated by the simulation examples.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
1999年第1期32-37,共6页
Pattern Recognition and Artificial Intelligence
基金
国家教委博士点基金
关键词
遗传算法
遗传算子
组合优化
模式聚类
Genetic Algorithm, Genetic Operator, Combinatorial Optimization, Pattern Clustering