3-1 Tensorflow --非线性回归-神经网络

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
import tensorflow as tf
import numpy
import matplotlib.pyplot as plt

# 生成200个随机样本点  作为X
# x_data = numpy.linspace(-0.5, 0.5, 200) # 从-0.5到0.5产生200个点，均匀分布的
# print(x_data)   # 这个样是一维的，我们线性回归需要二维的
x_data = numpy.linspace(-0.5, 0.5, 200)[:, numpy.newaxis]
# print(x_data)

# 构造噪音
noise = numpy.random.normal(loc=0, scale=0.02, size=x_data.shape)  # 符合高斯分布的值（loc:期望,scale:标准差   μ=0,σ=1正态分布）
# 构造Y
y_data = numpy.square(x_data) + noise

# 定义两个占位符placeholder
x = tf.placeholder(tf.float32, [None, 1])
y = tf.placeholder(tf.float32, [None, 1])

# 定义神经网络中间层
Weights_L1 = tf.Variable(tf.random_normal([1, 10])) # 权重
biases_L1 = tf.Variable(tf.zeros([1, 10])) # 偏置
Wx_plus_b_L1 = tf.matmul(x, Weights_L1) + biases_L1
L1 = tf.nn.tanh(Wx_plus_b_L1)

# 定义神经网络输出层
Weights_L2 = tf.Variable(tf.random_normal([10, 1])) # 权重
biases_L2 = tf.Variable(tf.zeros([1, 1])) # 偏置
Wx_plus_b_L2 = tf.matmul(L1, Weights_L2) + biases_L2
prediction = tf.nn.tanh(Wx_plus_b_L2)

# 代价函数
loss = tf.reduce_mean(tf.square(prediction - y))
# 梯度下降
train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)

with tf.Session() as sess:
    # 初始化变量
    sess.run(tf.global_variables_initializer())
    for _ in range(2000):
        sess.run(train_step, feed_dict={x:x_data, y:y_data})  
    # 活得预测值
    prediction_values = sess.run(prediction, feed_dict={x:x_data})
    
    # 画图
    plt.scatter(x_data, y_data)
    plt.plot(x_data, prediction_values, 'r-', lw=5) # r- 红色实线  lw 线宽
    plt.show()