九、SVM支持向量机
1. Model
其实从预测的模型上来说和感知机是一样的
z = w ⃗ ⋅ x ⃗ + b = w 1 x 1 + w 2 x 2 + . . . + w n x n + b y = s i g n ( z ) = { 1 , z ≥ 0 0 , z < 0 z = \vec w \cdot \vec x + b=w_1x_1+w_2x_2+...+w_nx_n+b\\ y=sign(z)= \begin{cases} 1, z\geq0\\ 0, z<0 \end{cases} z=w⋅x+b=w1x1+w2x2+...+wnxn+by=sign(z)={
1,z≥00,z<0
2. 损失函数
SVM的损失函数
J ( w ) = ∑ x i ∈ M [ y ( i ) ∗ c o s t 1 ( w ⃗ ⋅ x ⃗ ( i ) + b ) + ( 1 − y ( i ) ) ∗ c o s t 2 ( w ⃗ ⋅ x ⃗ ( i ) + b ) ] J(w) = \sum_{x_i\in{M}}{[y^{(i)}*cost_1( \vec w \cdot \vec x^{(i)}+b)+(1-y^{(i)})*cost_2(\vec w \cdot \vec x^{(i)}+b)]} J(w)=xi∈M∑[y(i)∗cost1(w⋅x(i)+b)+(1−y(i))∗cost2(w⋅x(i)+b)]
令 z = w ⃗ ⋅ x ⃗ ( i ) + b z=\vec w \cdot \vec x^{(i)}+b z
最低0.47元/天 解锁文章
591
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
