使用Fisher 现行判别函数对给定的样本进行训练，对于两类，Fisher线性判别函数很好做，就是 w = Sw (m1 - m2), 其中：Sw为总类内离散度矩阵， m1, m2, 为两个模式类的均值-Fisher discriminant function using the current sample of a given training for two, Fisher linear discriminant function is well done, that is, w = Sw