tensorflow初学代码系列一（基于莫烦视频）

（1）一次线性函数拟合

代码下载http://download.csdn.net/download/seven_year_promise/10126396

import tensorflow as tf
import numpy as np


#create data
x_data = np.random.rand(100).astype(np.float32)
y_data = x_data*0.1+0.3


""" structure """


Weights = tf.Variable(tf.random_uniform([1],-1.0,1.0))
biases=tf.Variable(tf.zeros([1]))


y=Weights*x_data+biases


Loss=tf.reduce_mean(tf.square(y-y_data))
optimizer=tf.train.GradientDescentOptimizer(0.5)
train=optimizer.minimize(Loss)


init = tf.initialize_all_variables()


""" end   """


sess = tf.Session()
sess.run(init)   #activate the structure


for step in range(201):
  sess.run(train)
  if step % 20 == 0:
     print(step,sess.run(Weights),sess.run(biases))
（2）变量学习

代码下载http://download.csdn.net/download/seven_year_promise/10126397

import tensorflow as tf
state = tf.Variable(0, name='counter')
print state.name


one = tf.constant(1)




new_value=tf.add(state, one)
update = tf.assign(state, new_value)


init = tf.initialize_all_variables()


with tf.Session() as sess:
   sess.run(init)
   for _ in range(3):
      print sess.run(update)
      print(sess.run(state))
      print state

（3）会话session

import tensorflow as tf
matrix_1=tf.constant([[3,3]])
matrix_2=tf.constant([[2],
                      [2]])
product = tf.matmul(matrix_1,matrix_2)


#method1
sess = tf.Session()
result = sess.run(product)
print result
sess.close()




#method2
with tf.Session() as sess:
   result2=sess.run(product)
   print result2

（4）placeholder

import tensorflow as tf
input_1 = tf.placeholder(tf.float32)
input_2 = tf.placeholder(tf.float32)


output = tf.multiply(input_1,input_2)


with tf.Session() as sess:
   print sess.run(output, feed_dict={input_1:[7.],input_2:[2.]})

（5）生成一个自己的网络结构

代码下载http://download.csdn.net/download/seven_year_promise/10126399

import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt


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
#data for training 
x_data = np.linspace(-1,1,300)[:, np.newaxis]
noise = np.random.normal(0,0.05,x_data.shape)
y_data = np.square(x_data) - 0.5 + noise
#data holder
xs = tf.placeholder(tf.float32,[None, 1])
ys = tf.placeholder(tf.float32,[None, 1])
#networks
l_1 = add_layer(xs,1,10,activation_function = tf.nn.relu)
prediction = add_layer(l_1,10,1,activation_function = None)
#loss
loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys - prediction), reduction_indices=[1]))
#train method
train_step = tf.train.GradientDescentOptimizer(0.01).minimize(loss)
#initialization
init = tf.initialize_all_variables()
#Session
sess = tf.Session()
sess.run(init)
#visulization
fig = plt.figure()
ax = fig.add_subplot(1,1,1)
ax.scatter(x_data, y_data)
plt.ion()
plt.show()


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}))


    try:
      ax.lines.remove(lines[0])
    except Exception:
      pass
    prediction_value = sess.run(prediction,feed_dict={xs:x_data})
    lines = ax.plot(x_data, prediction_value,'r-',lw=5)
    
    plt.pause(0.8)
    （6）tensorboard

代码下载http://download.csdn.net/download/seven_year_promise/10126400

import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt


def add_layer(inputs, in_size,out_size, n_layer, activation_function=None):
   layer_name = 'layer%s'% n_layer
   with tf.name_scope('layer_name'):
      with tf.name_scope('Weights'):
         Weights = tf.Variable(tf.random_normal([in_size, out_size]),name='W')
         tf.summary.histogram(layer_name+'/weights',Weights)
      with tf.name_scope('biases'):
         biases = tf.Variable(tf.zeros([1,out_size]) + 0.1, name='b' )
         tf.summary.histogram(layer_name+'/biases',biases)
      with tf.name_scope('Weights'):
         Wx_plus_b = tf.add(tf.matmul(inputs, Weights),biases)
   
      if activation_function is None:
         outputs = Wx_plus_b
      else:
         outputs = activation_function(Wx_plus_b)
         tf.summary.histogram(layer_name+'/outputs',outputs)
      return outputs


#data for training 
x_data = np.linspace(-1,1,300)[:, np.newaxis]
noise = np.random.normal(0,0.05,x_data.shape)
y_data = np.square(x_data) - 0.5 + noise


#data holder
with tf.name_scope('inputs'):
   xs = tf.placeholder(tf.float32,[None, 1],name='x_input')
   ys = tf.placeholder(tf.float32,[None, 1],name='y_input')


#networks
l_1 = add_layer(xs,1,10,n_layer=1,activation_function = tf.nn.relu)
prediction = add_layer(l_1,10,1,n_layer=2,activation_function = None)




#loss
with tf.name_scope('Loss'):
   loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys - prediction), reduction_indices=[1]), name='L')
   tf.summary.scalar('loss',loss)
#train method
with tf.name_scope('train'):
   train_step = tf.train.GradientDescentOptimizer(0.01).minimize(loss)




#initialization
#init = tf.initialize_all_variables()
#Session
sess = tf.Session()
merged = tf.summary.merge_all()
writer = tf.summary.FileWriter("logs/",sess.graph)
sess.run(tf.global_variables_initializer())


"""
#visulization
fig = plt.figure()
ax = fig.add_subplot(1,1,1)
ax.scatter(x_data, y_data)
plt.ion()
plt.show()
"""
for i in range(1000):
  sess.run(train_step, feed_dict={xs:x_data,ys:y_data})
  if i%50==0:
    result = sess.run(merged, feed_dict={xs:x_data,ys:y_data})
    writer.add_summary(result, i)