摘要
提出了一种面向高维资源的分布式相似资源搜索机制.针对传统的分布式对等(P2P)网络无法解决高维资源的相似性搜索问题,通过基于主成分分析的降维算法将高维资源向量模型映射到低维空间,以低维空间中资源向量模型为索引,映射到P2P网络里的分布式散列表中,以一种完全基于P2P网络和路由机制的简单有效方式实现分布式相似性资源搜索,同时避免资源维数过高引发搜索的维数灾难.对降维处理后资源相似性信息保留情况进行了分析,并通过基于内容寻址网络的仿真验证了降维算法对于构建低维资源索引的有效性.对于具有一定聚类特征的高维资源,该方法可以在分布式的相似性搜索中获得较高的查准率.
A distributed semantic resources search mechanism for high-dimensional resources is presen- ted. Faced with the problem that the similarity search with high-dimensional resources couldn＇t be effec- tively achieved in traditional peer-to-peer （P2P） network, a high-dimensional resource vector model is mapped to the low dimensional space based on dimensionality reduction algorithm based on principal com- ponent analysis and then projected to distributed hash table in P2P network which is a simple and effec- tive way to achieve distributed similarity search. Meanwhile, the curse of dimensionality owing to the high dimension of resources could be prevented in the search. The maintenance of the similarity information af- ter processing of dimensionality reduction is analyzed. Simulation based on content addressable network is shown the effectiveness of low-dimensional index built by dimensionality reduction algorithm. The mecha- nism will achieve a high precision ratio in distributed similarity search for the clustered high-dimensional resources.
出处
《北京邮电大学学报》
EI
CAS
CSCD
北大核心
2013年第2期74-78,共5页
Journal of Beijing University of Posts and Telecommunications
基金
杭州华星--北邮信通院2011研究生创新基金
国家科技重大专项项目(2012ZX03005008)
关键词
向量模型
坐标空间
降维
资源搜索
对等网络
vector model
coordinate space
dimension reduction
resources search
peer-to-peer net-work
