摘要
使用遗传算法来实现径向基概率神经网络 (RBPNN)的全结构遗传优化 ,包括优选网络第一隐层节点数和求取匹配的核函数控制参数 .提出了适用于RBPNN的染色体编码方式 ,不仅使得所选隐中心矢量充分体现了模式样本的空间分布特征 ,同时还能够获得隐中心矢量的最佳数目及匹配的核函数控制参数 .新构造的适应度函数能够有效地控制网络输出的误差精度 .实验结果表明 ,该算法有效地简化了RBPNN模型的结构 .
The genetic algorithm was used to optimize the full structure radial basis probabilistic neural networks(RBPNN), including selecting the hidden centers vectors of the first hidden layer and determining the matching controlling parameters of kernel function of RBPNN. The proposed genetic encoding method not only completely embodies the space distribution characterizes of pattern samples, but also simultaneously achieves the optimum number of the selected hidden centers vectors and the matching controlling parameters of the kernel function. The novelly constructed fitting function can efficiently control the error accuracy of the RBPNN output. The experimental results show that the algorithm effectivelfies simpliy the structure of PBPNN.
出处
《红外与毫米波学报》
SCIE
EI
CAS
CSCD
北大核心
2004年第2期113-118,共6页
Journal of Infrared and Millimeter Waves
基金
国家自然科学基金资助项目 (60 173 0 5 0 )
