1 import numpy as np
2 import pylab
3
4
5 def plot_data(data, b, m):
6 x = data[:, 0]
7 y = data[:, 1]
8 y_predict = m*x + b
9 pylab.plot(x, y_predict, 'k-')
10 pylab.plot(x, y, 'o')
11 pylab.show()
12
13
14 def gradient(data, initial_b, initial_m, learning_rate, num_iter):
15 b = initial_b
16 m = initial_m
17 x = data[:, 0]
18 y = data[:, 1]
19 n = float(len(data))
20 for i in range(num_iter):
21 b_gradient = -(1/n)*(y - m*x-b)
22 b_gradient = np.sum(b_gradient, axis=0)
23 m_gradient = -(1/n)*x*(y - m*x - b)
24 m_gradient = np.sum(m_gradient)
25 theta0 = b - (learning_rate*b_gradient)
26 theta1 = m - (learning_rate*m_gradient)
27 b = theta0
28 m = theta1
29 if i % 100 == 0:
30 j = (np.sum((y - m*x - b)**2))/n
31 print("参数b:{},参数m:{},损失值：{}".format(b, m, j))
32 return [b, m]
33
34
35 def linear_regression():
36 # 导入数据
37 data = np.loadtxt('data.csv', delimiter=',')
38
39 # 线性回归参数设定
40 learning_rate = 0.001
41 initial_b = 0.0
42 initial_m = 0.0
43 num_iter = 1000
44
45 b, m = gradient(data, initial_b, initial_m, learning_rate, num_iter)
46 plot_data(data, b, m)
47
48 if __name__ == '__main__':
49 linear_regression()
测试结果：
