import tensorflow as tf
import numpy as np
#添加神经层
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
#设置输入数据
x_data = np.linspace(-1,1,300)[:,np.newaxis]
#设置噪音数据
noise = np.random.normal(0,0.05,x_data.shape)
#设置预期输出
y_data = np.square(x_data)-0.5 + noise
#设置传入变量
xs = tf.placeholder(tf.float32,[None,1])
ys = tf.placeholder(tf.float32,[None,1])
#第一层，隐藏层，1个输入，10个输出（10个神经元）
l1 = add_layer(xs,1,10,activation_function=tf.nn.relu)
#输出层，一个输出
prediction = add_layer(l1,10,1,activation_function=None)
#误差/代价
loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys-prediction),reduction_indices=[1]))
#最优化过程
train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)
init = tf.initialize_all_variables()
sess = tf.Session()
sess.run(init)
训练和输出
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})
