摘要
结构选择是神经网络研究的热点,文章提出了一种基于相容粗糙集技术的ELM(Extreme LearningMachine)网络结构选择方法,给定一个含有很多隐含层结点的前馈神经网络。该方法用相容依赖度度量隐含层结点的重要性,将不重要的隐含层结点逐一去掉,直到满足预定义的终止条件为止。实验结果表明,该文提出的方法是行之有效的。
Architecture selection is hot topic in the study of neural networks. Based on tolerance rough set technique, a method of architecture selection for ELM network is proposed in this paper. Given a single layered forward neural network with many hidden nodes, the method employs the concept of tolerance dependency to measure the significance of hidden nodes and keeps removing the nodes which are not of great importance until the predefined termination condition is satisfied. The experimental results show that the proposed method is effective.
出处
《合肥工业大学学报（自然科学版）》
CAS
CSCD
北大核心
2012年第12期1628-1632,共5页
Journal of Hefei University of Technology：Natural Science
基金
国家自然科学基金资助项目(61170040)
河北省自然科学基金资助项目(F2010000323)
河北省高校科技重点基金资助项目(ZD2010139)
河北大学自然科学基金资助项目(2011-228043)
关键词
粗糙集
相容粗糙集
神经网络
网络结构选择
rough set
tolerance rough set
neural network
network architecture selection
