摘要
提出一种多目标遗传算法 ,将均匀设计技术应用于适应度函数合成和交叉算子构造 ,以提高遗传算法的空间搜索均匀性、子代质量和运算效率 .分析和实验结果表明 ,该方法可缩短算法运行时间和得到分布较均匀的Pareto有效解集 ;配合基于元件标称值的网表级高效编码方案和考虑基因位差异的遗传概率调整策略 ,可实现模拟电路自动设计 ,通过单次运行即获得对应不同偏好的多种实用化设计结果 .
We propose a novel multi-objective genetic algorithm based on the Uniform Design Techniques (UDT), which features a fitness function construction approach using the UDT to obtain a set of uniformly scattered search directions toward the Pareto frontier, and a multi-parents crossover operator using the UDT to improve the quality of offspring and decrease the computation cost. It is proved by experimental results that the method is capable of bringing out more uniformly scattered Pareto optimal solutions within a shorter execution time, and that when combined with an efficient representation scheme of circuits based on standard industrial values of components and a genetic parameters adaptation technique which takes into account loci' different effects and tracks the development of evolution and individuals diversity, it can be expected to realize automated design of analog circuits and to provide a set of effective results via a single execution.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2004年第10期1723-1725,1729,共4页
Acta Electronica Sinica
基金
国家自然科学基金 (No .60 1 330 1 0
No .60 3740 63)
