PERFORMANCE AND GENERALIZATION OF THE CLASSIFICATION FIGURE OF MERIT CRITERION FUNCTION

被引：11
作者：
BARNARD, E
机构：
[1] Department of Electronics and Computer Engineering, University of Pretoria, Pretoria
来源：
IEEE TRANSACTIONS ON NEURAL NETWORKS | 1991年 / 2卷 / 02期
关键词：
D O I：
10.1109/72.80345
中图分类号：
TP18 [人工智能理论];
学科分类号：
081104 ; 0812 ; 0835 ; 1405 ;
摘要：
A new criterion function for training neural networks, introduced by Hampshire and Waibel [1], is studied. It is shown that this criterion function has some highly desirable properties. However, these properties are not directly related to generalization (as was suggested in [1]; it is shown that systematic improvement of generalization involves a different class of modifications.
引用
收藏
页码：322 / 325
页数：4
相关论文
共 7 条
[1]   A COMPARISON BETWEEN CRITERION FUNCTIONS FOR LINEAR CLASSIFIERS, WITH AN APPLICATION TO NEURAL NETS [J].
BARNARD, E ;
CASASENT, D .
IEEE TRANSACTIONS ON SYSTEMS MAN AND CYBERNETICS, 1989, 19 (05) :1030-1041
[2]  
CHAUVIN Y, 1990, ADV NEURAL INFORMATI, V2, P642
[3]  
Duda R. O., 1973, PATTERN CLASSIFICATI, V3
[4]  
Hampshire J B, 1990, IEEE Trans Neural Netw, V1, P216, DOI 10.1109/72.80233
[5]  
HINTON GE, 1986, PARALLEL DISTRIBUTED, V1, P318
[6]  
STINCHCOMBE M, 1989, JUN P IEEE INT C NEU, pI613
[7]   CONSTRUCTING A GENERALIZER SUPERIOR TO NETTALK VIA A MATHEMATICAL-THEORY OF GENERALIZATION [J].
WOLPERT, DH .
NEURAL NETWORKS, 1990, 3 (04) :445-452
1