import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
#定义层,add_layer(输入数据,输入数据维数,输出数据维数,激励函数)
def add_layer(inputs,in_size,out_size,n_layer,activate_function=None):
layer_name = 'layer%s'%n_layer
with tf.name_scope(layer_name): #层名称
with tf.name_scope('weights'):
Weights = tf.Variable(tf.random_normal([in_size,out_size]),name='W')
tf.summary.histogram(layer_name+'/weights',Weights) #画图
with tf.name_scope('biases'):
bias = tf.Variable(tf.zeros([1,out_size])+0.1,name='b')
tf.summary.histogram(layer_name+'/biases',bias)
with tf.name_scope('Wx_plus_b'):
Wx_plus_b = tf.add(tf.matmul(inputs,Weights),bias)
if activate_function is None:
outputs = Wx_plus_b
else:
outputs = activate_function(Wx_plus_b)
tf.summary.histogram(layer_name+'/outputs',outputs)
return outputs
#自定义数据集, [:,np.newaxis]可以生成shape=(300,1)的张量;[np.newaxis:,]可以生成shape=(1,300)的张量
x_data = np.linspace(-1,1,300)[:,np.newaxis]
noise = np.random.normal(0,0.05,x_data.shape) #噪声
y_data = np.square(x_data) - 0.5 + noise
with tf.name_scope('inputs'): #可视化图输入名称
xs = tf.placeholder(tf.float32,[None,1],name='x_input')
ys = tf.placeholder(tf.float32,[None,1],name='y_input')
#添加隐藏层
l1 = add_layer(xs,1,10,n_layer=1,activate_function=tf.nn.relu)
#添加输出层
prediction = add_layer(l1,10,1,n_layer=2,activate_function = None)
#reduction_indices=[1]将其维数降到一维,即将其相加求和
with tf.name_scope('loss'):
loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys-prediction),reduction_indices=[1]))
tf.summary.scalar('loss',loss) #loss的收敛图使用tf.summary.scalar
with tf.name_scope('train'):
train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)
#初始化变量variable
init = tf.global_variables_initializer()
sess = tf.Session()
sess.run(init)
merged=tf.summary.merge_all()
writer = tf.summary.FileWriter('logs/',sess.graph) #mac环境下,logs路径为Users/用户名/logs
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}))
result = sess.run(merged,
feed_dict = {xs:x_data,ys:y_data})
writer.add_summary(result,i)
tips:tf.summary.scalar(‘loss’,loss) #loss的收敛图使用此方法
tf.summary.histogram(layer_name+’/weights’,Weights) #画图,其他数据用此方法