摘要
小波神经网络在处理具体问题时需要对其各个参数进行优化训练,才能成为一个要处理具体问题的小波神经网络,针对小波神经网络的参数优化训练问题,提出了免疫遗传算法,对比引入小生境技术的改进遗传算法,它利用了生物免疫系统的浓度抑制规律,有效的提高了种群的多样性,加忆了收敛速度,有更好的计算稳定性,同时提出了线性递减策略的遗传变异方式,使变异操作更加科学.将改进的免疫遗传算法用于小波神经网络的参数优化训练,通过具体的实现数据,证明了免疫遗传算法在解决小波神经网络的参数优化训练问题方面的优良效果.最后通过二级倒立摆仿真实验,验证了所设计的控制器优良的控制效果.
In order to deal with specific question by the wavelet neural network, it is necessary to optimize training the varios parameters. For optimizing the parameters, the improved immune genetic algorithm that introduces niche technology is presented. This improved immune genetic algorithm that use the rule of the concentration of the biological immune system suppression effectively improves the diversity of population, speeds up theconvergence speed and keeps better stability calculation. At the same time ,putting forward the strategies of thelinear decreasing genetic variation, it makes the mutation operation more scientific. Through the implementation of specific data, it proves the immune genetic algorithm in solving the parameters of wavelet neural network optimization has a good performance. At last, through the simulation results of double inverted pendulumexperiment verifies that the control- ler by designed has a good control effect.
出处
《哈尔滨理工大学学报》
CAS
北大核心
2015年第4期55-59,共5页
Journal of Harbin University of Science and Technology
关键词
小波变换
人工神经网络
小波神经网络
免疫遗传算法
二级倒立摆
the wavelet transform
artificial neural network
wavelet neural network
immune genetic algorithm
double inverted pendulum
