对比这么复杂的推导过程，SVM的思想确实那么简单。它不再像logistic回归一样企图去拟合样本点（中间加了一层sigmoid函数变换），而是就在样本中去找分隔线，为了评判哪条分界线更好，引入了几何间隔最大化的目标。-The derivation process of the contrast is so complex, the idea of SVM is so simple. It is no longer tried to like logistic regression fitting