1 import tensorflow as tf
2 import numpy as np
3 import matplotlib.pyplot as plt
4
5 #使用numpy生成200个随机点
6 x_data = np.linspace(-0.5,0.5,200)[:,np.newaxis]
7 noise = np.random.normal(0,0.02,x_data.shape)
8 y_data = np.square(x_data) + noise
9
10 #定义两个placeholder
11 x = tf.placeholder(tf.float32,[None,1])
12 y = tf.placeholder(tf.float32,[None,1])
13
14 #输入层1个神经元节点，中间层10个神经元节点，输出层1个神经元节点
15 #定义神经网络中间层
16 Weights_L1 = tf.Variable(tf.random_normal([1,10]))
17 biases_L1 = tf.Variable(tf.zeros([1,10]))
18 Wx_plus_b_L1 = tf.matmul(x,Weights_L1) + biases_L1
19 L1 = tf.nn.tanh(Wx_plus_b_L1)
20
21 #定义神经网络输出层
22 Weight_L2 = tf.Variable(tf.random_normal([10,1]))
23 biases_L2 = tf.Variable(tf.zeros([1,1]))
24 Wx_plus_b_L2 = tf.matmul(L1,Weight_L2) + biases_L2
25 prediction = tf.nn.tanh(Wx_plus_b_L2)
26
27 #二次代价函数
28 loss= tf.reduce_mean(tf.square(y-prediction))
29 #使用梯度下降法训练网络
30 train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)
31
32 with tf.Session() as sess:
33 #变量初始化
34 sess.run(tf.global_variables_initializer())
35 for _ in range(2000):
36 sess.run(train_step,feed_dict={x:x_data,y:y_data})
37 #print(sess.run(Weights_L1))
38 #获得预测值
39 prediction_value = sess.run(prediction,feed_dict={x:x_data})
40 #画图
41 plt.figure()
42 plt.scatter(x_data,y_data)
43 plt.plot(x_data,prediction_value,'r-',lw=5)
44 plt.show()
2019-05-30 10:58:12