【作者】 李世峰； 黄磊； 刘昌平；

【Author】 LI Shi-feng, HUANG Lei, LIU Chang-ping Institute of Automation, Chinese Academy of Sciences

【机构】 中国科学院自动化研究所；

【摘要】 聚类已经不仅仅限于经典的模式识别领域,而广泛应用于统计理论、机器学习和数据挖掘。除了提取样本的统计分布特征作为识别的有力手段,聚类在高维、大规模数据库的数据分析方面也发挥着越来越重要的作用,推动了相关理论的快速发展。本文介绍了聚类中常用的若干方法及其改进,提出一种新的聚类方法,并应用于脱机手写数字的识别,以期通过比较反映这些方法的某些特性。更多还原

【Abstract】 In addition to application to classical pattern recognition domain,clustering has been widely used in statistical theory, machine learning and data mining. As well as a powerful tool in recognition by extracting statistical features lying in input patterns, clustering is playing a more and more important role in data analysis of high-dimensional, large-scale databases, thus promoting a rapid theoretic development. The paper summarizes some commonly used algorithms and new improvement in clustering, including one which is advanced for the first time by us, which are implemented in off-line hand-written digits recognition, to evaluate the property.更多还原