摘要
在对经典GMDH算法和几种改进的GMDH算法论述的基础上,研究了其理论基础——K-G多项式和参考函数所带来的大量冗余变量项问题;提出二重GMDH算法来简化最终模型,剔除冗余的变量项,以得到更简洁的、健壮的模型。
Based on the classical GMDH algorithm and other several improved GMDH algorithms, the paper studies the GMDH＇s theory, especial the K - G polynomial and the transfer function which induce the problems of mass and redundant polynomial variables in the output model. An improved GMDH algorithm which runs GMDH algorithm twice is given, and be named Bi - GMDH algorithm. Features of this algorithm are analysed. It shows that the Bi -GMDH algorithm is more compact and robust than original algorithms.
出处
《软科学》
CSSCI
2008年第4期17-20,共4页
Soft Science
基金
国家自然科学基金资助项目(70271073)
