1 # -*- coding: utf-8 -*-
 2 """
 3 Created on Sun Jul  2 19:59:43 2017
 4 
 5 @author: Administrator
 6 """
 7 import tensorflow as tf
 8 import numpy as np
 9 from tensorflow.examples.tutorials.mnist import input_data
10 
11 def compute_accuracy(v_xs,v_ys):
12     global prediction
13     y_pre = sess.run(prediction,feed_dict = {xs:v_xs,ys:v_ys,keep_prob: 1})
14     correct_prediction = tf.equal(tf.arg_max(y_pre,1),tf.arg_max(v_ys,1))
15     accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))
16     result = sess.run(accuracy,feed_dict = {xs:v_xs,ys:v_ys})
17     return result
18     
19 def getWeights(shape):
20     Weights = tf.Variable(tf.truncated_normal(shape,stddev = 0.1))
21     return Weights
22 def getBias(shape):
23     return tf.Variable(tf.constant(0.1,shape = shape))
24 
25 def conv2d(x,W):
26     return tf.nn.conv2d(x,W,strides= [1,1,1,1],padding = ‘SAME‘)
27 def maxpool(x):
28     return tf.nn.max_pool(x,ksize = [1,2,2,1],
29                           strides = [1,2,2,1],padding=‘SAME‘)
30 
31 mnist = input_data.read_data_sets("‘MNIST_data‘, one_hot=True")
32 xs = tf.placeholder(tf.float32,[None,28*28])
33 ys = tf.placeholder(tf.float32,[None,10])
34 
35 keep_prob = tf.placeholder(tf.float32)
36 x_image = tf.reshape(xs,[-1,28,28,1])
37 
38 
39 W_c1 = getWeights([5,5,1,32])
40 b_c1 = getBias([32])
41 h_c1 = tf.nn.relu(conv2d(x_image,W_c1)+b_c1)
42 h_p1 = maxpool(h_c1)
43 #这里注意的是maxpooling会将原来的28*28 变为14*14
44 
45 W_c2 = getWeights([5,5,32,64])
46 b_c2 = getBias([64])
47 h_c2 = tf.nn.relu(conv2d(h_p1,W_c2)+b_c2)
48 h_p2 = maxpool(h_c2)
49 #经过这次maxpooling，这时候将变7*7*64
50 
51 W_fc1 = getWeights([7*7*64,1024])
52 b_fc1 = getBias([1024])
53 h_p2_flat = tf.reshape(h_p2,[-1,7*7*64])
54 h_fc1 = tf.nn.relu(tf.matmul(h_p2_flat,W_fc1)+b_fc1)
55 h_fc1_drop = tf.nn.drop(h_fc1,keep_prob)
56 
57 W_fc2 = getWeights([1024,10])
58 b_fc2 = getBias(10)
59 prediction = tf.nn.relu(tf.matmul(h_fc1_drop,W_fc2)+b_fc2)
60 
61 cross_entropy = tf.reduce_mean(-tf.reduce_sum(ys*tf.log(prediction),
62                                               reduction_indices= [1]))
63 train_step = tf.train.AdamOptimizer(0.001).minimize(cross_entropy)
64 
65 with tf.Session() as sess:
66     sess.run(tf.initialize_all_variables())
67     for i in range(1000):
68         batch_xs,batch_ys = mnist.train.next_batch(100)
69         sess.run(train_step,feed_dict = {xs:batch_xs,ys:batch_ys,keep_prob:0.5})
70         if i % 50:
71             print (compute_accuracy( mnist.test.images,mnist.test.labels))
72