tensorflow实现简单的卷积神经网络

 1 # MNIST训练
 2 
 3 import tensorflow as tf
 4 import matplotlib.pyplot as plt
 5 from tensorflow.examples.tutorials.mnist import input_data
 6 import numpy as np
 7 
 8 mnist = input_data.read_data_sets('MNIST_data/',one_hot=True)
 9 
10 def weight_variable(shape):
11     initial = tf.truncated_normal(shape,stddev=0.1)
12     return tf.Variable(initial)
13 
14 def bias_variable(shape):
15     initial = tf.constant(0.1,shape=shape)
16     return tf.Variable(initial)
17 
18 def conv(x,w):
19     return tf.nn.conv2d(x,w,strides=[1,1,1,1], padding='SAME')
20 
21 def max_pool(x):
22     return tf.nn.max_pool(x,ksize = [1, 2, 2, 1],strides=[1,2,2,1],padding='SAME')
23 
24 x = tf.placeholder(tf.float32,shape=[None,784])
25 y_ = tf.placeholder(tf.float32,shape=[None,10])
26 x_image = tf.reshape(x,[-1,28,28,1])
27 
28 #卷积层1-池化层1
29 w_conv1 = weight_variable([5,5,1,32])
30 b_conv1 = bias_variable([32])
31 h_conv1 = tf.nn.relu(conv(x_image,w_conv1)+b_conv1)
32 h_pool1 = max_pool(h_conv1)
33 
34 #卷积层2-池化层2
35 w_conv2 = weight_variable([5,5,32,64])
36 b_conv2 = bias_variable([64])
37 h_conv2 = tf.nn.relu(conv(h_pool1,w_conv2)+b_conv2)
38 h_pool2 = max_pool(h_conv2)
39 
40 #全连接层
41 w_fc1 = weight_variable([7 * 7 *64,1024])
42 b_fc1 = bias_variable([1024])
43 h_pool_flat = tf.reshape(h_pool2,[-1,7 * 7 *64])
44 h_fc1 = tf.nn.relu(tf.matmul(h_pool_flat,w_fc1)+b_fc1)
45 
46 #dropout层
47 keep_drop = tf.placeholder(tf.float32)
48 h_fc1_drop = tf.nn.dropout(h_fc1,keep_drop)
49 
50 #softmax层
51 w_fc2 = weight_variable([1024,10])
52 b_fc2 = bias_variable([10])
53 y = tf.matmul(h_fc1_drop,w_fc2)+b_fc2
54 
55 #loss
56 loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=y,labels=y_))
57 train_step = tf.train.AdamOptimizer(0.0001).minimize(loss)
58 #计算模型预测的准确率
59 correct_prediction = tf.equal(tf.argmax(y,1),tf.argmax(y_,1))
60 accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))
61 
62 sess = tf.InteractiveSession()
63 init = tf.global_variables_initializer()
64 sess.run(init)
65 losses = []
66 acc = []
67 for i in range(2000):
68     batch = mnist.train.next_batch(50)
69     if i % 100 == 0:
70         train_accuracy = accuracy.eval(feed_dict={x:batch[0],y_:batch[1],keep_drop:1.0})
71         print('step %d,training accuracy %g' %(i,train_accuracy))
72         acc.append(train_accuracy)
73         loss_tmp = sess.run(loss,feed_dict={x:batch[0],y_:batch[1],keep_drop:1.0})
74         losses.append(loss_tmp)
75     sess.run(train_step,feed_dict={x: batch[0], y_: batch[1], keep_drop: 0.5})
76 print("test accuracy",accuracy.eval(feed_dict={x:mnist.test.images,y_:mnist.test.labels,keep_drop:1.0}))
参考文章：
1.https://www.cnblogs.com/willnote/p/6874699.html
作者：舟华520

出处：https://www.cnblogs.com/xfzh193/

本文以学习，分享，研究交流为主，欢迎转载，请标明作者出处！