摘要
分析了遗传多峰搜索领域内现有方法的不足 .对传统的遗传算法引入了梯度算子和聚类算子 ,将近似导数平方和的倒数作为评价函数 ,并定义了罚项 .用改进后的遗传算法搜索多峰 .实测结果表明 ,该算法搜索速度明显加快 ,精度有很大提高 .对等高等距。
This paper analysis the shortage of the existing methods of searching multi maximum GA. Basing on traditional genetic algorithms, it use GA to search multi maximum by introducing gradient calculator and merging calculator and using the inverse of the sum of square of differential coefficient as fitness function and defining the punish function to distinct other situation. The test results contrast with “simple subpopulation schemes” show that the SMMGA is efficient on all types function and its convergence speed is quicker and its convergence precision is higher than other method.
出处
《南开大学学报(自然科学版)》
CAS
CSCD
北大核心
2000年第3期17-22,共6页
Acta Scientiarum Naturalium Universitatis Nankaiensis
基金
天津自然科学基金!( 993 60 0 81 1 )资助项目
关键词
遗传算法
多峰搜索
梯度算子
聚类算子
极值点
genetic algorithms
optimization of multimodal function
gradient calculator
merge calculator