典型相关分析综述  

Survey on canonical correlation analysis

作　　者：李有梅[1] 梁珣[1] 

机构地区：[1]中国计量大学理学院,浙江杭州310018

出　　处：《中国计量大学学报》2017年第1期113-118,共6页Journal of China University of Metrology

基　　金：国家自然科学基金资助项目(No.11301494)

摘　　要：数据之间的相关性分析是大数据处理的重要组成部分,典型相关分析及其扩展方法在多个领域得到了广泛应用.主要有用于解决多数据集特征融合的多集合典型相关分析,用于处理特征之间非线性关系的核典型相关分析,用于处理有类别特征数据时的判别典型相关分析,用于处理有噪声数据时的稀疏典型相关分析等扩展方法.本文全面综述了典型相关分析原理及其各种扩展方法,最后对这一方法的研究前景给出讨论和展望.Correlation analysis between data has become an important part of large data processing. The canonical correlation analysis method and its extensions have been widely used in various fields. The multiple canonical correlation analysis is used to solve the feature fusion for multi-data sets. The kernel canonical correlation analysis is used to find out the non-linear relationship between the data. The discriminant canonical correlation analysis is used to analyze the data which carry category information. The sparse canonical correlation analysis is used to solve the data with too many characteristics. In this paper, the principles of the canonical correlation analysis method and its various extensions are introduced. At the end of this paper, the prospects and outlook of the canonical correlation analysis are discussed.

关 键 词：典型相关分析 多变量特征融合 广义特征值问题 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] O212.4[自动化与计算机技术—控制科学与工程]