基于遗传算法和最速下降法函数优化混合数值算法.pdfVIP

1997 年 7 月 系统工程理论与实践 第 7 期 基于遗传算法和最速下降法的函数优化混合数值算法 赵明旺 (武汉冶金科技大学自动化系, 430081) 摘要　在遗传算法中嵌入一个最速下降算子, 并定义适当的适应度函数和子代个体的选择算子, 从而 可结合遗传算法和最速下降法两者的长处, 得到既有较快收敛性, 又能以较大概率得到全局极值的新 的用于连续函数全局优化的混合数值算法。数值计算结果表明了本文方法显著优于求解函数优化的 遗传算法和最速下降法。 关键词　遗传算法　最速下降法　函数优化　适应度 A H yb rid N um erical A lgo rithm fo r Function Op tim ization Based on Genetic A lgo rithm and Steepest D ecen t A lgo rithm Zhao M ingw ang ( . , , 430081) D ep t of A utom ation W uhan Yejin U niversity of Science T echno logy Abstract　In th is paper, th rough a steepest decent operato r is em bedded into the genet ic algo rithm and a p roper fitness function and a selecting operato r fo r son generation are defined, a hybrid algo rithm fo r global op tim ization of continuous function, com bined the advances of bo th of genetic algo rithm and steepest decent algo rithm , is go t w ith fast convergence and great p robability fo r global op tim ization. T he num erical computing re sults show n that the m ethod is distinctly superio r to the genetic algo rithm and steepest decent algo rithm. Keywords　genetic algo rithm ; steepest decent algo rithm ; function op tim ization ; fit ness 1　引言 ( ) 对连续可微函数的优化问题, 传统数值优化方法 如最速下降法、 法和共轭梯度法等 有相对 N