深度学习案例（二）：非线性回归案例

import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
#使用numpy生成200个随机点
x_data=np.linspace(-0.8,0.8,200)[:,np.newaxis]#200行1列矩阵
noise=np.random.normal(0,0.01,x_data.shape)#200行1列矩阵
y_data=np.square(x_data)+noise#200行1列矩阵

#定义两个placeholder
x=tf.placeholder(tf.float32,[None,1])#任意行1列矩阵
y=tf.placeholder(tf.float32,[None,1])#任意行1列矩阵

#定义神经网络中间层
Weights_L1=tf.Variable(tf.random_normal([1,10]))
biases_L1=tf.Variable(tf.zeros([1,10],tf.float32))
W_plus_b_L1=tf.matmul(x,Weights_L1)+biases_L1
L1=tf.nn.tanh(W_plus_b_L1)


#定义神经网络输出层
Weights_L2=tf.Variable(tf.random_normal([10,1]))
biases_L2=tf.Variable(tf.zeros([1,1]))
W_plus_b_L2=tf.matmul(L1,Weights_L2)+biases_L2
output=tf.tanh(W_plus_b_L2)

#二次代价函数
loss=tf.reduce_mean(tf.square(y-output))

#使用梯度下降法训练
train_step=tf.train.GradientDescentOptimizer(0.1).minimize(loss)

with tf.Session() as sess:
    init=tf.global_variables_initializer()
    sess.run(init)
    for i in range(5000):
        sess.run(train_step,feed_dict={x:x_data,y:y_data})
    #获得预测值
    output_value=sess.run(output,feed_dict={x:x_data})
    #画图
    plt.figure()
    plt.scatter(x_data,y_data)
    plt.plot(x_data,output_value,'b-',lw=5)
    plt.show()