遗传算法群体规模第研究.pdfVIP

第35 卷 第1 期 计 算 机 工 程 2009 年1 月 Vol.35 No.1 Computer Engineering January 2009 ·人工智能及识别技术 · 文章编号：1000—3428(2009)01—0162—03 文献标识码：A 中图分类号：TP311 遗传算法群体规模的研究 1 1 2 黎 明 ，龙佳丽 ，盛伟翔 (1. 南昌航空大学无损检测技术教育部重点实验室，南昌 330063 ；2. 江西司法警官职业学院司法信息系，南昌 330013) 摘 要：遗传群体规模的选择是使用遗传算法优化计算时的首要问题，直接影响遗传算法全局收敛率和收敛速度等。该文研究二进制和自 然数编码遗传算法的群体规模，结合偏好函数和模式定理，利用前向及后向差分方程，得到这 2 种编码的群体规模下限值，证明其存在性。 通过对 2 个典型多模函数的优化测试，验证所得群体规模理论优化值的有效性。 关键词：遗传算法；群体规模；差分方程；函数优化 Study on Population Size of Genetic Algorithm 1 1 2 LI Ming , LONG Jia-li , SHENG Wei-xiang (1. Key Laboratory of Nondestructive Testing, Ministry of Education, Nanchang Hangkong University, Nanchang 330063; 2. Department of Law Information, Jiangxi Vocational College of Politics and Law, Nanchang 330013) 【Abstract 】The quality of the initial population size directly affects the performance and efficiency of the Genetic Algorithm(GA), and how to choose the initial population size “N ”is the important problem. The initial population size is studied based on the defined partial function, favor and back difference equations, and the schema theorem. The inner relations――two excellent inequations between the initial population size and the code length are obtained when binary and natural number codes are used in GA operations. Experimental results of this method on two classical complex multimodal functions show its validity and superiority. 【Key words 】Genetic Algorithm(GA); population size; difference equation; function optimization 1 概述 规模下限值，并证明