CAN FUZZY NEURAL NETS APPROXIMATE CONTINUOUS FUZZY FUNCTIONS

被引:76
作者
BUCKLEY, JJ [1 ]
HAYASHI, Y [1 ]
机构
[1] IBARAKI UNIV,DEPT COMP & INFORMAT SCI,HITACHI,IBARAKI 316,JAPAN
关键词
FUZZY NEURAL NETWORKS; FUZZY FUNCTIONS; UNIVERSAL APPROXIMATOR; HYBRID FUZZY NEURAL NETWORKS;
D O I
10.1016/0165-0114(94)90283-6
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
We first argue that regular fuzzy neural nets, based on standard fuzzy arithmetic and the extension principle, cannot be universal approximators. This negative result is surprising since (regular) neural nets are universal approximators. We then argue that hybrid fuzzy neural nets, not necessarily based on standard fuzzy arithmetic, can be universal approximators.
引用
收藏
页码:43 / 51
页数:9
相关论文
共 42 条