SVM 原理推导
机器学习就是找决策边界
1.have u ? if w * u + b 〉= 0 them is + 正样本（W*u =U的图影，b原点到边界的值）
if w * u >=c
if w * u +b <0 them is - 样本
2.yi(w * x +b) -1 >=0
yi(w * x + b) -1 =0
3. width =2/|w|
min 1/2 |w| sqr yi(w*x +b) -1 =0
4.key ida
L = 1/2 ||W||sqr - z[y]
kernel smo,qp,kkt
dataset = numpy.loadtxt('path',delimiter=',')
x = dateset[:,0:8]
y = dataset[:,8]
dt.finllna(mean_cols)
all_dt = isnull().sum().sum()