论文部分内容阅读
最小类方差支持向量机(MCVSVM)充分考虑数据的分布信息,但是在小样本情况下却仅利用类内散度矩阵非零空间中的信息.为了综合利用类内散度矩阵非零空间和零空间中的信息来进一步提高分类性能,文中首先在零空间中建立一种分类器——零空间分类器(NSC),然后再把MCVSVM和NSC进行融合,从而进一步提出集成分类器(EC).不同于MCVSVM和NSC,EC综合考虑非零空间和零空间中的信息,体现出更强的泛化能力.最后通过实验验证算法的有效性.
The minimum class variance support vector machine (MCVSVM) takes the distribution information of the data into full consideration, but only uses the information of the non-zero spatial divergence matrix in the case of small samples.In order to comprehensively utilize the non-zero space and zero Space to further improve the classification performance. Firstly, a classifier named zero space classifier (NSC) is built in zero space, and then the MCVSVM and NSC are merged to further propose an integrated classifier (EC). Different from MCVSVM and NSC, EC considers the information in non-zero space and zero space, which shows more generalization ability.Finally, the validity of the algorithm is verified through experiments.