|Table of Contents|

A Genetic Design for Feedforward Neural Network(PDF)

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

Issue:
1999年06期
Page:
486-489
Research Field:
Publishing date:

Info

Title:
A Genetic Design for Feedforward Neural Network
Author(s):
LuJianfeng ShangShang YangJingyu
School of Information,NUST,Nanjing 210094
Keywords:
neural netw ork network topology alg orithm genet ic algorithm
PACS:
TP183
DOI:
-
Abstract:
During applicat ion of neural netw ork, there exist some problems, including diff-i cult determinat ion of the size and structure of neural netw ork in advance, the learning speed of neural network is slow , and it’ s easy to converge to local optimum. In the v iew of these problems, some references proposed to use Genet ic Algorithms ( GAs) to design and train neural network, the top structure and related parameters ( weig hts and thresholds) can be obtained s-i multaneously w ith this method. On this basis, this presentat ion made some improvement on proposed method. Main improvement includes that f loat-point matrix is adopted to encode, evolution of GAs itself is modif ied and the proposed method can sat isfy some constrain cond-i t ions. For feedforw ard neural network, this method can f ind suitable netw ork structure and corresponding parameter ( w eight s and thresholds) simultaneously under certain constrain condit ion. New method has g reat improvement over the old one in both accuracy and speed.

References:

1 Saha S, Chrisensen J P. Genetic desig n of sparse feedforw ar d neural netw ork1 Information Science, 1994, 79: 191~ 200
2 刘勇, 康立山, 陈毓屏.非数值并行算法—— 遗传算法. 北京: 科学出版社, 1995
3 金希东, 李治. 遗传- 灾变算法及其在神经网络和控制系统中的应用. 见: 靳蕃. 神经网络理论与应用研究. 成都: 西南交通大学出版社, 1996. 255~ 260

Memo

Memo:
-
Last Update: 2013-03-29