摘要
基于核技巧提出的新的非线性鉴别分析方法在最小二乘意义上与基于核的Fisher鉴别分析方法等效,相应鉴别方向通过一个线性方程组得出,计算代价较小,相应分类实现极其简便。该方法的最大优点是,对训练数据进行筛选,可使构造鉴别矢量的“显著”训练样本数大大低于总训练样本数,从而使得测试集的分类非常高效;同时,设计出专门的优化算法以加速“显著”训练样本的选取。实验表明,该方法不仅具有明显的效率上的优势。
The least squares solution of novel discriminant analysis method, based on Kernel trick, was equivalent to kernel-based Fisher discriminant analysis. The discriminant vector of the novel method was efficiently solved from linear equations. Moreover, corresponding classifying strategy was very simple. The most striking advantage of the novel method was that only a few original training samples were sorted as 'significant' nodes for constructing discriminant vector. As a result, corresponding testing was much efficient. In addition an appropriative, optimized algorithm was developed to improve the efficiency of selecting 'significant' nodes. Experiments show that the performance of the novel method is equal to kernel-based Fisher discriminant analysis, with superiority in efficiency.
出处
《解放军理工大学学报（自然科学版）》
EI
2005年第2期127-131,共5页
Journal of PLA University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目 ( 6 0 0 72 0 34 )
