So they have better performance than adaptive Gaussian kernel. 因而，相对于自适应高斯核函数而言，两者性能更优。
As for the undivided linear sample space, the kernel function is needed to map onto another high dimension linear space. 对于线性不可分的样本空间，需要寻找核函数，将线性不可分的样本集映射到另一个高维线性空间。
