RNN之NLSTM递归神经网络

Halcom

RNN之NLSTM递归神经网络
百度网盘代码分享：http://pan.baidu.com/s/1eRQcCuu
电脑：Win7旗舰版+64Bit+AMD Athlon(tm)X2 DualCore QL-64 2.10GHz   RAM2.75GB
Anaconda3-4.2.0-Windows-x86_64

1. import time
   
2. #from tensorflow.examples.tutorials.mnist import input_data
   
3. import tensorflow as tf
   
4. import Get_Mnist_Data
   
5. 
   
6. start=time.clock()
   
7. #mnist = input_data.read_data_sets('/temp/', one_hot=True)
   
8. mnist = Get_Mnist_Data.read_data_sets('Get_Mnist_Data', one_hot=True)
   
9. end=time.clock()  
   
10. print('Runing time = %s Seconds'%(end-start))
    
11. 
    
12. def compute_accuracy(v_x, v_y):
    
13.     global pred
    
14.     #input v_x to nn and get the result with y_pre
    
15.     y_pre = sess.run(pred, feed_dict={x:v_x})
    
16.     #find how many right
    
17.     correct_prediction = tf.equal(tf.argmax(y_pre,1), tf.argmax(v_y,1))
    
18.     #calculate average
    
19.     accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
    
20.     #get input content
    
21.     result = sess.run(accuracy,feed_dict={x: v_x, y: v_y})
    
22.     return result
    
23. 
    
24. def LSTM_cell():
    
25.     return tf.contrib.rnn.BasicLSTMCell(n_hidden_units, forget_bias=1.0, state_is_tuple=True)
    
26. 
    
27. def Drop_lstm_cell():
    
28.     return tf.contrib.rnn.DropoutWrapper(LSTM_cell(), output_keep_prob=0.5)
    
29. 
    
30. def Mul_lstm_cell():
    
31.     return tf.contrib.rnn.MultiRNNCell([Drop_lstm_cell() for _ in range(lstm_layer)], state_is_tuple=True)
    
32. 
    
33. def RNN(X,weights,biases):
    
34.     # hidden layer for input
    
35.     X = tf.reshape(X, [-1, n_inputs])
    
36.     X_in = tf.matmul(X, weights['in']) + biases['in']
    
37.     X_in = tf.reshape(X_in, [-1,n_steps, n_hidden_units])
    
38.    
    
39.     # cell
    
40.     #lstm_cell = tf.contrib.rnn.BasicLSTMCell(n_hidden_units, forget_bias=1.0, state_is_tuple=True)
    
41.     lstm_cell = Mul_lstm_cell()
    
42.     _init_state = lstm_cell.zero_state(batch_size, dtype=tf.float32)
    
43.     outputs,states = tf.nn.dynamic_rnn(lstm_cell, X_in, initial_state=_init_state, time_major=False)
    
44.    
    
45.     #hidden layer for output as the final results
    
46.     #results = tf.matmul(states[2][1], weights['out']) + biases['out']
    
47.     # or
    
48.     outputs = tf.unstack(tf.transpose(outputs, [1,0,2]))
    
49.     results = tf.matmul(outputs[-1], weights['out']) + biases['out']
    
50. 
    
51.     return results
    
52.    
    
53. 
    
54. #load mnist data
    
55. #mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
    
56. 
    
57. # parameters init
    
58. lstm_layer = 3
    
59. l_r = 0.001
    
60. training_iters = 100
    
61. batch_size = 128
    
62. 
    
63. n_inputs = 28
    
64. n_steps = 28
    
65. n_hidden_units = 128
    
66. n_classes = 10
    
67. 
    
68. #define placeholder for input
    
69. x = tf.placeholder(tf.float32, [None, n_steps, n_inputs])
    
70. y = tf.placeholder(tf.float32, [None, n_classes])
    
71. 
    
72. # define w and b
    
73. weights = {
    
74.     'in': tf.Variable(tf.random_normal([n_inputs,n_hidden_units])),
    
75.     'out': tf.Variable(tf.random_normal([n_hidden_units,n_classes]))
    
76. }
    
77. biases = {
    
78.     'in': tf.Variable(tf.constant(0.1,shape=[n_hidden_units,])),
    
79.     'out': tf.Variable(tf.constant(0.1,shape=[n_classes,]))
    
80. }
    
81. 
    
82. pred = RNN(x, weights, biases)
    
83. cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=pred,labels=y))
    
84. train_op = tf.train.AdamOptimizer(l_r).minimize(cost)
    
85. 
    
86. correct_pred = tf.equal(tf.argmax(pred,1),tf.argmax(y,1))
    
87. accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
    
88. 
    
89. #init session
    
90. sess = tf.Session()
    
91. #init all variables
    
92. sess.run(tf.global_variables_initializer())
    
93. #start training
    
94. 
    
95. # x_image,x_label = mnist.test.next_batch(500)
    
96. # x_image = x_image.reshape([500, n_steps, n_inputs])
    
97. 
    
98. for i in range(training_iters):
    
99.     #get batch to learn easily
    
100.     batch_x, batch_y = mnist.train.next_batch(batch_size)
     
101.     batch_x = batch_x.reshape([batch_size, n_steps, n_inputs])
     
102.     sess.run(train_op,feed_dict={x: batch_x, y: batch_y})
     
103.     if i % 50 == 0:
     
104.         print(sess.run(accuracy,feed_dict={x: batch_x, y: batch_y,}))
     
105.       #  print(sess.run(accuracy,feed_dict={x: x_image, y: x_label}))
     
106. sess.close()
     

复制代码