摘要
基于Fisher的投影降维的思想,讨论了线性分类器设计问题,提出最优分类超平面,克服了Fisher等线性分类器的偏向性与误分类现象.利用优化技巧把确定最优分类超平面问题转化为求解半正定的二次规划.随机数据模拟实验表明文中算法对设计高性能线性分类器是有效的.
The linear discrimination of pattern classification is discussed. Based on Fishers ideal of reducing dimensionality by projecting feature vectors onto a straight line, the optimal separating hyperplane is presented. And then a semipositive quadratic programming is offered to find the optimal separating hyperplane. The deflection and misclassificaition of the Fishers Linear Discriminant is improved. The effectiveness of the method is shown by random data experiments.
出处
《西安电子科技大学学报》
EI
CAS
CSCD
北大核心
2002年第6期791-795,共5页
Journal of Xidian University
基金
十五"国家部委科技(电子)预研资助项目(413160501)
关键词
模式识别
FISHER线性判别
最优分类超平面
二次规划
pattern classification
Fisher's linear discriminant
optimal separating hyperplane
quadratic programming
