WRBF网络的设计及其在混沌时间序列中的应用  

Design WRBF Neural Network and Its Application in Chaos Time Sequence Prediction

作　　者：陈得宝[1,2] 赵春霞[1] 

机构地区：[1]南京理工大学计算机系 [2]淮北煤炭师范学院物理系,安徽淮北235000

出　　处：《系统仿真学报》2005年第3期574-576,586,共4页Journal of System Simulation

基　　金：安徽省高等学校青年教师科研资助计划项目(2004jq180);教育部优秀青年教师资助计划项目(2001 年度)

摘　　要：结合小波网络良好的时频特性和 RBF 网络良好的局部特性,设计小波径向基函数网络(WRBF),该四层网络较三层小波网络和三层 RBF 网络有更优的特性。网络的第一隐层对输入样本进行小波映射,完成对输入信息的压缩,第二隐层实现径向基函数的非线性计算,克服了 RBF网络在处理多维样本时神经元的中心点和宽度难以确定及网络结果往往较复杂的弱点。为实现对网络结构和参数的同时优化,提出用二进制—复数混合编码的自适应进化规划,利用双倍体基因拓展染色体的信息量,加快算法的收敛速度,实现全局优化。在算法研究的基础上,利用 WRBF 网络对混沌时间序列进行预测,验证了方法的有效性。The Wavelet network has the best character in time domain and frequency domain, and RBF network has the best local character. With combination of their merits, a new WRBF (wavelet radial basis function network) network is designed. It has four layers cells. The input samples are compressed by the first hidden layer (wavelet layer) and the second layer operating with the compressed outputs of the first layer. So the larger number samples that should be used to determine the centers and the widths of the RBF's cells was relieved. In order to optimize structure and parameters of WRBF network, a new adaptive evolutionary programming with binary and complex-valued encoding method is proposed. The information is expanded at large scale with diploid genes, the convergent speed of the algorithm is fastened and the global optimization is realized. The effectiveness of the proposed method is proved by simulation with chaos time sequence based on the research of algorithm.

关 键 词：混合编码进化规划 WRBF网络 小波网络 RBF网络 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]