梯度下降代码

https://blog.csdn.net/panghaomingme/article/details/79384922
 1 #!usr/bin/python3
 2 # coding:utf-8
 3  
 4 # BGD 批梯度下降代码实现
 5 # SGD 随机梯度下降代码实现
 6 import numpy as np
 7  
 8 import random
 9  
10  
11 def batchGradientDescent(x, y, theta, alpha, m, maxInteration):
12     x_train = x.transpose()
13     for i in range(0, maxInteration):
14         hypothesis = np.dot(x, theta)
15         # 损失函数
16         loss = hypothesis - y
17         # 下降梯度
18         gradient = np.dot(x_train, loss) / m
19         # 求导之后得到theta
20         theta = theta - alpha * gradient
21     return theta
22  
23  
24 def stochasticGradientDescent(x, y, theta, alpha, m, maxInteration):
25     data = []
26     for i in range(4):
27         data.append(i)
28     x_train = x.transpose()
29     for i in range(0, maxInteration):
30         hypothesis = np.dot(x, theta)
31         # 损失函数
32         loss = hypothesis - y
33         # 选取一个随机数
34         index = random.sample(data, 1)
35         index1 = index[0]
36         # 下降梯度
37         gradient = loss[index1] * x[index1]
38         # 求导之后得到theta
39         theta = theta - alpha * gradient
40     return theta
41  
42  
43 def main():
44     trainData = np.array([[1, 4, 2], [2, 5, 3], [5, 1, 6], [4, 2, 8]])
45     trainLabel = np.array([19, 26, 19, 20])
46     print(trainData)
47     print(trainLabel)
48     m, n = np.shape(trainData)
49     theta = np.ones(n)
50     print(theta.shape)
51     maxInteration = 500
52     alpha = 0.01
53     theta1 = batchGradientDescent(trainData, trainLabel, theta, alpha, m, maxInteration)
54     print(theta1)
55     theta2 = stochasticGradientDescent(trainData, trainLabel, theta, alpha, m, maxInteration)
56     print(theta2)
57     return
58  
59  
60 if __name__ == "__main__":
61     main()