摘要
提出了一种基于竞争的自激励神经网络学习算法SIN ,该算法综合了自适应谐振理论和竞争型神经网络的特点 ,并在隐含层采用了Hebb学习规则 ,既能保证原有记忆不受影响 ,又能对新的信息加以记忆 ,同时又克服了ART网络对噪音敏感的缺点 ,学习速度快 ,分类性能好 ,具有在线学习的功能 将该算法应用于Web日志挖掘能够有效地剔除噪音 ,得到很好的用户聚类和页面聚类的结果 。
A novel neural classification algorithm named self inspiring network based on competition(SIN) is proposed in this paper It combines the advantages of adaptive resonance theory and the competitive neural networks Overcoming the sensibility to the noise which ART exists, it can either hold the former memory, or can memorize the new information Hebb learning rule is used in the implicit layer resulting in fast learning speed, good capability in classification and learning on line Web log mining used in this algorithm can omit the noise and get good result of the clusters in users and pages It also provides the effective decision making for the webmaster to devise the personalized web site
出处
《计算机研究与发展》
EI
CSCD
北大核心
2003年第5期661-667,共7页
Journal of Computer Research and Development
基金
教育部博士点科研基金项目 (2 0 0 10 3 3 5 0 49)
关键词
神经网络
聚类
WEB挖掘
无指导学习
neural networks
clustering
web mining
non supervised learning
