|
一种基于信息熵的神经网络规则提取方法
A method for rule extraction from trained neural networks based on information netropy
查看参考文献11篇
文摘
|
从训练后的神经网络中提取规则已成为当前研究热点。已有的网络规则提取方法常需网络修剪和再训练过程,因而计算成本较高。本文提出一种基于信息熵的神经网络规则提取方法,它在网络无需重复训练的情况下能够从训练过的神经网络中快速提取规则。其算法主要有四个过程组成:网络训练、决策树构建和相关隐单元识别、相关输入连接的识别及规则产生。文章以异或问题和棉花病害诊断规则提取为例进行实验,结果表明,基于信息熵的神经网络规则提取方法是有效可行的。 |
其他语种文摘
|
That extracting rules from trained neural networks has become one of the hottest topics in the research field currently. the computational cost of existing methods is very high because of the need for pruning and retraining of networks. In this paper, a method for rule extraction from trained networks is proposed based on information entropy. It can speed up rule extraction without pruning and retraining. The algorithm includes four procedures: training the network, building a decision tree and identifying the relevant hidden units, identifying the relevant input connestions, and generating rules. XOR problem and cotton disease rule extraction are taken as examples to show that the proposed method is effective and efficient for extracting rules from trained networks. |
来源
|
模式识别与人工智能
,2002,15(2):246-252 【核心库】
|
关键词
|
神经网络
;
信息熵
;
决策树
;
规则提取
|
地址
|
1.
中国科学技术大学自动化系, 合肥, 230026
2.
中国科学院合肥智能机械研究所, 合肥, 230031
|
语种
|
中文 |
文献类型
|
研究性论文 |
ISSN
|
1003-6059 |
学科
|
自动化技术、计算机技术 |
基金
|
国家自然科学基金
;
中国科学院研究生科学与社会实践专项(创新研究类)项目
|
文献收藏号
|
CSCD:948631
|
参考文献 共
11
共1页
|
1.
Carpenter G A. Rule Extraction: From Neural Architecture to Symbolic Representation.
Connection Science,1995,7(1):3-28
|
被引
2
次
|
|
|
|
2.
Setiono R. Neurolinear: From Neural Networks to Oblique Decision Rules.
Neurocomputing,1997,17:1-24
|
被引
5
次
|
|
|
|
3.
Setiono R. Extracting M-of-N Rules from Trained Neural Networks.
IEEE Trans on Neural Networks,2000,11(2):512-519
|
被引
5
次
|
|
|
|
4.
Mitra S. Knowledge-Based Fuzzy MLP for Classification and Rule Generation.
IEEE Trans on Neural Networks,1997,8(6):1338-1349
|
被引
3
次
|
|
|
|
5.
Tsukimoto H. Extracting Rules from Trained Neural Networks.
IEEE Trans on Neural Networks,2000,11(2):377-389
|
被引
13
次
|
|
|
|
6.
Lu H J. NeuroRule: A Connectionist Approach to Data Mining.
Proc of the 21st International Conference on Very Large Data Bases,1995:478-489
|
被引
1
次
|
|
|
|
7.
熊范纶. KDD中规则提取的收敛网络方法及其应用.
软件学报,2000,11(12):1635-1641
|
被引
3
次
|
|
|
|
8.
Ooyen A V. Improving the Convergence of the Backpropagation Algorithm.
Neural Networks,1992,5(3):465-471
|
被引
21
次
|
|
|
|
9.
Hertz J.
Introduction to the Theory of Neural Computation,1991
|
被引
16
次
|
|
|
|
10.
Setiono R. A Neural Network construction algorithm Which Maximizes the Likelihood Function.
Connection Science,1995,7(2):147-166
|
被引
2
次
|
|
|
|
11.
Battiti R. First-and Second-Order Methods for Learning: Between Steepest Descent and Newton's Method.
Neural Computation,1992,4:141-166
|
被引
30
次
|
|
|
|
|
|