摘要
信息时代,数据信息日渐丰富。人们获取更多信息的同时,也给机器学习领域带来挑战,降维是避免"维数灾难"的有效手段。笔者主要对特征选择的降维方法进行总结,搜索策略和评价准则两个方面在特征子集形成的过程中很重要,从这两个方面对特征选择算法进行阐述并比较了它们的特点。最后结合实际应用对特征选择研究方向进行展望。
In the information age,the data and information are becoming more and more abundant.The increasing data make people get more information,but also bring challenges to machine learning.Dimensionality reduction technology can be used to effectively avoid the“Curse of Dimensionality”.The author gives detailed description about feature selection.In this paper,the feature selection algorithms are described from searching strategy and evaluation criterion,and their characteristics are compared.Finally,the direction for future research of feature selection is discussed based on its practical application.
作者
黄铉
Huang Xuan(Department of Microelectronics,Chengdu College of University of Electronic Science and Technology of China,Chengdu Sichuan 611731,China)
出处
《信息与电脑》
2017年第24期67-68,共2页
Information & Computer
关键词
降维
特征选择
模式识别
dimensionality reduction
feature selection
pattern recognition