说明：  线性判别分析法（LDA），LDA以提高样本在子空间中的可分类为目标。寻找一组基向量，在这些基向量张成的子空间中，不同类别的训练样本能有最小的类内离散度，最大的类间离散度。
(Linear discriminant analysis (LDA), LDA in order to improve the sample in the sub-space can be classified as a target. Find a group-based vector, vector-based Zhang in these sub-space, different types of training samples can have the smallest dispersion category, the largest between-class dispersion.)