摘要
可分离性判据在统计模式识别的特征提取和聚类分析中有重要应用。论文定义了基于类距离的可分离性判据,发现可比基于类内类间距离的可分离性判据更好地反映类别分离性,也有更好的几何意义,并将它用于特征提取。
The separative criterion is important to the feature extraction and clustering analysis in statistical pattern recognition.In this paper,a separative criterion based on class distance is given.It can reflect the separability better than the criterion based on within-class and among-class distance,and has better geometric meaning.Its application of feature extra IS ALSO GIWEN
出处
《计算机工程与应用》
CSCD
北大核心
2003年第26期97-99,共3页
Computer Engineering and Applications
基金
四川省科技厅应用基础研究项目资助(编号:01SY051-09)
关键词
类距离
可分离性
特征提取
Class distance Separability Feature extaction
