摘要
提出了一种新型局部连接网络———K型局部连接网络。首先给出所提网络的结构与算法,然后对所提网络隐含层激活函数———K型函数的性质,以及网络的性能进行了理论上的分析。最后将K型网络应用到函数逼近以及建模上,通过与其他局部连接网络的性能对比验证K型网络所具有的优越性。
A new type of local connection neural network - K-type local connection neural network (shorted by KLCNN) is proposed in this paper. First the structure and algorithm of the proposed network is given. Then the K-type active function in the implication layer of KLCNN, and its properties are analyzed theoretically. The KLCNN is used in the function approximation and modeling. Compared with other local connection neural networks, K neural network is proved to have more advantages than others.
出处
《计算机应用》
CSCD
北大核心
2004年第2期44-46,共3页
journal of Computer Applications
基金
安徽省自然科学基金资助项目 (0 3 0 42 3 0 1 )
关键词
局部网络
K型网络
系统建模
函数逼近
local neural network
K neural network
system modeling
function approximation
