使用tensorflow实现最简单的线性回归算法

 1         #线性回归：用线性模型y=Wx+b拟合sin
 2         import numpy as np
 3         import matplotlib.pyplot as plt
 4         import tensorflow as tf
 5         
 6         #数据，标签
 7         x_data  = np.linspace(-2*np.pi,2*np.pi,300)
 8         noise = np.random.normal(-0.01,0.05,x_data.shape)
 9         y_label = np.sin(x_data) + noise
10         plt.rcParams['font.sans-serif']=['FangSong'] # 用来正常显示中文标签
11         plt.rcParams['axes.unicode_minus']=False# 用来正常显示负号
12         plt.title('线性模型y=Wx+b拟合sin')
13         plt.legend()
14         plt.grid(True)
15         plt.plot(x_data, y_label, 'b.', label='测试数据')
16         
17         #静态图定义
18         mg = tf.Graph()
19         with mg.as_default():
20             #图输入
21             X = tf.placeholder("float")
22             Y = tf.placeholder("float")
23             
24             #训练权重w,b
25             W = tf.Variable(np.random.randn(), name="weight")
26             b = tf.Variable(np.random.randn(), name="bias")
27             
28             #线性模型y=Wx+b
29             pred = tf.add(tf.multiply(X, W), b)
30             
31             #损失函数:使用样本方差
32             cost = tf.reduce_sum(tf.pow(pred-Y, 2)) / (len(x_data)-1)
33             
34             #使用梯度下降法优化
35             optimizer = tf.train.GradientDescentOptimizer(0.01).minimize(cost)
36             
37             #初始化图变量
38             init = tf.group(tf.global_variables_initializer(),
39                             tf.local_variables_initializer())
40             
41         with tf.Session(graph=mg) as sess:
42             sess.run(init)
43             for epoch in range(500):
44                 for (x, y) in zip(x_data, y_label):
45                     sess.run(optimizer, feed_dict={X: x, Y: y})
46                 if (epoch+1) % 10 == 0:
47                     c = sess.run(cost, feed_dict={X: x_data, Y:y_label})
48                     print('epoch={:} cost={:0.6f} W={:0.6f} b={:0.6f}'.format(epoch+1,c,sess.run(W),sess.run(b)))
49             training_cost = sess.run(cost, feed_dict={X: x_data, Y: y_label})
50             print('训练结果 cost={:0.6f} W={:0.6f} b={:0.6f}'.format(training_cost,sess.run(W),sess.run(b)))
51             plt.plot(x_data, sess.run(W) * x_data + sess.run(b),'r--' , label='拟合数据')
52             plt.legend()
53         plt.show()

---