用tensorflow实现线性回归问题

# -*- coding: UTF-8 -*-

import tensorflow as  tf
import numpy as  np

import  matplotlib.pyplot as plt

import os

os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'  #忽略警告



num_points = 1000#定义噪点的个数

vectors_set = []

for i  in range(num_points):#生成1000个随机点
    x1 = np.random.normal(0.0,0.55)
    y1 = x1 * 2 + 3 + np.random.normal(0.0,0.9)

    vectors_set.append([x1,y1])


x_data = [v[0] for v in vectors_set]#想数据
y_data = [v[1] for v in vectors_set]#y数据



W = tf.Variable(tf.random_uniform([1],-1.0,1.0),name = 'W')#定义权值W,范围在-1~1之间


b = tf.Variable(tf.zeros([1]),name = 'b')#定义偏置b，默认为0


y = W * x_data + b

#以均方差来作为损失
loss = tf.reduce_mean(tf.square(y-y_data),name = 'loss')
#用梯度下降法来训练，括号里的参数为学习率
optimizer = tf.train.GradientDescentOptimizer(0.5)
#训练过程就是最小化这个差值
train = optimizer.minimize(loss,name = 'train')
sess = tf.Session()
init = tf.global_variables_initializer()


sess.run(init)
print('W = ',sess.run(W),'    b=',sess.run(b),'  loss = ',sess.run(loss))


for i  in range(500):
    sess.run(train)
    print('W = ',sess.run(W),'    b=',sess.run(b),'  loss = ',sess.run(loss))

plt.figure()
plt.scatter(x_data,y_data,c='r')
plt.plot(x_data,sess.run(W)*x_data +  sess.run(b),'b')
plt.show()
sess.close()