学习日记-机器学习4-多特征/向量化

8 Multiple Features 多特征

$x_j=j^{th}$ feature
$n=$ number of features
${\vec{x}}^{(i)}=$ features of $i^{th}$ training example
$x_j^{(i)}=$ value of feature $j$ in $i^{th}$ training examples

Model:
Previously: $f_{w,b}(x)=wx+b$
now: $f_{w,b}(x)=w_1{x}_1+w_2{x}_2+...+w_n{x}_n+b$
$\vec{w}=[w_1{w}_2{w}_3...w_n]$
$b$ is a number
$\vec{w}$ and $b$ are parameters of the model.
$\vec{x}=[x_1{x}_2{x}_3...x_n]$
$f_{\vec{w},b}(\vec{x})=\vec{w}\cdot \vec{x}+b$
multiple linear regression 多元线性回归

9 Vectorization 向量化

Ex.
$\vec{w}=[w_1{w}_2{w}_3]$
$b$ is a number
$\vec{x}=[x_1{x}_2{x}_3]$
linear algebra: count from 1

w = np.array([1.0, 2.5, -3.3])
b = 4
x = np.array([10, 20, 30])

code: count from 0

WIthout vectorization
$f_{\vec{w},b}(\vec{x})=w_1{x}_1+w_2{x}_2+w_3{x}_3+b$

f = w[0] * x[0] + w[1] * x[1] + w[2] * x[2] + b

$f_{\vec{w},b}(\vec{x})={\sum }_{j=1}^n{w}_j{x}_j+b$

f = 0
for j in range(0, n):
	f = f + w[j] * x[j]
f = f + b

Vectorization
$f_{\vec{w},b}(\vec{x})=\vec{w}\cdot \vec{x}+b$

f = np.dot(w, x) + b

Two distinct benefits:
1.It makes the code shorter.
2.It results in your code running much faster.

Gradient descent
$\vec{w}=(w_1{w}_2...w_{16})$
derivatives: $\vec{d}=(d_1{d}_2...d_{16})$
(without $b$)

w = np.array([0.5, 1.3, ..., 3.4])
d = np.array([0.3, 0.2, ..., 0.4])

compute $w_j=w_j-0.1d_j$ for $j=\mathrm{1...16}$
Without vectorization
$w_1=w_1-0.1d_1$
$w_2=w_2-0.1d_2$
$⋮$
$w_{16}=w_{16}-0.1d_{16}$

for j in range(0, 16):
	w[j] = w[j] - 0.1 * d[j]

With vectorization
$\vec{w}=\vec{w}-0.1\vec{d}$

w = w - 0.1 * d

学习来源：吴恩达机器学习课程