摘要
本文讨论了二阶神经网络的映射能力．主要内容包括：（１）从理论上严格地证明了二阶神经网络能以任意精度逼近任意连续函数．（２）给出了二阶神经网络的ＢＰ算法．（３）模拟实验结果．模拟实验表明：在中间隐层单元数目相同的条件下，二阶神经网络的误差函数比一阶神经网络下降得快；在误差精度相同的条件下。
The approximation capability of the second-order neural network is investigated in this paper and the following results have been obtained: 1)It has been prove4 that the second-order neural network can approximate any continuous function with any degree of accuracyl 2)the BP algorithm for se cond-order neural network and the simulated results are given in this paper. The simulated experiments show. that when the number of hidden neurons in the second-order neural network is equal to the first-order one's, the error of the second-order neural network decreases faster than the first-order one's; when the accuracies of both'of the second-order and first-order neural networks are equal, the number of hidden neurons in the second-order is far smaller than the first-order one's.
出处
《控制理论与应用》
EI
CAS
CSCD
北大核心
1996年第4期516-520,共5页
Control Theory & Applications
关键词
二阶神经网络
BP算法
映射能力
神经网络
second-order neural network
approximation
BP algorithm for the second-order neural network
