随机数发生器质量对简单遗传算法性能的影响  

Effects of Random Number Generator Quality on Simple GA Performances

作　　者：陈科[1] 龚子彬[1] 钟智攀[1] 

机构地区：[1]合肥工业大学机械与汽车工程学院,合肥230009

出　　处：《系统仿真学报》2009年第23期7473-7478,共6页Journal of System Simulation

摘　　要：伪随机数发生器(PRNG)对遗传算法(GA)性能的影响主要体现为低质量PRNG使得GA性能呈现不稳定性。提出平均海明距离方法来度量初始种群的多样性,建立理论模型分析短周期PRNG(一种典型低质量PRNG)周期、GA个体编码长度等因素对GA初始种群多样性的影响规律。仿真实验结果表明初始种群多样性对GA性能有显著影响;短周期PRNG通过影响初始种群多样性从而影响GA性能,且使得GA性能在不同条件下呈现出多种不稳定的变化状态。实验结果验证了理论模型的有效性。Unstable GA performances are often observed when a poor pseudo random number generator（PRNG） with a short period was used.An average hamming distance method was proposed for measuring the initial population diversity.Effects of period of PRNG and individual length on the initial population diversity of GA were analyzed.The simulation results show that GA performances are affected by initial population diversity significantly;the short-period PRNG affects the GA performance by affecting the initial population diversity,and unstable GA performances are due to different change patterns of initial population diversity.The experimental results also validate the effectiveness of the proposed theoretical model.

关 键 词：伪随机数发生器 遗传算法 海明距离 初始种群多样性 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]