摘要
1 引言神经网络技术已经在很多领域得到了成功的应用.但由于神经网络并不具有一个统一的理论框架,其经验性成分相当高,这对其进一步发展造成了极大的阻碍。如果能为神经网络应用提供一些指导性的分析方法,不仅将促进该领域的理论研究。
; There is some relationship between the architecture and generalization of the neural network, and the size of the training set. As the kernel of PAC learning and the measure of the learning a-bility of learning machine,VC-dimension plays an important role in the relationship. So,the research of VC-dimension will highy profit neural network theory and technology. In this paper,based on the statistical learning theory,the up-to-the-minute achievements and research evolvement of the VC-dimension computation of neural networks are introduced,and some future research issues are also discussed.
出处
《计算机科学》
CSCD
北大核心
2000年第7期59-62,78,共5页
Computer Science
基金
国家自然科学基金(编号:69625103)
关键词
神经网络
VC维
计算
VC-dimension, Neural networks, Pattern recognition, Machine learning, Computational learning theory
