Tensorflow编程基础之Mnist手写识别实验+关于cross_entropy的理解

Posted 2021-01-19 kerwins-ac

好久没有静下心来写点东西了，最近好像又回到了高中时候的状态，休息不好，无法全心学习，恶性循环，现在终于调整的好一点了，听着纯音乐突然非常伤感，那些曾经快乐的大学时光啊，突然又慢慢的一下子出现在了眼前，不知道我大学的那些小伙伴们现在都怎么样了，考研的刚刚希望他考上，实习的菜头希望他早日脱离苦海，小瑞哥希望他早日出成果，范爷熊健研究生一定要过的开心啊！天哥也哥早日结婚领证！那些回不去的曾经的快乐的时光，你们都还好吗！

 

最近开始接触Tensorflow，可能是论文里用的是这个框架吧，其实我还是觉得pytorch更方便好用一些，仔细读了最简单的Mnist手写识别程序，觉得大同小异，关键要理解Tensorflow的思想，文末就写一下自己看交叉熵的感悟，絮叨了这么多开始写点代码吧！  1 # -*- coding: utf-8 -*-

  2 """
  3 Created on Sun Nov 11 16:14:38 2018
  4 
  5 @author: Yang
  6 """
  7 
  8 import tensorflow as tf 
  9 from tensorflow.examples.tutorials.mnist import input_data 
 10 
 11 mnist = input_data.read_data_sets("/MNIST_data",one_hot=True) #从input_data中读取数据集，使用one_hot编码
 12 
 13 import pylab #画图模块
 14 
 15 tf.reset_default_graph()#重置一下图 图代表了一个运算过程，包含了许多Variable和op，如果不重置一下图的话，可能会因为某些工具重复调用变量而报错
 16 
 17 x = tf.placeholder(tf.float32,[None,784])#占位符，方便用feed_dict进行注入操作
 18 y = tf.placeholder(tf.float32,[None,10])#占位符，方便用feed_dict进行注入操作

 20 
 21 W = tf.Variable(tf.random_normal([784,10]))#要学习的参数统一用Variable来定义，这样方便进行调整更新
 22 b = tf.Variable(tf.zeros([10]))
 23 
 24 
 25 #construct the model 
 26 pred = tf.nn.softmax(tf.matmul(x,W) + b)
 27 
 28 cost = tf.reduce_mean(-tf.reduce_sum(y*tf.log(pred),reduction_indices=1))
 29 
 30 learning_rate = 0.01
 31 
 32 optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)
 33 
 34 #set parameters about thee model 
 35 training_epoch = 25 
 36 batch_size = 100
 37 display_step = 1
 38 saver = tf.train.Saver()
 39 model_path = "log/kerwinsmodel.ckpt"
 40 
 41 #start the session 
 42 
 43  with tf.Session() as sess :
 44      sess.run(tf.global_variables_initializer())
 45      
 46      for epoch in range(training_epoch):
 47          avg_cost = 0
 48          total_batch = int(mnist.train.num_examples/batch_size)
 49          print(total_batch)
 50          for  i in range(total_batch):
 51              batch_xs,batch_ys = mnist.train.next_batch(batch_size)
 52              
 53              _,c = sess.run([optimizer,cost],feed_dict={x:batch_xs,y:batch_ys})
 54              
 55              avg_cost += c/ total_batch
 56          if (epoch +1 ) % display_step ==0:
 57              print("Epoch:",‘%04d‘ %(epoch+1),"cost=","{:.9f}".format(avg_cost))
 58              
 59      print("Finish!")
 60      
 61      correct_prediction = tf.equal(tf.argmax(pred,1),tf.argmax(y,1))
 62      accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))
 63      print("Accuracy:",accuracy.eval({x:mnist.test.images,y:mnist.test.labels}))
 64      
 65      save_path = saver.save(sess,model_path)
 66      print("Model saved in file: %s" % save_path)
 67  #    
 68 
 69     
 70 #读取模型程序
 71 
 72 print("Starting 2nd session...")
 73 with tf.Session() as sess:
 74     sess.run(tf.global_variables_initializer())
 75     saver.restore(sess,model_path)
 76     
 77     #测试model
 78     correct_prediction = tf.equal(tf.arg_max(pred,1),tf.argmax(y,1))
 79     #计算准确率
 80     accuracy = tf.reduce_mean(tf.cast(correct_prediction,tf.float32))
 81     print("Accuracy:",accuracy.eval({x:mnist.test.images,y:mnist.test.labels}))
 82     
 83     output = tf.argmax(pred,1)
 84     batch_xs,batch_ys = mnist.train.next_batch(2)
 85     outputval,predv = sess.run([output,pred],feed_dict={x:batch_xs})
 86     print(outputval,predv,batch_ys)
 87     
 88     im = batch_xs[0]
 89     im = im.reshape(-1,28)
 90     pylab.imshow(im)
 91     pylab.show()
 92     
 93     im = batch_xs[1]
 94     im = im.reshape(-1,28)
 95     pylab.imshow(im)
 96     pylab.show()
 97     
 98     
 99     
100         

 

#占位符，方便用feed_dict进行注入操作