|Table of Contents|

Design and Application of WRBF Neural Network Based on Improved GA

《南京理工大学学报》（自然科学版）[ISSN:1005-9830/CN:32-1397/N]

Issue:
2007年03期
Page:
370-374
Research Field:
Publishing date:
2007-06-30

Info

Title:
Design and Application of WRBF Neural Network Based on Improved GA
Author(s):
CHEN De-bao1; 2; ZHAO Chun-xia1
1.School of Computer Science and Technology,NUST,Nanjing 210094,China;2.Physical Department,Huaibei Coal Industry Teachers College,Huaibei 235000,China
Keywords:
w ave let netwo rk;  rad ial basis function netw ork;  w avelet radial basis function network;  genet ic algo rithm
PACS:
TP183
DOI:
-
Abstract:
In o rder to overcome the demanding samples of design ing RBF ( rad ial basis funct ion ne-t w ork) neural netw ork and w avelet netw ork automatically and the overlap phenomenon of linear summat ion in output layer, a new four-layer netw ork namedWRBF ( w ave let radia l basis function ne-t w ork) is designed based on the simple structures of RBF and w avelet netw ork. In the netw ork, the first h idden layer is used to compress input space and the second h idden layer is to process the compressed variables. In training m ethods, a h igh-rank chromosome for indiv idual in improved GA is used to realize selecting connection betw een tw o h idden layers. The structure and parameters o f WRBF netw ork are optim ized simu ltaneously. The experimental results from M ISO system and coeff-i cient model o f therma lm eter demonstrate that the improved GA reduces the premature phenomenon and that theWRBF netw ork has h igher prec ision formode ling.

References:

[ 1] 陈天平. 神经网络及其在系统识别应用中的逼近问 题[ J]. 中国科学( A辑), 1994, 37( 4): 414- 421.
[ 2] 韩敏, 席剑辉, 程磊. RBPN 在模式识别中的应用 [ J] . 控制理论与应用, 2002, 19( 6): 940- 944.
[ 3] Chester D. W hy two h idden layers are better than one [ A]. IEEE In ternational Jo in t Conference on Neu ra l Ne tw ork [ C]. W ashington, DC: IEEE pub lications, 1990. 265- 268.
[ 4] M ary J, Green G R. Approx im ation theo ry and fin ite prec ision com putation [ J ]. IEEE T rans On Neu ra l Ne tw orks, 1995, 8 ( 1): 31- 37.
[ 5] 王晓兰, 陈得宝, 邱熔胜, 等. 进化规划在挖掘最优 语言模糊规则基中的应用[ J]. 兰州大学学报, 2001, 37( 2): 71- 76.
[ 6] 陈长征, 王楠. 遗传算法中交叉和变异概率选择的 自适应方法及作用机理[ J] . 控制理论与应用, 2002, 19( 1): 41- 43. [ 7 ] 陈得宝, 刘暾东, 魏辉. 基于三阶染色体结构的自 适应进化规划及应用[ J]. 系统仿真学报, 2004, 16 ( 2): 322- 325.

Memo

Memo:
-
Last Update: 2007-06-30