基于主体的分布式聚类系统

Agent-based distributed clustering system

随着网络的发展，对分布式数据处理的需求日益提高．分布式聚类系统是典型的分布式数据挖据，是分布式数据检索的重要部分．在多主体环境（MAGE）的基础上，构架一个分布式的聚类系统，核心学习算法是生长分层自组织映射网络．MAGE系统为从分布式环境中信息的获取以及知识的共享提供平台．生长分层自组织映射（GHSOM）网络动态生长的特点适合分布环境下对分布数据总体信息的缺乏；它的分层网络结构能更好组织数据类别；学习得到的生长分层白组织映射网络也可以作为检索的接口．该系统可以很好地被用到分布式数据检索系统中去，例如分布式图像检索．

With the development of networks, the distributed data processing is increasingly becoming important. A distributed clustering system is a typical distributed data mining, and is an important part for distributed information retrieval. Based on the multi-agent environment （MAGE）, the distributed clustering system will be developed, which takes the growing hierarchical self-organizing map （GHSOM） as the main learning algorithm. The MAGE system offers a well platform for acquiring information and sharing knowledge in the distributed environment. The dynamic growing architecture of GHSOM well matches the deficiency condition of lacking the complete information in distributed environment; the hierarchical architecture organizes the clusters of data better; also the resulting map can serve as retrieval interfaces. This distributed clustering system can be well utilized in distributed information retrieval systems such as a distributed image retrieval system.