tensorflow 案例

import tensorflow as tf
import numpy as np
#添加一层inputs输入的数据，in_size为输入节点数，out_size为输出节点数，下一个为激励函数
def add_layer(inputs,in_size,out_size,activation_function=None):
    Weights=tf.Variable(tf.random_normal([in_size,out_size]))   #权重
    biases=tf.Variable(tf.zeros([1,out_size]+0.1))              #偏移量
    Wx_plus_b=tf.matmul(inputs,Weights)+biases                  #计算公式
    if activation_function is None:                             #是否用激励函数
	outputs=Wx_plus_b
    else:
	outputs=activation_function(Wx_plus_b)
    return outputs

x_data=np.linspace(-1,1,300)[:,np.newaxis]                 #初始输入值
noise=np.random.normal(0,0.05,x_data.shape)                #干扰大小 计算
y_data=np.square(x_data)-0.5+noise                         #输出值

xs=tf.placeholder(tf.float32,[None,1])                      #输入占位符
ys=tf.placeholder(tf.float32,[None,1])                      #输出占位符

#l1=add_layer(x_data,1,10,activation_function=tf.nn.relu)
l1=add_layer(xs,1,10,activation_function=tf.nn.relu)        #添加一层中间计算层，使用激励函数
prediction=add_layer(l1,10,1,activation_function=None)      #添加输出层，

#loss 是估计值和真实值之映射到某一空间的误差
#loss=tf.reduce_mean(tf.reduce_sum(tf.square(y_data-predition),reduction_indices=[1]))
loss=tf.reduce_mean(tf.reduce_sum(tf.square(ys-prediction),reduction_indices=[1]))

train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)

init=tf.initialize_all_variable()
sess = tf.Session()
sess.run(init)

for i in range(1000):
    sess.run(train_step,feed_dict={xs:x_data,ys:y_data})
    if i % 50:
		print(sess.run(loss,feed_dict={xs:x_data,ys:y_data}))