摘要
神经网络的结构冗余的原因的基础上,提出了一种利用粗集优化网络结构的原理与方法,并用实例证明,与现有的权消去法,灵敏度剪枝法,相关性剪枝法等方法相比,该方法不仅优化了网络的拓扑结构,而且加快了网络的收敛速度,从而增强了BP神经网络的适应能力。
The reason of redundancy of neural network＇s structure is analyzed and discussed. On this condition, it is tried to optimize the structure of neural network with rough sets, and the basic principle of optimized method is described concretely. Compared with weight elimination, sensitivity pruning algorithm, correlation pruning algorithm. The method＇s excellence is proved with an example, which shows that the method can be used to not only optimize the structure of neural network, but also quicken constringency rate, and it would improve the adaptive abilities of BP.
出处
《计算机工程与设计》
CSCD
北大核心
2007年第17期4210-4212,共3页
Computer Engineering and Design
关键词
神经网络
粗糙集
结构
优化设计
权消去法
灵敏度剪枝法
相关性剪枝法
neural network
rough sets
structure
optimization design
weight elimination algorithm
sensitivity pruning algorithm
correlation pruning algorithm
作者简介
朱万富（1974-），男，安徽淮南人，硕士研究生，讲师，研究方向为人工智能.E—mail：wfzhu2000@yahoo．com．cn
赵仕俊（1957-），男，山东东营人，博士，研究员，研究方向为人工智能及计算机网络。
