In [98]:

    

import tensorflow as tf


    

In [99]:

    

import numpy as np


    

In [100]:

    

x_data = np.linspace(-1,1,300)[:,np.newaxis]


    

In [101]:

    

noise = np.random.normal(0,0.05,x_data.shape)


    

In [102]:

    

y_data = np.square(x_data) - 0.5 + noise


    

In [103]:

    

xs = tf.placeholder(tf.float32,[None,1])


    

In [104]:

    

ys = tf.placeholder(tf.float32,[None,1])


    

In [105]:

    

def add_layer(inputs,in_size,out_size,activation_function=None):
    weights = tf.Variable(tf.random_normal([in_size,out_size]))
    biases = tf.Variable(tf.zeros([1,out_size]) + 0.1)
    Wx_plus_b = tf.matmul(inputs,weights) + biases
    if activation_function is None:
        outputs = Wx_plus_b
    else:
        outputs = activation_function(Wx_plus_b)
    return outputs


    

In [106]:

    

hl = add_layer(xs,1,20,activation_function=tf.nn.relu)  # 构建隐藏层 假设隐藏层有10个神经元


    

In [107]:

    

predication = add_layer(hl,20,1,activation_function=None) # 构建输出层 假设输出层和输入层一样 有一个神经元


    

In [108]:

    

loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys - predication),reduction_indices=[1]))


    

In [109]:

    

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


    

In [110]:

    

init = tf.global_variables_initializer()  # 初始化所有变量


    

In [111]:

    

sess = tf.Session()


    

In [112]:

    

sess.run(init)


    

In [113]:

    

for i in range(1000):
    sess.run(train_step,feed_dict={xs:x_data,ys:y_data})
    if i % 50 == 0:
        print(sess.run(loss,feed_dict={xs:x_data,ys:y_data}))


    

    




1.45352
0.0117755
0.00750508
0.00554429
0.00462293
0.00419196
0.00385518
0.00360303
0.00345502
0.00335338
0.0032715
0.00321185
0.00316425
0.00313139
0.00310651
0.00308562
0.0030634
0.00304262
0.00302476
0.0029991